3. Practical Work: SEM Search Engine Marketing - Applying SEO Techniques to Improve Website Rankings
4. Practical Work: Using GA4 (Google Analytics) for E-commerce Traffic Analysis
5. Implementation: Using A/B Testing for E-commerce Design Decisions and Analyzing Advertising Effectiveness
5. Practical Work: Using GTM (Google Tag Manager) for Customer Behavior Tracking
6. Practical Work: Using Google Looker Studio for Marketing Data Visualization
2.【GA4 Self-Learning Platform: Google Analytics Learning Center Courses】：
16. Google provides GA Learning Center Courses:
101: Getting Started with Google Analytics 4 (GA4) (Beginner)
102: Managing GA4 Data & Understanding Reports (Beginner)
201: Deep Dive into GA4 Data & Reports (Intermediate)
301: Integrating GA4 with Other Tools & Data Sources (Advanced) 【GA Learning Center Course URL】
3.【GA Certification Exam URL】：
Google offers GA certification exams for free, allowing multiple attempts online. The exam consists of 70 multiple-choice questions with a 60-minute time limit. A score of 80% or higher is required to pass.
A GA certification is highly beneficial for marketing professionals. 【GA4 Certification Exam - Certification Center URL】
4.【Various GA4 Practice Question Banks & Past Exam Questions】：
2. Haran’s Marketing Notes:
Google Analytics 4 Official Certification & Latest Past Exam Questions (50 Questions): Detailed breakdown of 50 official GA4 certification questions with explanations.
◉【Haran's Marketing Notes: Past Exam Questions】
1. Automated Generation:
2. Diverse Styles:
3. Personalized Customization:
4. Lower Creation Barriers:
5. Usage Scenarios: Presentation illustrations, short video illustrations...
6. Advantages: Google product, stable website, no copyright issues, quick and easy
(2).【Google AutoDraw】 Illustration Tool:
1.【Website】AutoDraw
2. Features:
【Automatic Recognition & Guessing Drawn Images】: AutoDraw's core feature is intelligent recognition. When you sketch freely on the canvas, the system automatically identifies what you are drawing and provides multiple precise suggestions.
1. LDM (Latent Diffusion Model):
Feature 1: LDM is a latent variable diffusion model based on diffusion processes.
Feature 2: Image generation speed is slow, at least 30 seconds.
2. LCM (Latent Consistency Model):
Feature 1: LCM emphasizes maintaining consistency in latent space for stable image generation.
Feature 2: Image generation speed is fast.
(2).【Krea】 AI Illustration Tool: Text-to-Image, Photorealistic Images, Free
1. 【Website】Krea
2. Feature 1: Requires registration via Gmail account
3. Feature 2: Free (with a daily AI usage limit of 3 minutes), no copyright issues
4. Feature 3: 【Text-to-Image】 Enter an English phrase to generate an image
5. Feature 4: 【Free PNG Downloads】
6. Feature 5: Krea’s 4 Membership Tiers & Pricing
7. Feature 6: Krea uses LCM AI technology, making image generation faster.
(1). Use 【AI Image Generator】 ➜ 【Sketch to Image】 mode.
(2). Text-to-Image + Shape-Guided Image Generation.
Example: Pikachu.
(3). White background.
(4). Adjust AI Imagination Strength.
(5). 【Take a screenshot / or download PNG image】 and upload to Zuvio.
(6). **Note:** Free members only have 3 minutes of AI computation per day.
(15). Goal: Draw NTUT’s mark along with book and computer icons
(16). Select 【Image】
(17). Choose 【>】 ➜ My Style ➜ Recraft v3 ➜ 【Icon】
(18). Enter the English description: "Draw the logo of NTUT and the logo of books and computer"
(19). 【Generate Image】 - Icon
(20). 【Export】 ➜ 【PNG】
(21). 【Download PNG Icon】
(22). 【Application】: Use Recraft to quickly create icons for presentations and reports
4. 【Assignment Submission】:
(1). Use 【Image】 ➜ 【Vector Art】
(2). Goal: Draw a vector logo for your department
(3). 【Download PNG Image】 and upload it to Zuvio
◉The drawing function of ChatGPT 4o is supported by DALL-E 3 technology
◉Note 1: The free version of ChatGPT 4o does not support drawing; only the paid version offers this feature
◉Note 2: ChatGPT's text-to-image works well for physical objects but performs poorly when generating logos
◉Gemini currently (2024) offers drawing capabilities
◉Copilot currently (2024) offers drawing capabilities
◉Three-Round Comparison:
Round 1: Compare which generated image most resembles a real photo
Round 2: Compare which generated image can create a logo pattern
Round 2: Compare which generated image can create celebrity photos with arbitrary designs (not sourced from internet photos)
(2).【Round 1 Comparison】: Generate an image, a dog playing with a ball in the park in the afternoon
(3).【Round 2 Comparison】: Generate an image, draw the logo of The Department of Business Management, NTUT
draw the logo of The Department of Business Management, NTUT
◉Three-Round Comparison Conclusion:
Round 1: Compare which generated image most resembles a real photo (Grok > Gemini > ChatGPT, Copilot)
Round 2: Compare which generated image can create a logo pattern (Gemini > Grok > ChatGPT, Copilot)
Round 2: Compare which generated image can create celebrity photos with arbitrary designs (not sourced from internet photos) (Grok > Gemini, ChatGPT)
(7).【Submit Assignment】:
(1). Use 【ChatGPT or Copilot's drawing function】
(2). Goal: Text-to-image, any object image
(3). 【Download PNG image】, upload to zuvio
2. Feature 1: Direct Login (using Gmail account)
3. Feature 2: Free (but with a daily limit of 10,000 credits), no copyright issues
4.Illustration: Claim daily 10,000 credits
5. Feature 3: Software has a Chinese interface (but prompts must still be entered in Chinese)
(2). Drawing: Use 【Recommended/Popular/Daily Rankings/Most Popular】 to load and modify existing 2D characters created by others
◉(F). Set a new 【Model】 and configure a new 【LoRA】
Copy the referenced English prompt,
1 lady, long hair, brown hair, off shoulder, white shirt, simple background, Character sheet, clothes sheet Set a new 【Model】: Daydream
Set a new 【LoRA】: Realistic Enhancer (Realistic Enhancement)
Remove the old 【LoRA】, as the old LoRA is incompatible with the 【Daydream Model】
Output Result Illustration Result Image
Set a new 【LoRA】: Chocoskin (Chocolate-colored skin)
Remove the old 【LoRA】, as the old LoRA is incompatible with the 【Daydream Model】
Output Result Illustration Result Image
(3). Explanation: The three elements of generating images 【Prompt, Model, LoRA】
(A). Simple Metaphor 1: Imagine ordering food:
◉Prompt: Your order to the waiter ("I want a steak with red wine").
◉Model: The kitchen's basic menu (standard Western cuisine preparation).
◉LoRA: Special seasoning (e.g., adding truffle or a specific sauce).
(B). Similar to Krea AI Drawing:
◉Prompt: Similar to inputting Garfield
◉Model: Similar to selecting a Style from the Gallery
◉LoRA: Similar to adjusting the intensity of the Style (0 to 0.99)
(C). Simple Metaphor 3: These three are like the three parts of a concert
◉Prompt: The "score," telling the AI what to play (theme and scene).
◉Model: The "instrument," determining the tone and quality (basic style and capability).
◉LoRA: LoRA is the "special effect," adding character to the performance (specific style or detail).
(4). Explanation: 【Special Prompt Type 1】: Double brackets ((...)), Single brackets (...)
(A). Return to the homepage, view 【Recommended/Popular/Daily Rankings/Most Popular】 to load and modify existing 2D characters created by others:
Illustration 1,
(B). Observe an example:
Illustration 1,
◉Prompt:
1girl, ((19th century)), young, ((arms crossed)), ((pouty)), ((dark gray dress)), ((young)), ((13 years old)), ((petite body)), ((dark beige hair)), ((standing)), ((gray eyes)), ((cute)), ((ruffled white skirt)), wavy hair, ((black ribbon in hair)), ((Medium length hair)), ((barmaid dress)), ((shoulder length hair)), ((ruffled white sleeves)), ((19th century apron)), ((slender)), narrow hips, narrow shoulders, ((modest)), ((small Bavarian village)), short sleeves,
◉Chinese Translation: 1 girl, ((19th century)), young, ((arms crossed)), ((pouty)), ((dark gray dress)), ((young)), ((13 years old)), ((petite body)), ((dark beige hair)), ((standing)), ((gray eyes)), ((cute)), ((ruffled white skirt)), wavy hair, ((black ribbon in hair)), ((Medium length hair)), ((barmaid dress)), ((shoulder length hair)), ((ruffled white sleeves)), ((19th century apron)), ((slender)), narrow hips, narrow shoulders, ((modest)), ((small Bavarian village)), short sleeves.
(C). What are (()) and ()?
◉Explanation: (), Single brackets: Single brackets () indicate a standard weight (approximately 1.0),
◉Explanation: Double brackets ((...)), usually increase the influence of the term by about 1.1 to 1.3 times (the exact multiplier varies by tool, PixAI typically defaults to 1.2 times).
◉Explanation: (()), Double brackets: Used to emphasize the importance of a term or phrase.
(D). What is ((19th century))?
◉Explanation: ((19th century)) is a way of expressing weight control (Weight Adjustment) in AI image generation tools (e.g., PixAI or Stable Diffusion) to adjust the influence of specific words or phrases.
For example, ((19th century)) more strongly instructs the AI to align the image style or elements with 19th-century characteristics (e.g., clothing, scenes, or atmosphere) compared to just 19th century.
(5). Explanation: 【Special Prompt Type 2】: Double brackets ((...)), Single brackets (...)
(A). Return to the homepage, view 【Recommended/Popular/Daily Rankings/Most Popular】 to load and modify existing 2D characters created by others:
Illustration 1,
(B). Observe an example:
Illustration 1,
◉Prompt:
(score_9, score_8_up, score_7_up), headphones, cap, sitting a bench, training wear, (ultra detailed perfect piece:1.2), illustration, masterpiece, (extremely detailed CG 8k), (very fine 8K CG), (1girl:1.2), (red hair), middle hair, wavy hair, hair over one eye, sparkling, red eyes,, ◉Chinese Translation: Headphones, cap, sitting on a bench, training wear, (ultra detailed perfect piece:1.2), illustration, masterpiece, (extremely detailed 8K CG), (very fine 8K CG), (1girl:1.2), (red hair), medium-length hair, wavy hair, hair over one eye, sparkling, red eyes, lora:more_details:0.5.
(C). Explanation of Special Prompt
◉Prompt: (ultra detailed perfect piece:1.2)
Meaning: "Ultra detailed perfect piece" is a quality description, instructing the AI to generate a highly artistic, detailed image close to "masterpiece" level.
Meaning: :1.2 is the weight value, increasing the influence of this term by 20% compared to the standard weight (1.0), telling the AI to prioritize ultra-high detail and perfection.
(D). Prompt: (extremely detailed 8K CG)
Meaning: "Extremely detailed 8K CG" refers to generating a very high-resolution (8K, approximately 7680x4320 pixels) computer-generated image.
Meaning: No explicit weight inside the brackets, default weight is approximately 1.0,
Meaning: But double brackets ((...)) (if present) would increase it to 1.2 times.
(E). Prompt: (1girl:1.2)
Meaning: "1girl" specifies generating a single female character.
Meaning: :1.2 indicates a weight of 1.2, increasing influence by 20%, emphasizing this as the main subject to avoid generating multiple characters or shifting focus.
(F). Prompt: (red hair)
Meaning: Specifies the girl's hair color as red, no explicit weight, default is 1.0
Meaning: Single brackets indicate standard influence, equal to other descriptions.
(D). Prompt: (score_9, score_8_up, score_7_up)
◉Explanation: Users typically rate images based on aesthetics, detail, and artistry on a scale of 1 to 10.
◉score_9 indicates a near-perfect image (9 points), while score_8_up and score_7_up indicate images scoring 8 or above or 7 or above.
◉score_9: Represents top-tier quality, nearly perfect images, usually with extreme detail, artistry, and visual impact.
◉score_8_up: Indicates images scoring 8 or above, within the excellent range, possibly with minor flaws but overall outstanding.
◉score_7_up: Indicates images scoring 7 or above, within the good to excellent range, suitable for most high-standard needs.
◉Conclusion: (score_9, score_8_up, score_7_up) instructs the AI to prioritize generating images with a quality of 7 or above (ideally close to 9).
(6). Self-designed pattern: A girl from NTUT Business Administration playing volleyball
(A). Chinese Target: I want to design a 1K image on Pixai, featuring a Taiwanese university girl wearing sportswear (short sleeves, shorts), with a ponytail, coffee-brown dyed hair, playing volleyball, with an NTUT logo on her clothes. Please provide an English prompt.
(B). Note: The more prompts provided, the more credits will be deducted.
(C). Prompt 1:
1girl, Taiwanese university student, (ponytail, coffee brown dyed hair:1.2), wearing sports outfit (short-sleeve top, shorts), NTUT logo on chest, playing volleyball, dynamic pose, mid-air action, (athletic build:1.1), youthful face, bright smile, (detailed sports uniform:1.2), outdoor volleyball court, sunny day, blue sky, crowd in background, (highly detailed:1.2), (high quality:1.2), 1K, vibrant colors, sharp focus, perfect lighting, dynamic angle, sweat drops, realistic texture, ((masterpiece:1.3)), ((best quality:1.3)), anime style, source_anime
1 girl, Taiwanese university student, (ponytail, coffee brown dyed hair:1.2), wearing sportswear (short-sleeve top, shorts), NTUT logo on chest, playing volleyball, dynamic pose, mid-air action, (athletic build:1.1), youthful face, bright smile, (detailed sports uniform:1.2), outdoor volleyball court, sunny day, blue sky, crowd in background, (highly detailed:1.2), (high quality:1.2), 1K, vibrant colors, sharp focus, perfect lighting, dynamic angle, sweat drops, realistic texture, ((masterpiece:1.3)), ((best quality:1.3)), anime style, source_anime.
(D). Negative Prompt:
low quality, blurry, distorted face, extra limbs, modern casual clothing, bad anatomy
Low quality, blurry, distorted face, extra limbs, modern casual clothing, bad anatomy.
(8). Self-designed pattern: Create a 2D anime image with a specified body pose
(A).【Download a Jumping Photo】 (B). Set the body pose 【ControlNet】: Add image control ➜
Illustration ➜【OpenPose Full (Full Body)】:
Illustration ➜【Upload an Image】:
Illustration,
Illustration (C). Goal: I want to design a 1K image on Pixai, featuring a Taiwanese university girl wearing sportswear (short sleeves, shorts), with a ponytail, coffee-brown dyed hair, playing volleyball, with an NTUT logo on her clothes. Please provide an English prompt.
(D). Prompt:
A Taiwanese university girl playing volleyball, wearing a sporty outfit (short-sleeved shirt and shorts), brown-dyed hair tied in a ponytail, athletic and energetic, her shirt has the "NTUT" logo, in an outdoor volleyball court, mid-action pose, intense expression, bright and dynamic lighting, high-quality, 1K resolution, ultra-detailed, realistic textures, sharp details, vivid colors, cinematic shot, perfect anatomy.
(D). Select the 1st set of 【Model, LoRA】:
Model Selection Illustration,
Output Result Illustration Result Image
1. 【Website】ideogram
2. Feature 1: Requires signing up (using a Gmail account)
3. Feature 2: Free (first 10 days, with a daily limit of 40 images), no copyright issues
(2). Using 【ideogram】 to Create the Logo of National Taipei University of Technology
(3). Using 【Gemini】 to Create the Logo of National Taipei University of Technology
◉ In 【prompt】 enter: draw the logo of National Taipei University of Technology with the text NTUT
◉ 【Generated Image 1】 ◉ Gemini can generate logos with text, but they are not as refined and professional as 【ideogram】.
◉ 【ideogram】 is the top choice for generating text-based logos.
(4). Using 【Recraft】 to Create the Logo of National Taipei University of Technology
◉ In 【prompt】 enter: draw the logo of National Taipei University of Technology with the text NTUT
◉ 【Generated Image 1】 ◉ 【Recraft】 can also generate text-based logos professionally, but slightly inferior to 【ideogram】.
◉ 【ideogram】 remains the best choice for generating text-based logos.
(5). 【Assignment Submission】:
(1). Use 【ideogram's logo creation feature】
(2). Goal: Create a logo image for your department
(3). 【Download PNG image】 and upload it to Zuvio
☎chp11. AI Video Generation (Vidu, Sora, inVideoAI, Kling, Luma, Runway, HailuoAI)
(1). Text-to-Video: The simplest, most stable, and completely free method: Vidu
1. Vidu AI is an AI video generation tool developed by Tsinghua University and Shengshu Technology, offering both free and paid versions.
2. Meaning of Vidu: (Video do, or We do)
3.◉【Vidu Website】 4. Free Version Usage Rights:
Video Length: The free version can generate 【4-second videos】.
Video Resolution: Supports ultra-clear output up to 1080p.
Generation Limit: Upon successful registration, free users get 【80 points per month】, consuming 4 points per video, allowing the creation of 20 videos per month.
Usage Restrictions: The free version is limited to personal use and cannot be used commercially.
5. Paid Version Advantages:
Video Length: The paid version allows generating videos up to 8 seconds long.
Point Acquisition: Paid users can obtain more points depending on the subscription plan.
Commercial Use: The paid version allows videos to be used for commercial purposes.
6. In summary, Vidu AI's free version is suitable for individuals to experience AI video generation, but it has limitations in video length, generation count, and commercial use. For higher flexibility and commercial usage, upgrading to the paid version is recommended.
7.◉【Vidu Introduction: China's First Indigenous Video AI Model】
(1). Text-to-Video: The simplest, most stable, and completely free method: Vidu
1. Vidu AI is an AI video generation tool developed by Tsinghua University and Shengshu Technology, offering both free and paid versions.
2. Meaning of Vidu: (Video do, or We do)
3.◉【Vidu Website】
(2). 【Reference-Based Video Generation】: Jay Chou Eating Watermelon in a Park
(1). Text-to-Video: The Simplest, Most Stable, and Completely Free Method - Vidu
1. Vidu AI is an AI video generation tool launched by Tsinghua University in Beijing and Shengshu Technology, offering both free and paid versions.
2. The meaning behind Vidu's name: (Video do, or We do)
3. ◉【Vidu Official Website】
(2). 【Text-to-Video】: A Bomb Drops and Explodes, Destroying the Cars on the Roadside
Using the【Text-to-Video】feature:
Capture a serene moment of a corgi puppy gracefully swimming in a large pool under the sunlight.
A submerged perspective showcases the puppy's gentle smile, as soft golden light pierces through the water, forming rippling patterns on the pool floor.
The scene is dominated by soft pink tones, creating a dreamy and ethereal atmosphere.
A high-resolution cinematic shot captures the water texture and the joyful expressions of the corgi in exquisite detail, resulting in a simple yet cinematic portrayal of tranquility and purity.
This minimalist yet emotionally rich setting conveys a sense of calm and happiness, making it an ideal choice for a visually engaging cinematic shot.
◉【Illustration】 ◉【AI-Generated Video Result】
Using the【Text-to-Video】feature:
Describing a serene and beautiful scene where a corgi puppy swims gracefully in a sunlit pool. Below is a detailed breakdown of this text:
Scene Setting: Describes a corgi puppy swimming in a large pool, capturing a tranquil moment.
Underwater Perspective: The submerged view allows us to see the puppy’s gentle smile, adding depth and dimension to the frame.
Lighting Effects: Soft golden light penetrates the water’s surface, creating mesmerizing reflections on the pool floor.
Color Composition: The scene primarily uses soft pink hues, enhancing a dreamy and ethereal ambiance that evokes warmth and tranquility.
High-Resolution Cinematic Shot: Captures the fine details of water textures and the joyful expressions of the corgi, making the image more vivid and lifelike.
Emotional Expression: Through this simple yet emotionally rich setup, it conveys a sense of peace and joy, which is crucial for film compositions.
Cinematic Feel: The entire scene is described as a simple yet filmic portrait of tranquility and purity, making it suitable for both still photography and motion pictures, boasting strong visual appeal.
◉【Illustration】 ◉【AI-Generated Video Result】
Camera Language and Technical Details (Cinematography and Lighting)
Format
Literary Description
Technical Description
Purpose
Story Design
Filming Execution
(2).【Ask ChatGPT to Write a Storyboard】: First, provide ChatGPT with the screenplay/story summary, then ask it to generate a detailed storyboard.
1. Step 1: First, provide ChatGPT with the screenplay/story summary, then ask it to create a storyboard.
◉ Input Prompt:
I want to create a 40-second microfilm that requires 10 scenes. Please help me design a storyboard.
Story Summary: Jay Chou, in the microfilm, is a musician searching for inspiration. Through everyday city scenes, he discovers that musical inspiration is hidden in every moment of life. The story unfolds in reverse chronology, with the final scene capturing a quiet moment of his musical creation.
2. Step 2: Copy and paste each scene's cinematography description into Vidu to generate AI videos.
Using the【Image-to-Video】feature:
◉【Upload Image: Jay Chou】 ◉ Input Prompt:
Scene One: Chasing Through the Bustling City (Seconds: 0-4)
Visual Description:
The camera starts with a close-up of a pair of leather shoes before zooming out to reveal Jay Chou navigating through a busy city street.
The background is filled with hurried pedestrians, speeding vehicles, and flashing lights, creating a sense of urgency.
He carries a notebook and occasionally looks up at the sky as if searching for something.
Sound Effects:
The background music consists of a fast-paced piano melody, adding tension.
The sounds of car horns and hurried footsteps blend into the atmosphere.
3. Step 3: Scene Two Storyboard Description
Using the【Image-to-Video】feature:
◉【Upload Image: Jay Chou】 ◉ Input Prompt:
Scene Two: Inspiration Strikes at the Subway Station (Seconds: 4-8)
Visual Description:
Jay Chou stands on the subway platform, a train rushes past, creating a gust of wind.
His gaze stops at the wall of the platform, covered with graffiti and advertisements. A phrase "The rhythm of life is under your feet" catches his attention.
He smiles slightly and quickly jots down a few notes in his notebook.
Sound Effects:
The deep rumbling of the passing train suddenly stops, transitioning the background music into a soft string melody, marking the moment of inspiration.
4. Step 4: Scene Three Storyboard Description
Using the【Image-to-Video】feature:
◉【Upload Image: Jay Chou】 ◉ Input Prompt:
Scene Three: Contemplation in a Café (Seconds: 8-12)
Visual Description:
Jay Chou sits by the window of a café, light rain falling outside, with raindrops trickling down the glass.
A cup of coffee sits on the table. He frowns slightly as he sketches musical notes on a staff in his notebook.
A nearby music box suddenly starts playing a tune, causing him to pause and listen attentively.
Sound Effects:
The gentle music box melody blends harmoniously with the background sound of rain.
Flaw 2: When prompts include "many dogs," "many cars," or "many people," Sora tends to generate errors.
【Example Screenshot】
Flaw 3: Cannot correctly display requested text titles, often misspells words.
【Example Screenshot】
Flaw 4: Certain keywords impose style restrictions. For example, if "magic" is mentioned, the scene defaults to a game-like virtual environment instead of realistic imagery.
【Example Screenshot】
Flaw 5: Simple scripts work best. Complex scripts tend to confuse Sora, leading to unrealistic and chaotic camera movements.
【Example Screenshot】
Flaw 6: Low success rate in generating accurate prompts, quickly consuming monthly credits.
Flaw 7: Uploaded images and videos must not contain people (only animals or objects are allowed).
(4). Sora's Advantages:
Advantage 1: Supports 【Text-to-Video, Image-to-Video, Video-to-Video】 (but does not allow people in uploads).
Advantage 2: The 【Remix】 function is powerful, allowing successful object modifications in videos (e.g., adding glasses or replacing a dog with a cat).
【Example Screenshot】
Advantage 3: The 【Blend】 function effectively merges two videos (e.g., blending a real cat with a cartoon cat effect).
【Example Screenshot】
Advantage 4: The storyboard feature 【Story Board】 allows setting different script segments within a 5-second video.
【Example Screenshot】
Advantage 5: The 【Re-Cut】 feature allows users to delete unwanted sections and regenerate missing video parts automatically.
【Example Screenshot】
(1). 【Text-to-Video】: A bomb drops, explodes, and destroys cars on the roadside.
Using the 【Text-to-Video】 function:
Enter command: A bomb drops, explodes, and destroys the cars on the roadside, causing a big fire.
◉ 【Example Screenshot】
(4). 【Text-to-Video】: A corgi puppy swimming underwater (1)
Using the 【Text-to-Video】 function:
Capturing a serene moment of a corgi puppy gracefully swimming in a large sunlit pool.
The underwater perspective showcases the puppy’s gentle smile, with soft golden light filtering through the water, casting a dance of light and shadow at the pool’s bottom.
The scene is dominated by soft pink hues, creating a dreamy and ethereal atmosphere.
High-resolution photography captures every delicate detail of the water’s texture and the corgi’s joyful expressions, resulting in a tranquil and cinematic portrait of innocence.
This simple yet emotionally rich setup conveys a sense of peace and happiness, making it an ideal choice for a visually appealing cinematic shot.
◉ 【Example Screenshot】
(5). 【Text-to-Video】: A corgi puppy swimming underwater (2)
Using the 【Text-to-Video】 function:
The scene describes a corgi puppy elegantly swimming in a large, sunlit pool.
Scene Setup: Capturing a peaceful moment of a corgi puppy swimming in a large pool, giving a sense of tranquility.
Underwater Perspective: Showing the puppy’s gentle smile from an underwater viewpoint.
Light Effects: Golden light filtering through the water creates dynamic and artistic visuals.
Color Palette: Soft pink hues enhance the dreamy and ethereal atmosphere.
High-Resolution Photography: Capturing intricate details of the water and the corgi’s joyful expressions.
Emotional Expression: The setup conveys a sense of peace and happiness.
Cinematic Feel: The scene is designed for a visually appealing cinematic shot.
◉ 【Example Screenshot】
◉ 【AI Generated Video Result 1】 ◉ Result: The corgi walks on water instead of swimming.
◉ Reason: Sora often misinterprets complex scripts and details.
(6). 【Text-to-Video】: A gentleman smiles and waves at me.
Using the 【Text-to-Video】 function:
A gentleman smiles and waves at me.
◉ 【Example Screenshot】
◉ 【AI Generated Video Result 2】 ◉ Result: When displaying multiple people, hand movements often appear unnatural (bending, sudden disappearance).
◉ Reason: Sora currently struggles with realistic hand design in multi-person scenes.
(8). 【Text-to-Video】: A horse running on a city road, with the camera following its movement.
Using the 【Text-to-Video】 function:
A horse running on a city road, with the camera following its movement.
Using the 【Story Board】 function:
Enter the first scene: A boy is running in the park. ◉ 【Example Screenshot】 Enter the second scene: The boy picks up a $100 bill from the ground. Enter the third scene: The boy jumps up excitedly to celebrate. Click 【Create】.
◉ 【Example Screenshot】
(2). If the video is unsatisfactory, modify it using 【Re-Cut】.
In the previous video, the final scene does not show the boy jumping in celebration, so it needs modification:
Use the 【Re-Cut】 function, select the first video.
◉ 【Example Screenshot 1】
, 【Remix Example Screenshot 2】 Drag the last second of the frames forward ➜ Click 【Create】.
◉ 【Example Screenshot】
(3). If the video is still unsatisfactory, modify it using 【Edit-Story】.
In the previous video, the final scene still does not include the boy jumping in celebration, so it needs further modification:
Use the 【Edit-Story】 function, select the first video.
◉ 【Example Screenshot 1】 Move the third scene’s frame forward by one position ➜ Click 【Create】.
◉ 【Example Screenshot】
Drawback: The free version requires a 3-hour wait for generating one video.
Text-to-Video: Firework explosion scene
◉【AI Generated Video Result 1】 Result: Unexpectedly, the final output contained fireworks???
(2). 【Kling AI - Video Generation】
Drawback: The free version requires a 700-minute (13-hour) wait for generating one video.
Text-to-Video: Waves crashing on the shore
◉【AI Generated Video Result 1】 Result: Normal
(3). 【Luma Labs AI - Video Generation】
Drawback: The free version does not require waiting, but the results are strange.
Text-to-Video: Dog playing with a ball in the park
◉【AI Generated Video Result 1】 Result: Strange outcome
(4). 【InVideo AI - Video Generation】
Advantages: Generates a 1-minute or 1.5-minute video instantly based on a text script.
Advantages: No need for video editing software. InVideo AI can create a video with automatic voice-over in just 30 minutes.
Advantages: Can process Chinese text and automatically generate Chinese voice-overs.
Drawbacks: Cannot specify reference images.
Drawbacks: Contains two watermarks.
Example: Jay Chou eating watermelon in a park.
◉【AI Generated Video Result 1】 Result: Due to insufficient input data, the video lacked significant variation.
1. Suno is an AI music generation tool with the following free user privileges.
2.◉【Suno Website】 3. Daily Limit: Free users can generate up to 【10 songs】 per day.
4. Generation Speed: Free users share computing resources, which may result in slower processing.
5. Commercial Use Restriction: Music generated with the free version is for personal use only and 【cannot be used for commercial purposes】.
6. For more features or commercial use rights, consider upgrading to the paid version.
(2). 【Explore】: Mood-Based Song Selection (Randomly listen to user-created songs)
Using the 【Explore】 feature: Roll the dice to randomly listen to user-generated songs.
◉【Illustration】,
【Illustration】
(3). 【Text-to-Music in Chinese】: "I Don't Want to Eat Cilantro" Song
Using the 【Create】 feature:
Input prompt: Write a song expressing 'I don’t want to eat cilantro.' Result: Inputting a Chinese prompt generates a Chinese song with Chinese lyrics.
Features: The lyrics are rhymed, aesthetically appealing, and structured with verses and choruses.
Features: Two different arrangements and melodies are generated for the same lyrics.
◉【Illustration】,
【Generated Lyrics Illustration】 ◉【Download MP3 Audio】 ◉【Download MP4 Video】
(4). 【Text-to-Music in English】: My Girlfriend Wants to Break Up with Me
Using the 【Create】 feature:
Input prompt: My girlfriend wants to break up with me. Write a song that describes how I feel right now. Result: Inputting an English prompt generates an English song with English lyrics.
◉【Illustration】 ◉【Download MP3 Audio】 ◉【Download MP4 Video】
(1). Text-to-Music Generation: Free Method Using Suno
1. Suno is an AI music generation tool that provides the following features for free users.
2. ◉ 【Suno Official Website】 3. Daily Limit: Free users can generate 【10 songs】 per day.
4. Generation Speed: Free users share computing resources with others, which may result in slower processing speeds.
5. Commercial Use Restriction: Music generated by free users is strictly for personal use 【not for commercial purposes】.
6. If you need more features or plan to use the music commercially, consider upgrading to the paid version.
(2). Using 【sunoprompt.com】: Professional Song Style Generator
◉ 【SunoPrompt Song Style Generator Website】 Describe your song theme: I want to create a song to celebrate getting first place in my math exam today. Select various parameters: theme, melody, harmony, rhythm, structure, instrumentation, style/genre, emotion, dynamics, production, originality, and vocal style.
◉ 【Example Screenshot】
, 【Example Screenshot】
, 【Example Screenshot】 Copy the song generation prompt from SunoPrompt, paste it into Suno to generate an English AI song.
Translate the SunoPrompt song generation prompt into Chinese, then paste it into Suno to generate a Chinese AI song.
----------------------------
Title: I want to create a song to celebrate getting first place in my math exam today.
Song Style: A joyful rock song with a driving electric guitar and off-beat rhythms, featuring a soft piano backdrop enhanced by reverb and delay effects, following a classic AABA structure.
◉ 【Result: English Song 1】 ◉ 【Result: English Song 2】
----------------------------
Title: I want to create a song to celebrate getting first place in my math exam today.
Song Style: A rock song with strong rap vocals, full of cheerful emotions, featuring a gradually rising melody and syncopated rhythms. The electric guitar plays the main motif, supported by soft piano chords and a rich harmonic texture. The classic song structure blends rock and rap elements, making the song highly creative.
◉ 【Result: Chinese Song 1】 ◉ 【Result: Chinese Song 2】
----------------------------
Title: I want to create a song to celebrate getting first place in my math exam today.
Song Style: A combination of rock, rap, and electronic effects with a raw production style. This high-energy track features a fast tempo and a gradually rising melody, with electronic beats serving as the primary instrumental background.
◉ 【Result: Chinese Song 3】
(3). Using 【Suno Chirp v1 Model Supported Keywords】: Various Music Styles and Terms
----------------------------
Title: I want to create a song about helping stray dogs.
Song Style: This song incorporates R&B and soul music elements.
◉ 【Result: Chinese Song 1】
(1). Text-to-Music Generation: Free Method Using Suno
1. Suno is an AI music generation tool that provides the following features for free users.
2. ◉ 【Suno Official Website】 3. Daily Limit: Free users can generate 【10 songs】 per day.
4. Generation Speed: Free users share computing resources with others, which may result in slower processing speeds.
5. Commercial Use Restriction: Music generated by free users is strictly for personal use 【not for commercial purposes】.
6. If you need more features or plan to use the music commercially, consider upgrading to the paid version.
2. 【Setting Lyrics】: ◉ Method 1: Copy lyrics from an existing song, classical poetry, or modern poetry.
◉ Method 2: Ask ChatGPT to generate lyrics:
Example 1: Write lyrics for a song about secretly liking a classmate but being too afraid to confess. The song should be two minutes long, in Chinese, and must rhyme.
Example 2: Assume you are a professional lyricist. Write lyrics about secretly liking a classmate and wanting to confess but hesitating. The song style should be "a sweet but anxious feeling," two minutes long, in Chinese, and must rhyme.
◉ Method 3: Use Suno to randomly generate lyrics.
◉ Method 4: Write original lyrics yourself.
3. 【Setting Music Style】: ◉ 【Reference: Suno's Supported Music Styles】 Select different styles/genres, and emotions: It's best to input style names in English, as Suno understands them better.
Example of a single style: jazz
Example of multiple styles: jazz, emotional, epic
Example of Chinese styles: 爵士, 情感音樂, 史詩音樂
Style breakdown of Jay Chou's "Seven Miles Fragrance": Pop (80), Ambient (60), Indie (40), Electronic (30), Rock (20)
Style breakdown of "Blue and White Porcelain": Pop (80), Soul (70), Folk (60), Ballad (50), Acoustic (40)
Style breakdown of "Rice Fragrance": Pop (80), Soul (70), Acoustic (60), Ballad (50), Folk (40)
Style breakdown of "Nunchucks": Pop (80), Rock (70), Punk (60), Electronic (20), Hip hop (10)
What is 【Ambient Music】:
Ambient music focuses on atmosphere rather than structure, often having a slow or non-existent rhythm, making it suitable for relaxation and meditation.
(3). How to Identify a Song's Style and Genre: 【SonoTeller】 AI Music Analysis
1. ◉ 【SonoTeller】 AI Music Analysis Tool 2. Function: Input a 【Song Name】 or 【YouTube URL】 to analyze the song's style and genre.
3. Limitation: Daily usage is limited to 1-3 times.
Example 1: Analyzing "Nunchucks"
◉ Input YouTube URL: https://www.youtube.com/watch?v=OR-0wptI_u0
◉ Result: "Nunchucks" genre classification: Pop (80), Rock (70), Punk (60), Electronic (20), Hip hop (10)
【Example Screenshot】
Example 2: Analyzing "Rice Fragrance"
◉ Input song name: "Rice Fragrance"
◉ Result: "Rice Fragrance" genre classification: Pop (80), Soul (70), Acoustic (60), Ballad (50), Folk (40)
2. CapCut is a versatile and user-friendly video editing software suitable for personal creation, social media video production, and professional editing. Developed by ByteDance, it is available for mobile (iOS, Android) and desktop (Windows, Mac).
Features:
Easy to use: Provides an intuitive interface for beginners.
Rich asset library: Includes various effects, filters, music, and subtitle templates.
Multi-platform sync: Allows cross-device editing and access to projects anytime, anywhere.
Supports multiple formats: Capable of exporting in high resolutions (e.g., 4K).
AI-powered features: Includes AI subtitle generation, speech-to-text, and background removal.
3. Free Version Capabilities
Basic Editing Features:
Video splitting, merging, and trimming.
Transitions, filters, and basic animations.
Adding subtitles and music.
Templates & Asset Library: Provides some free templates, music, and effects. Premium content requires a subscription.
AI Subtitle Generation: Automatically generates subtitles, but advanced styles require the Pro version.
Sound Effects & Music: Includes free sound effects and background music, though some are restricted for non-commercial use.
Advanced Feature Limitations: AI effects like motion tracking and 3D images may have restrictions. Free exports may include a watermark.
4. Free vs. Pro Version Comparison
Feature
Free Version
Pro Version
Asset Library
Some free assets
Full access
AI Features
Basic features (e.g., auto subtitles)
Advanced effects & premium styles
Export Quality
Up to 1080p (some with watermark)
Up to 4K, no watermark
Music & Sound Effects
Free music available
Licensed premium music
Advertisements
No ads
No difference
5. CapCut vs. JianYing Comparison
Comparison
JianYing
CapCut
Market
Greater China (including Taiwan)
International
Language Support
Supports Simplified and Traditional Chinese
Supports English and other languages, but not Traditional Chinese
Functionality
More localized content, catering to Asian users
More international content, suited for global users
Music Licensing
Includes more Asian-region music
More globally licensed music
Template Preference
Supports Asian-style templates
More international-style templates
Interface
Supports Traditional Chinese
Mainly in English
Target Users
Users in Taiwan and Greater China
Global users
6. Recommendation for Use in Taiwan
If you mainly use it in Taiwan:
Choose "JianYing" as it supports Traditional Chinese and provides more localized assets and templates.
If you work internationally or create content for a global audience:
Choose "CapCut" as it better meets international needs with more global templates and assets.
7. Using CapCut in Taiwan Requires an Apple ID (Cannot Use Douyin Account)
(2). 【Using ChatGPT to Generate Atmosphere Description for Lofi Music】
Input Prompt: Generate five prompt words for an image that fits the background of Lofi music, creating a warm, nostalgic Lofi atmosphere.
◉ 【Lofi Description Generated with ChatGPT】 ◉ First Result:
“A cozy room with a warm desk lamp illuminating a wooden table, an open notebook, a coffee cup, and an old radio, with soft moonlight streaming through the window and a misty forest in the background, exuding nostalgia.”
(3). 【Using the Above Description to Generate Instrumental Music with Suno】:
Using the "Create" function for text-to-music generation:
Input Prompt: A cozy room with a warm desk lamp illuminating a wooden table, an open notebook, a coffee cup, and an old radio, with soft moonlight streaming through the window and a misty forest in the background, exuding nostalgia.
◉ Remember to check the "Instrumental" option to generate pure instrumental music.
◉ 【Instrumental Music Generated with Suno】
(4). 【Using the Above Description to Generate a Scene Image with Krea】:
Using the "Image Generation" function:
Input Prompt: A cozy room with a warm desk lamp illuminating a wooden table, an open notebook, a coffee cup, and an old radio, with soft moonlight streaming through the window and a misty forest in the background, exuding nostalgia.
◉ 【Background Image Generated with Krea】
(5). 【Using the Image + Music to Create a Lofi Music Video with CapCut】:
1. Apply 【AI Effects】 to the Image ➜ 【Check: Effects】 ➜ 【Camera Movements】 ➜ 【3D Image】
2. Shorten the image duration on the timeline to match the 3D animation effect.
3. Duplicate the first image and apply 【Reverse Playback】 to create a seamless transition.
4. Set 【Speed Adjustment】 for both images ➜ 【Speed Multiplier: 0.3】.
5. Enable 【Smart Frame Interpolation】 ➜ Select 【Optical Flow Method】.
6. Duplicate the two-image sequence multiple times to match the music duration.
7. Apply 【Special Effects】 ➜ 【Atmosphere】 ➜ Select 【Floating Light Particles】.
8. Adjust 【Transparency and Speed】 of the light particles.
(3). [Using CapCut to Convert Simplified Subtitles to Traditional Chinese - Method 1: Set Traditional Fonts]
Convert Simplified to Traditional Chinese Subtitles:
◉ Step 1: Open the [Font] list
◉ Step 2: Select from the list: [Wenyi Traditional], [Hong Kong Style Traditional], or [Modern Traditional] fonts
(If you can’t find them, search for the keyword: “Traditional”)
◉ [Illustration 1],
[Result Illustration 2] ◉ Note: AI recognition may still produce errors such as simplified characters, typos, awkward line breaks, or unsatisfactory font appearance
◉ [Illustration 1]
(4). [Using CapCut to Convert Simplified to Traditional Chinese - Method 2: Export Subtitles, Translate, Then Re-import]
Convert Simplified to Traditional Chinese Subtitles:
◉ Step 1: Export the video
◉ [Illustration 1] ◉ Step 2: Uncheck [Export Video]
◉ [Illustration 1] ◉ Step 3: Check [Export Subtitles], choose format [txt]
◉ [Illustration 1] ◉ Step 4: Requires VIP membership
◉ Step 5: Log in with an Apple account (but subtitle export has usage limits even with Apple login)
Recommendation: Use the CapCut International Cloud version to access AI subtitle recognition and SRT export for free
Not Recommended: Signing up with Apple account due to limited uses
◉ [Illustration 1],
[Illustration 2] ◉ Step 6: Subtitle export complete
◉ [Illustration 1] ◉ Step 7: Open txt subtitles, copy to Word, and convert to Traditional Chinese
◉ [Illustration 1],
[Illustration 2] ◉ Step 8: Copy Traditional Chinese from Word
◉ Step 9: Delete original subtitles in CapCut (Select All ➜ Delete)
◉ Step 10: [Text] ➜ [Auto Captions] ➜ [Script Matching] ➜ [Start Matching] ➜ Paste Script ➜ [Start Matching]
◉ [Illustration 1],
[Illustration 2],
[Illustration 3],
[Result Illustration 4]
Demonstration Using CapCut International Cloud Version: Free access to [AI Subtitle Recognition & SRT Export]
(2). [Using CapCut to Import ESG Video and Auto-Recognize Subtitles with AI]:
AI Auto Subtitle Recognition:
◉ Step 1: Drag the video to the timeline and select it
◉ Step 2: [Text] ➜ [Add Text] ➜ [Default Text]
◉ Step 3: Input three test subtitle lines:
Welcome to the ESG Charging Station
This episode will share insights on corporate ESG and sustainable operations
We aim to provide a more comprehensive understanding of how ESG operates in companies
◉ Step 4: [Set subtitle text color]
◉ [Illustration 1] ◉ Step 5: [Set subtitle font size and position]
◉ [Illustration 1] ◉ Step 6: Adjust timeline position
◉ [Illustration 1]
(3). [Using CapCut to Replace Subtitles with Custom Voiceover]
Replacing subtitles with custom voiceover:
◉ Step 1: Lower the original video volume to minimum
◉ [Illustration 1] ◉ Step 2: [Test voice tone] ➜ Select a [Free Voice Character] (Do not choose paid voice options)
◉ [Illustration 1] ◉ Step 3: Use the mouse to click the selection square [Select All] for all subtitles
◉ [Illustration 1] ◉ Step 4: For all subtitles ➜ Click [Export] in the top right ➜ [Read Aloud] ➜ Choose a free [Voice Character] (member-exclusive voices are marked with a blue icon)
◉ Step 5: [Start Reading Aloud]
◉ [Illustration 1] ◉ Step 6: [Export] the final video
◉ [Illustration 1]
(2). [Using CapCut to Import Video and Automatically Recognize Lyrics for Subtitles]:
AI Subtitle Recognition:
◉ Step 1: Drag the video to the timeline and select it.
◉ Step 2: [Text] ➜ [Recognize Lyrics] ➜ [Recognize Lyrics]
◉ [Illustration 1] ◉ Step 5: [Adjust the font size and position of the subtitles]
◉ [Illustration 1]
(3). Note: Subtitling (AI function, [Subtitle Recognition, Lyrics Recognition]) is limited to 4 free uses per month
Note: CapCut's free user limitation for [Subtitle Recognition] feature
As of March 27, 2025, CapCut's [Subtitle Recognition] function (AI speech-to-text to generate subtitles from video audio) is no longer available to free users. According to online sources, this feature became paid around June 2024 and is now limited to VIP or Pro members. Free users must upgrade to paid plans (e.g., CapCut VIP) to access it.
Conclusion: [Current Status] – You can use [Subtitle Recognition, Lyrics Recognition], but cannot export the video.
Past Status: In 2021, CapCut offered [Subtitle Recognition] free to all users with high accuracy in recognizing Chinese and English speech, supported exporting SRT files, and was well-received by content creators.
CapCut Subtitle and Lyrics Recognition Feature Access
1. Free Membership Access
5 Free Uses Per Month: Newer versions of CapCut (e.g., 6.0.1 and above) limit subtitle and lyrics recognition to 5 times per month for free users. Further use requires a VIP membership.
Subtitle File Export Not Limited: Exporting subtitle files (e.g., SRT format) is not restricted and remains free.
Older Versions Still Free:
CapCut version 5.9: Fully free, no VIP requirement, unlimited subtitle/lyrics recognition.
Early Beta Versions: Some official desktop beta versions still allow free use of subtitle recognition.
2. Methods to Bypass Charges
Manually Copy Lyrics: After recognition, manually copy and paste each line into a new text layer to avoid export fees.
Use Older Versions: Download CapCut version 5.9 or earlier to avoid limitations.
Use Alternatives: Try other tools such as BiJian (from Bilibili), Whisper (open-source speech recognition), or Cloud Editors.
New Version Encryption: After version 6.0.1, project files are encrypted and subtitles can no longer be extracted directly. Export must be done through official tools.
Long-term Free Solutions: Use older versions (e.g., 5.9), copy lyrics manually, or use alternative tools.
(4). Alternative Solution: Use CapCut International Cloud Version to Automatically Recognize Lyrics, Export Simplified SRT, Convert to Traditional SRT, then Import into CapCut
(5). First Export the Simplified Chinese Lyrics as an SRT Subtitle File
◉ Step 1: Drag the video to the timeline and select it
◉ Step 2: [Export] ➜ [Export Subtitles] ➜ [SRT]
◉ [Illustration 1] ◉ Step 3: Copy to Word ➜ Convert Simplified to Traditional Chinese
◉ [Illustration 1] ◉ Step 4: Save in Word as [Plain Text SRT File]
◉ [Illustration 1]
(6). Import the SRT Subtitle File under [Local Subtitles]
(3). [Create Opening Segment: Segment 1]: Introduce your department, name, and student ID
◉ Step 1: Top left corner [Text] ➜ [Text Template] ➜ [Opening Title]
◉ [Opening Title] ➜ [Winter Travel]
◉ [Illustration 1],
◉ Step 2: Enter 4 parameters: Business Management Department, Chen Qingwen, A00123, Introduction to Artificial Intelligence
◉ [Illustration 1],
◉ Step 3: Drag the segment duration to: 10 seconds
◉ [Illustration 1]
(4). [Create Video Theme Introduction Segment: Segment 2]: Introduce the theme of the video 'What content theme are you creating?'
◉ Step 1: Top left corner [Text] ➜ [Text Template] ➜ [Long Text]
◉ [Long Text] ➜ [Blind Box Description] (Free, no blue icon)
◉ [Illustration 1],
[Illustration 2]
◉ Step 2: Enter Parameter 1:
Create a music video for Jay Chou's song
Describe Jay Chou searching for creative inspiration across the city
Theme song: "Qilixiang"
◉ Step 3: Enter Parameter 1: Theme
◉ [Illustration 1],
◉ Step 3: Drag the segment duration to: 12 seconds
◉ [Illustration 1]
(5). [Add Voiceover to Segment 2 Text: Narration]
◉ Step 1: Select the second video segment
◉ Top right corner [Export] ➜ [Narration] ➜ Select [Free Character Voice] ➜ Select [Translated Film Male] ➜ [Start Narration]
◉ [Illustration 1]
(6). [Create Technology Introduction Segment: Segment 3]: Introduce the technologies used in the video 'What technologies did you use?'
◉ Step 1: Top left corner [Text] ➜ [New Text] ➜ [Default Text]
◉ Step 2: Enter parameters:
Technologies used include:
Using Krea to generate images
Using Suno to generate songs
Using Vidu to generate videos
Using ChatGPT to generate scripts
Using CapCut to edit videos
◉ [Illustration 1],
◉ Step 3: Drag the segment duration to: 12 seconds
◉ Step 4: Set the [Animation] for this segment to: [Shake]
➜ [Select video segment] ➜ Top right corner [Animation] ➜ [Loop] ➜ [Shake] (Free, no blue icon)
◉ [Illustration 1],
◉ Step 5: Set the narration for this segment
◉ [Illustration 1]
(7). [Create MV Cover Segment: Segment 4]: Introduce the MV video 'Cover'
◉ Step 1: Top left corner [Text] ➜ [New Text] ➜ [Default Text]
◉ Step 2: Enter parameters:
Qilixiang MV
Lyricist: Vincent Fang
Composer: Suno
MV Director: Chen Qingwen
◉ [Illustration 1],
◉ Step 3: Drag the segment duration to: 10 seconds
◉ Step 4: Set [Manual Voiceover] for this segment
◉ Step 5: Drag [Default Text] above the existing text
Enter parameters: The song title is Qilixiang, and the lyricist is Vincent Fang
◉ [Illustration 1],
◉ Step 6: Voiceover ➜ [Narration]
◉ [Illustration 1],
◉ Step 7: Set the upper [Text] to hidden
◉ [Illustration 1]
(8). [Add Manual Voiceover to MV Cover Segment 5]
◉ Step 1: Set the second [Manual Voiceover] for this segment
◉ Step 2: Drag [Default Text] above the existing text
Enter parameters: The composer is Suno, and the MV director is Chen Qingwen
◉ [Illustration 1],
◉ Step 3: Voiceover ➜ [Narration]
◉ [Illustration 1],
◉ Step 4: Set the upper [Text] to hidden
◉ [Illustration 1]
(9). [Drag 10 Jay Chou Video Segments with Qilixiang MP3]: Create a slow-motion playback video
◉ Step 1: Drag 10 Jay Chou video segments with Qilixiang MP3 to the timeline
◉ Step 2: Select all video segments
◉ Step 3: Top right corner [Speed] ➜ [Duration]: 8 seconds (2x duration, becomes slow motion)
◉ [Illustration 1]
(10). [Add Transition Effects Between Video Segments]
◉ Step 1: Drag 10 Jay Chou video segments with Qilixiang MP3 to the timeline
◉ Step 3: Top left corner [Transition] ➜ [Popular] ➜ Select one
(Click to preview and test; it indicates download completion (blue) before dragging to the timeline)
◉ [Illustration 1]
◉ Step 4: Drag [Transition] between all 10 video segments
◉ [Illustration 1]
(11). [Import Lyrics in Local Subtitles]: SRT Subtitle File
(12). [Extend the Duration of the 10th Video Segment to End with a Lyric Line]
◉ Step 1: Select the 10th video segment
◉ Step 2: Top right corner [Speed] ➜ [Duration]: 9 seconds
◉ [Illustration 1]
(13). [Trim the Song Audio Track to Match the Video Length]
◉ Step 1: Move the [Timeline Marker] ➜ to the end of the video
◉ Step 2: Select the [Audio] track
◉ Step 3: In the middle panel on the left, click [Split]
◉ [Illustration 1]
◉ Step 4: [Delete], remove the excess audio on the right
◉ [Illustration 1]
(14). [Set Fade-In and Fade-Out for the Song Audio]
◉ Step 1: Select the [Audio] track
◉ Step 3: Top right corner, click [Basic]:
[Fade-In Duration]: 3.0s
[Fade-Out Duration]: 4.0s
◉ [Illustration 1]
(15). [Set the Video Ending to Gradually Disappear]: Exit Animation
◉ Step 1: Select the last segment of the [Video] track
◉ Step 3: Top right corner, click [Animation] ➜ [Exit] ➜ Test various animations, select [Fade Out]
(Do not select premium effects; a blue icon indicates a premium effect, e.g., Screen Wipe is premium)
◉ [Illustration 1]
(16). [Set the Video Beginning to Fade In, Gradually Clear]: Entry Animation
◉ Step 1: Select the first segment of the [Video] track
◉ Step 3: Top right corner, click [Animation] ➜ [Entry] ➜ Test various animations, select [Fade In]
(Do not select premium effects; a blue icon indicates a premium effect, e.g., Screen Wipe is premium)
◉ [Illustration 1]
MyEdit is an online multimedia editing tool designed for processing audio, video, images, and text. It integrates various editing functions, making it suitable for content creators, designers, students, and general users.
Key Features
Audio Editing: Cut, merge, and adjust audio files, including noise reduction and sound quality enhancement.
Video Editing: Basic editing features such as adding subtitles, filters, and effects.
Image Editing: Crop, resize, and enhance image quality.
Text Processing: Extract text, convert formats, and more.
AI Support: AI-driven quick processing tools (e.g., speech-to-text, image enhancement).
Free User Limitations
Daily Usage Limit: Free users can perform up to 5 actions per day, depending on the selected function (e.g., audio or image processing).
File Size Limit: Each file must not exceed 50MB.
Feature Availability: Basic features (such as simple trimming and merging) are available for free, while advanced features (such as high-end filters, batch processing, and AI support) require a paid subscription.
Export Restrictions: Free users' exported files may contain watermarks (for video and image editing).
(2). Basic Usage Demonstration:
(A). Sign in to the website ➜ Click 【Sign In】 and log in with a Google account.
【Example Image】 (B). Change language to Traditional Chinese.
【Example Image】 (C). Select 【Audio Editing Tools】 ➜ 【AI Sound Effect Generator】.
【Example Image】 (D). Click 【Generate Free Sound Effects】.
【Example Image】 (E). Enter prompt text:
Sound of a motorcycle passing by.
◉2. NaturalReader is a professional text-to-speech application that converts any written text into spoken words.
It offers free and paid subscriptions to meet the needs of different users with various budgets.
(1). Paid Plus Subscription: Includes exclusive features and access to Plus and LLM (Large Language Model) voices, the latest and most advanced voices.
Using LLM technology, you can even clone your voice within minutes and make it speak in over 100 languages.
(2). Free Users: Can use Premium Voices for 20 minutes per day and Plus Voices for 5 minutes per day or have unlimited access to available free voices.
Mobile apps allow users to listen on the go.
With the mobile camera, the app can also read physical books and notes aloud.
3. For non-personal use (e.g., YouTube videos, e-learning, or other commercial or public uses), additional regulations apply.
4. The Chrome extension allows direct listening from web pages, Google Docs, online Kindle books, and emails.
You can add it to Chrome for free.
(2). Advantages and Disadvantages of NaturalReader:
◉1. Advantages:
(1). AI human-like voices that do not sound robotic.
(2). Uses professional broadcaster voices, similar to TV anchors.
(3). Offers a Chrome extension to listen to webpage text in a natural voice.
◉2. Disadvantages:
(1). The free version only allows listening, not downloading MP3 files.
(2). Paid voices are recorded by real people, while free voices sound robotic.
(3). Example: Common Influencer Voiceovers Cloned from Famous Broadcasters in China
Multilingual Support: Supports multiple languages, including Chinese, English, Spanish, Japanese, etc., meeting the needs of global users. Particularly suitable for creating cross-language content or international projects.
Multiple Voice Options: Offers a variety of voice styles, including male, female, and child voices. Voices can be adjusted for speed, pitch, tone, etc., to simulate real emotional expression.
High-Quality Audio: Generates high-quality audio with natural and fluent speech, suitable for professional scenarios such as video dubbing, course recording, and advertisement production.
Free and Commercial Licensing: Provides a free version for personal use, but for commercial purposes, users can upgrade to a paid plan for higher usage rights.
Easy Operation: Features a user-friendly web interface with no software download required. Users simply input text and select voice settings to generate audio.
Fast Generation: TTSMaker generates speech quickly, even for longer texts, completing the process in a short time.
Application Scenarios
Video Dubbing: Quickly generates video narration, suitable for social media, YouTube, and short ads.
Course Recording: Used for audio production in online education or explainer videos.
Voice Navigation: Ideal for navigation systems, customer service voice automation, etc.
E-books and Podcasts: Converts text into speech for e-book narration or podcast production.
Advantages
High Usability: Operable without professional skills.
Low Cost: Saves significant time and money compared to human recordings.
Quick Adjustments: Modify text and regenerate audio anytime.
Suitable Users
Content Creators: Video producers, podcasters, course instructors.
Businesses: For commercial advertisements or automated voice systems.
(2). Advantages, Disadvantages, and Limitations of TTSMaker:
◉1. Free Version Limitations:
(1). Text conversion limit per session: 1000 characters
◉2. Advantages:
(A). Offers 90 characters with various voice types.
(B). Provides diverse voice styles: cartoon, child (male, female), intellectual, emotional, cute, customer service, general, speech, reading, advertisement, deep, magnetic, and sweet voices.
[Illustration 1], [Illustration 2] (C). Certain characters (with yellow [Emotion] tags) offer various emotional tones: angry, furious, sad, happy, excited.
Users can preview, e.g., 74: 1510 - Ya Aya - Popular General Female Voice (Test: angry, happy, sad).
[Illustration] (D). Includes regional accents: Sichuan, Guangdong (Cantonese), Northeast, Shandong, Hong Kong-Taiwan, Taiwan accent, Hong Kong Cantonese.
[Illustration] (E). Features voices of well-known mainland Chinese broadcasters.
(F). Playback audio in MP3 format can be downloaded.
◉3. Disadvantages:
(A). Text conversion limit per session: 1000 characters.
(3). Example: Commonly Heard Online Influencer Voiceovers: Cloned from Famous Mainland Chinese Broadcasters
(A). For example: Characters 89 and 90 (professional dubbing for popular Douyin videos), frequently heard on YouTube videos.
[Illustration] (B). For example: Character 17, magnetic voice, often heard on YouTube videos.
[Illustration] (C). For example: Character 14, male news voice, frequently heard on YouTube videos.
[Illustration]
(4). Usage Demonstration
(A). First, input text (within 1000 characters).
Feature Highlights:
Multilingual Support: Supports multiple languages, including Chinese, English, Spanish, Japanese, etc., meeting global user needs. Ideal for cross-language content or international projects.
Multiple Voice Options: Offers various voice styles, including male, female, and child voices. Voices can be adjusted for speed, pitch, tone, etc., to simulate real emotional expression.
(B). Select a character (check the box next to Character 90 (professional dubbing for popular Douyin videos), a yellow circle will appear).
[Illustration] (C). Enter [graphic verification code], then [Start Conversion], listen to the preview, and download the MP3 to your local computer.
Fish Audio is a leading brand in music and audio technology, providing professional musicians and creators with high-quality music content and tools.
Key Features
High-Quality Audio Content: Fish Audio offers a rich library of sound effects, background music, and audio samples for movies, games, and advertisements.
Professional Tool Integration: Compatible with various digital audio workstations (DAWs), enhancing music production efficiency.
Innovative Technology: Incorporates AI-based audio processing and editing features.
Free User Limitations
Limited Generations: Free users can generate voice audio 20 times every 30 days.
Text Input Limit: Each request allows a maximum of 500 bytes, approximately 30 seconds of speech.
Quota Reset: The free quota resets at 00:00 UTC every 30 days.
Additionally, reports suggest that free users get 50 attempts per day, each limited to 500 bytes (~160 Chinese characters), generating approximately 35 seconds of audio per request.
(A). Click 【Explore More】:
【Example Screenshot】 (B). Select a character: Choose 【Jay Chou】 ➜ Click 【Use Voice】
【Example Screenshot】 (C). Enter text:
The song is called "San Li Xiang," lyrics by Fang Wenshan.
Composed by Suno, MV directed by Chen Qingwen.
The theme of this MV is to create a music video for Jay Chou's song,
The story follows Jay Chou as he roams the city seeking creative inspiration.
The theme song is: "San Li Xiang."
AI technologies used:
Using Krea to generate images.
Using Suno to compose music.
Using Vidu to create videos.
Using ChatGPT to generate scripts.
Using Fish Audio to clone voices.
Using CapCut to edit videos.
(D). Click 【Create】 ➜ Then 【Download MP3】
【Example Screenshot 1】
, 【Example Screenshot 2】
(E). Important Tip: Avoid Line Breaks!
◉ Example 1: This MV is about creating a music video for Jay Chou’s song, documenting his search for inspiration in the city. The theme song is "San Li Xiang." ◉ Example 2: The song is called "San Li Xiang," lyrics by Fang Wenshan. Composed by Suno, MV directed by Chen Qingwen. ◉ Example 3: Technologies used: Krea for images, Suno for music, Vidu for videos, ChatGPT for scripting, Fish Audio for voice cloning, CapCut for video editing.
(1). Convert English Videos into English Subtitles: The Easiest and Completely Free Method – Upload to YouTube, and It Will Automatically Generate Subtitles
2. Log in to YouTube and Upload the Video:
Sign in to your YouTube account, click the "Create" button in the upper right corner, then select "Upload Video."
Follow the prompts to upload your English video.
3. Wait for Automatic Subtitle Generation:
Once uploaded, YouTube will automatically recognize the speech in the video and generate subtitles.
This process may take a few minutes to several minutes (this video is about 5 minutes long), depending on the length of the video and server processing speed.
Processing in Progress - Illustration
4. View and Edit Subtitles:
After processing (about 5 minutes), check the subtitle function in the video.
Completed Subtitles - Illustration
6. Edit Subtitles:
You can find the automatically generated subtitles in YouTube Studio's "Subtitles" settings.
Click "Subtitles" to check the generated subtitles and edit or adjust them (if the automatic recognition contains errors).
Alternative Method: In YouTube Studio ➜ Click "Subtitles" ➜ Copy and Edit
Copy and Edit - Illustration Edit Subtitles - Illustration
8. Important Notes:
◉ The accuracy of automatic subtitle generation depends on audio quality and speech clarity. Background noise or strong accents may affect subtitle accuracy.
◉ YouTube automatic subtitles only support certain languages (such as English and Japanese). If the uploaded video is in Chinese, it will NOT generate Chinese subtitles.
◉ In summary, uploading to YouTube is a simple and free way to obtain English subtitles.
(1). Various Speech-to-Text Tools (Mostly Paid Services)
◉ Currently, most speech-to-text tools (which transcribe from audio/video files rather than live microphone input) require payment.
◉ Yating Transcription is developed in Taiwan and supports Taiwanese Hokkien; it offers 300 minutes of free transcription per Google account before requiring payment.
4. Create a New Transcription:
Click 【+ New Transcription】 ➜ Upload an audio/video file (supports Mandarin, English, and Taiwanese Hokkien)
Illustration
(1). Sign in to NotebookLM (Google's Professional Writing Tool, Free)
The entire Harry Potter novel series consists of about 1 million words.
NotebookLM can process 20 full sets of the Harry Potter series.
NotebookLM Feature 1: Converts uploaded notes into AI Podcast-style dual-person dialogues (only supports English audio).
NotebookLM Feature 2: Reads YouTube English videos to generate summaries, answer questions, and extract subtitles from specific time ranges (officially supports only videos with subtitles, but sometimes it can process videos without them).
NotebookLM Feature 3: Reads YouTube Chinese videos to generate summaries, answer questions, and extract subtitles from specific time ranges (currently the only tool capable of recognizing Chinese YouTube videos; however, videos uploaded within the last 24 hours or private videos cannot be processed).
NotebookLM Feature 4: Each created "notebook" acts as a "folder" where multiple reference files (PDF, Word, TXT), audio (MP3), video (MP4), YouTube URLs, and text content from web pages can be stored. It can also read recently uploaded Google Drive files.
NotebookLM Feature 5: Within a "notebook," you can input notes manually or generate AI-powered notes from uploaded files.
NotebookLM Feature 6: By adding up to 50 data sources in the left panel, you can generate AI notes that synthesize all the collected data.
NotebookLM Feature 7: Provides a built-in "Notebook Guide" feature to generate five common note-taking formats.
NotebookLM Feature 8: The five AI-generated notebook formats cannot be edited, but you can further refine them through interactive prompts.
NotebookLM Feature 9: Unlike other AI tools that may generate irrelevant or inaccurate content, NotebookLM strictly bases responses on the provided sources.
NotebookLM Feature 10: Responses always include citation links to the original sources you provided.
NotebookLM Feature 11: If you like an AI-generated response, you can click "Insert Note" to save it as a separate entry within your notebook.
NotebookLM Feature 12: Each notebook can store up to 1,000 note entries, and these entries cannot be modified once created.
NotebookLM Feature 13: The "Notebook Guide" feature allows AI to generate a dual-person AI Podcast (audio summary of your notebook contents).
1. Introduction to NotebookLM
◉ NotebookLM is an AI-powered research and writing assistant developed by Google.
◉ Purpose: Helps users explore and understand complex data more efficiently.
◉ Model: Powered by Google's Gemini 1.5 Pro AI model.
◉ Features: Speeds up draft writing, data organization, and summarization.
◉【Introduction Video: NotebookLM – AI-Powered Note Organization & Podcast Feature】
2. NotebookLM Use Cases:
◉ Students: Organize research materials and write papers.
◉ Researchers: Analyze data and compile reports.
◉ Writers: Gather writing inspiration and conduct research.
◉ Businesses: Knowledge management and document analysis.
3. Key Features of NotebookLM’s Notebook Guide: (1). Automated Content Summarization: When users upload documents (PDFs, Google Docs, etc.), NotebookLM analyzes and generates key takeaways, making it easier to grasp important information without reading the full text.
(2). Auto-Generated Q&A Sets: NotebookLM creates relevant questions and answers based on uploaded documents, helping users understand core concepts, key arguments, or data insights.
(3). Guided Document Reading & Analysis: Encourages users to focus on essential information and asks thought-provoking questions to aid comprehension and critical thinking.
(4). Learning Objectives & Summarization: NotebookLM helps set learning goals and compiles conclusions for future reference.
(5). Knowledge Transfer & Application: Converts data into structured learning notes, reports, or presentation content.
6. Practical Use Cases: ◉ Academic Research: Quickly understand academic materials with AI-generated Q&A.
◉ Knowledge Summarization: Converts complex documents into easy-to-read notes.
◉ Learning & Testing: Great for self-testing and knowledge retention.
◉ Report Writing: Helps structure reports and presentations.
4. Enter in the chat box: "Extract subtitles from this video between 00:01:22 and 00:05:00." Note: NotebookLM cannot process entire 2-hour videos at once. Extract subtitles in 10–20-minute segments instead. Enter: "Summarize the entire video into 10 key points." Illustration
5. Which Videos Cannot Be Recognized?
◉ It seems that older YouTube videos cannot be recognized by NotebookLM. Only videos uploaded after 2024 are supported. ◉ Example:
◉【Test Video 3: 2024 ESG Lecture】 (Recognized)
☎chp19. AI Talking Photo, AI Virtual Host, Marketing Video Template Creation
Vidnoz AI is an innovative AI-powered video creation platform designed for both individuals and businesses, enabling efficient and intelligent video production. This platform leverages AI technology to help users quickly generate high-quality video content without requiring professional video editing skills.
Key Features
AI-Powered Video Creation:
Provides multiple preset templates for quick professional video production.
Converts natural language into video scripts, simplifying content creation.
Text-to-Speech (TTS): Supports multiple languages and voice options, generating natural-sounding speech.
Virtual Character Generation: Allows virtual avatars to deliver presentations for educational content, product demonstrations, etc.
Automatic Subtitle Generation: Quickly generates subtitles for videos, with multi-language translation support.
Video Editing Tools: Includes basic editing, filters, effects, and more.
Use Cases: Online courses, business promotions, social media content creation.
Vidnoz AI Free User Features
Free Version Features
Basic Video Generation: Users can create up to 5 videos per month using preset templates.
Text-to-Speech: Supports limited languages and voices, allowing up to 10 minutes of speech generation per month.
Subtitle Function: Adds subtitles to short videos, limited to 2-minute clips.
Video Export: Free users' videos are watermarked with Vidnoz branding.
(2). Step-by-Step Usage Guide:
(A). Log in to the website ➜ After 3 seconds, it will automatically sign in ➜ Use your Google account to log in.
(B). Choose a video type ➜ Select any option ➜【Other】.
【Example Screenshot】 (C). Follow the four-step tutorial.
【Example Screenshot】 (D). You get 3 minutes of free video creation.
【Example Screenshot】 (E). Select 【AI Tools】 on the left.
【Example Screenshot】 (F). Choose 【AI Talking Photo】 (first option).
【Example Screenshot】 (G). Select 【Avatar】 and upload your own image (Note: You cannot upload celebrity images like Jay Chou, as the system will block them).
【Example Screenshot】 (H). Enter a text prompt (up to 300 words, about 1 minute):
I am Professor Chen Qingwen, currently teaching at the Department of Business Management at Taipei Tech.
The subjects I teach include: Digital Marketing Tools, E-Commerce, Customer Relationship Management, Business Ethics and Sustainable Management, Programming, Introduction to AI, Database Management, Business Models and Technological Innovation, AI in Business Management, AI in Market Forecasting, Internet Marketing, Business Multimedia Design, Business Data Analysis, and Journal Discussion—14 courses in total.
(I). Select 【chinese(TW)】 ➜ Choose 【Male】 ➜ Select 【Character】.
【Example Screenshot】 (J). Click 【Preview Audio】 (This only plays the voice, no animation).
(J). Click 【Generate Video】.
【Example Screenshot】 (K). The system will redirect to your account page ➜ Select 【My Creations】 on the left ➜ Wait 3–5 minutes for the video to be generated.
【Example Screenshot】 (L). Select the video ➜ Click 【Download】 (Free version exports at 720p resolution).
【Example Screenshot】
Vidnoz AI is an innovative AI-powered video creation platform designed for both personal and business users. It integrates AI technology to enable quick and high-quality video production without requiring professional video editing skills.
Key Features
AI-Powered Video Generation:
Provides various pre-designed templates for professional video creation.
Converts natural language into video scripts for easy content production.
Text-to-Speech (TTS): Supports multiple languages and voice selections to generate natural, fluent speech.
Virtual Avatar Creation: Offers virtual characters for presentations, suitable for course creation, product demos, and more.
Automatic Subtitle Generation: Quickly generates subtitles for videos and supports multi-language translations.
Applicable Scenarios: Educational courses, business promotions, social media content creation.
Free Version Limitations
Video Length Limit: Each video cannot exceed 3 minutes. The 3-minute quota refreshes every 24 hours.
Export Limit: Maximum of 3 video exports per month.
Export Quality: Only supports 720p resolution.
Template Usage: Only basic templates are available; premium templates require an upgrade.
(2). Step-by-Step Guide:
(A). Go to the website ➜ After 3 seconds, it will automatically sign in ➜ Use a Google account to log in
(B). "What type of video will you create?" ➜ Select any option ➜ Choose 【Other】
【Illustration】 (E). Select 【Template】 from the left panel
【Illustration】 (F). Choose 【Business】 ➜ 【Employee Benefit Overview】
【Illustration】 (G). Click 【Use This Template】
【Illustration】 (H). Click the 【Avatar】 on the screen ➜ Click the 【Avatar】 tab at the top ➜ Choose 【Upload Your Own Avatar】 ➜ Select 【Talking Photo】 ➜ Upload your own photo
【Illustration 1】,
【Illustration 2】,
【Illustration 3】 (I). Modify 【Your Avatar】 ➜ 【Choose Transparent Background, Round Frame, or Square Frame】
【Illustration】,
【Color Palette ➜ Set Background Transparency】 (J). Move the last slide to the first position ➜ Apply your avatar ➜ Modify 【Your Avatar】 ➜ 【Apply the New Avatar to All Scenes】
【Illustration 1】,
【Illustration 2】
(K). Download the presentation content: SEO Image Processing Methods
【Download PPT File】
(L). Enter the title for the first slide:
SEO Image Processing Methods
Supplement for the last class session that was not recorded
(M). Enter the first slide's 【Voiceover Subtitle】:
The final exam will cover both keyword SEO and image SEO. Since the last 10 minutes of the previous class were not recorded and some students left early for work, here is a brief supplement.
(N). Voiceover ➜ Choose language 【Chinese (TW)】 ➜ Select gender 【Male】 ➜ Choose 【YunJhe (Taiwan)】:
【Illustration 1】,
【Illustration 2】 (O). Click 【Preview/Play】 to test the audio, but the avatar will not move its mouth unless you generate the final video.
【Illustration】
(P-2). Follow the second slide of the 【PPT Presentation】, set the 【Title and Voiceover Subtitle】, and then preview.
【Illustration】 (P-3). Follow the third slide of the 【PPT Presentation】, copy the image into the template as a title, and add the voiceover subtitle, then preview.
【Illustration】 (P-6). Follow the sixth slide of the 【PPT Presentation】, copy the image as the title and add a subtitle, then preview.
【Illustration】
(Q). Finally, click 【Generate】 in the top-right corner ➜ It will redirect to the homepage ➜ Select 【My Creations】 from the left panel ➜ Wait approximately 15 minutes for the video to be generated.
【Illustration】
Vidnoz AI is an innovative AI-powered video creation platform designed for both personal and business users. It integrates AI technology to enable quick and high-quality video production without requiring professional video editing skills.
Key Features
AI-Powered Video Generation:
Provides various pre-designed templates for professional video creation.
Converts natural language into video scripts for easy content production.
Text-to-Speech (TTS): Supports multiple languages and voice selections to generate natural, fluent speech.
Virtual Avatar Creation: Offers virtual characters for presentations, suitable for course creation, product demos, and more.
Automatic Subtitle Generation: Quickly generates subtitles for videos and supports multi-language translations.
Applicable Scenarios: Educational courses, business promotions, social media content creation.
Free Version Limitations
Video Length Limit: Each video cannot exceed 3 minutes. The 3-minute quota refreshes every 24 hours.
Export Limit: Maximum of 3 video exports per month.
Export Quality: Only supports 720p resolution.
Template Usage: Only basic templates are available; premium templates require an upgrade.
(2). Step-by-Step Guide:
(A).【Download the prepared PPT file】 (A). Go to the website ➜ After 3 seconds, it will automatically sign in ➜ Use a Google account to log in
(B). "What type of video will you create?" ➜ Select any option ➜ Choose 【Other】
【Illustration】 (E). Click 【Create Video】 at the top right corner
【Illustration】 (F). Select 【Import PPT】 ➜ Upload 【Prepared PPT File: SEO Image Optimization Methods】
【Illustration】 (G). Adjust any misaligned text layouts (e.g., text that is too large and extends beyond the slide width)
【Illustration】 (H). Add your previously created 【Avatar】 ➜ Click 【Avatar】 at the top ➜ Select 【Upload Your Own Avatar】 ➜ Choose 【Talking Photo】 ➜ Upload your own image
【Illustration】 (I). Modify 【Your Avatar】 ➜ Choose 【Transparent Background, Round Frame, or Square Frame】
【Illustration】,
【Color Palette ➜ Set Background Transparency】
(J-1). Enter voiceover text for slide 1:
The final exam will cover both keyword SEO and image SEO. Since the last 10 minutes of the previous class were not recorded and some students left early for work, here is a brief supplement.
(J-2). Enter voiceover text for slide 2:
Why should we optimize images for SEO? Here are four reasons:
1. Helps search engines understand images better through alt and title text.
2. Improves search rankings and traffic; optimized images rank higher in image search.
3. Enhances user experience (especially when images fail to load).
4. Supports accessibility, helping visually impaired users via screen readers.
(J-3). Enter voiceover text for slide 3:
When optimizing images for SEO, configure three key img attributes:
1. SRC - Sets the image filename.
2. ALT - Defines alternative text when the image is unavailable.
3. TITLE - Displays a tooltip when hovering over the image.
(J-4). Enter voiceover text for slide 4:
The SRC attribute should contain an English filename, helping search engines better interpret the image's content.
(J-5). Enter voiceover text for slide 5:
The ALT attribute (short for "Alternative") serves as replacement text when an image fails to load due to network issues, file deletion, or for visually impaired users relying on screen readers.
(J-6). Enter voiceover text for slide 6:
The TITLE attribute displays a tooltip when users hover over an image, improving the browsing experience.
(K). Set voiceover settings for each slide ➜ Select language 【Chinese (TW)】 ➜ Choose gender 【Male】 ➜ Select 【YunJhe (Taiwan)】:
【Illustration 1】,
【Illustration 2】 (L). Click 【Preview/Play】 to test the audio. The avatar’s mouth will not move until you generate the final video.
【Illustration】
(M). Finally, click 【Generate】 at the top right ➜ The page will redirect to the homepage ➜ Select 【My Creations】 from the left panel ➜ Wait approximately 15 minutes for video generation.
【Illustration】
2020: DeepSeek was developed by a technology innovation company specializing in AI and big data, aiming to address the complexity and technical barriers in data analysis.
Inspiration: The founding team noticed that many small and medium-sized businesses, as well as non-technical users, struggled to extract insights from data, leading to the development of an AI-driven analytical tool powered by natural language.
(3). Key Features
Automated Chart Generation: Supports various chart types (line charts, bar charts, scatter plots, etc.), allowing users to choose the most suitable visualization.
Data Import & Cleaning: Supports multiple data formats (CSV, Excel, JSON, etc.) with built-in data cleaning capabilities to quickly process errors.
Efficient Sharing Features: Analysis results can be exported as PPT, PDF, or embedded directly into websites for easy sharing and reporting.
Open Source Model: The code is publicly available and completely open source.
(4). DeepSeek's Strengths:
Feature Comparison vs. Other Generative AIs: Includes DeepSeek, ChatGPT-4o, Claude, and Meta's LLaMA.
Math500 Competition (Evaluating Mathematical Reasoning Ability): DeepSeek outperformed ChatGPT-4o, Claude, and LLaMA.
【Example Screenshot】AIME 2024, American Mathematics Competition: DeepSeek demonstrated the best performance.
【Example Screenshot】CodeForce (Evaluating Coding Ability): DeepSeek performed the best.
【Example Screenshot】 MMLU-Pro (Multi-Task Understanding Ability): DeepSeek outperformed competitors.
GPQA-Diamond (Complex Problem-Solving Ability): DeepSeek achieved the highest scores.
(5). DeepSeek's Cost Advantages:
DeepSeek uses lower-tier GPUs but requires only 1/10 of the training time while achieving the best results. DeepSeek's training cost is only 1/100 of other AI models, yet it delivers superior performance. DeepSeek API pricing is 1/10 to 1/100 of other generative AI models, making it an attractive choice for enterprises building AI solutions.
(6). DeepSeek's Weaknesses:
DeepSeek's context memory is limited to 130,000 tokens, which is only 1/17 of other AIs. This means that after answering 20 consecutive questions, DeepSeek might forget the first question and require a new conversation.
(7). Numerical Comparison Test:
Question: Which is larger, 4.11 or 4.9? Test Objective: Evaluate numerical computation ability.
1. ChatGPT-4o has multiple features in data analysis, including:
◉ Data cleaning,
◉ Exploratory data analysis,
◉ Statistical testing,
◉ Data visualization,
◉ Ability to analyze data using natural language prompts without coding,
◉ Merging datasets,
◉ Cleaning datasets,
◉ Creating charts and uncovering data insights,
◉ ChatGPT-4o can assist Excel by generating formulas, providing code samples, and automating data processing.
2. While ChatGPT-4o is powerful in data analysis, it cannot fully replace Excel.
◉ Excel remains the primary tool for storing spreadsheets in businesses.
◉ However, ChatGPT-4o can replace Excel's formula calculations, eliminating the need to memorize formulas.
◉【Download Original Excel File】: Student Score Data ◉【Goal】: Display the average score of the "Chinese" subject in row 8 under the "Chinese" column.
◉【Input into ChatGPT】: Modify directly on the screen table and display the average Chinese score in row 8 under the "Chinese" column. ◉【Calculation Results: Image】 ◉【Key Note 1】: Specify where to modify (on-screen vs. in file). If not specified, ChatGPT may modify the file and provide a download link instead. ◉【Key Note 2】: Instead of traditional Excel notation (e.g., D8), use "Row 8 under the 'Chinese' column" as reference. ◉【Conclusion 1】: If specific locations can be modified, ChatGPT can replace Excel formula calculations anywhere (e.g., copying formulas, using statistical functions).
◉【Conclusion 2】: Display modifications on-screen first, then download as an Excel file when finalized.
(3). Using ChatGPT-4o: Automatically Generating a Bar Chart
◉ Upload the original Excel sample file.
◉Select the 【⟷】 symbol for expanded data analysis.
◉【Input into ChatGPT】: Generate a bar chart comparing the three subject scores of all students.
◉【Chart Output: Bar Chart】
◉【Issue Encountered】: Chinese characters were not displayed properly in the chart.
(4). Uploading Chinese Font Files to ChatGPT-4o
◉【Noto Sans TC Font Family (9 Variants)】 - Official Download Link
◉【Recommended Font: NotoSansTC-Medium】 - Download Link
◉【About NotoSansTC-Medium】: This is one of the nine font weights in the Noto Sans TC font family.
Noto Sans TC is a free and open-source font created by Google and Adobe, designed specifically for Traditional Chinese.
It supports multiple languages including Traditional Chinese, Japanese, and Korean, ensuring consistent readability across platforms.
This font aims to resolve display issues between different languages and fonts, ensuring proper character rendering on all devices.
◉【What is Font Weight?】: It refers to the thickness of the strokes in a font, providing various design options for different needs.
◉【Note】: The font file comes in a compressed format and needs to be extracted first.
◉【Uploading NotoSansTC-Medium to ChatGPT-4o】
◉【Input into ChatGPT】: Use this uploaded font file to regenerate the chart with all Chinese characters displayed using this font.
◉【Chart Output: Bar Chart with Proper Chinese Characters】
(6). Using ChatGPT-4o: Automatically Generating a Line Chart
◉ Upload the original Excel sample file.
◉ Ensure the 【⟷】 symbol is selected for expanded data analysis.
◉【Input into ChatGPT】: Generate a line chart comparing the three subject scores of all students.
◉【Chart Output: Line Chart】
◉【Issue Encountered】: The text on the X-Y axis is too small to read.
◉【Input into ChatGPT】: Increase the font size of the X-Y axis labels by 2x.
◉【Chart Output: Line Chart with Enlarged Axis Labels】
(7). Using ChatGPT-4o: Automatically Generating a Pie Chart
◉ Upload the original Excel sample file.
◉ Ensure the 【⟷】 symbol is selected for expanded data analysis.
◉【Input into ChatGPT】: Generate a pie chart for math scores, highlight the highest score, and display percentage shares.
◉【Chart Output: Pie Chart】
◉ Upload the original Excel file.
◉【Input in ChatGPT】: Generate a heat map to show correlations between parameters.
◉【Chart Output: Heat Map】
(4). Using ChatGPT-4o: Auto-Generate Pair Plot
◉ Upload the original Excel file.
◉【Input in ChatGPT】: Generate a pair plot showing the relationships between all parameters.
◉【Chart Output: Pair Plot】
◉【Issue】: Pair plot does not display Chinese fonts.
(5). Using ChatGPT-4o: Auto-Generate Pair Plot for Two Features
◉ Upload the original Excel file.
◉【Input in ChatGPT】: Generate a pair plot for two specific features: Petal Length vs. Sepal Width.
◉【Chart Output: Pair Plot for Two Features】
◉【Issue】: Pair plot does not display Chinese fonts.
◉ Upload the original Excel file.
◉【Input in ChatGPT】: Compare sales revenue across different business units.
◉【Cluster Analysis Chart Output】
(4). Using ChatGPT-4o: Two-Group Cross-Analysis
◉ Upload the original Excel file.
◉【Input in ChatGPT】: Perform a cross-analysis comparing different sales products and business units based on sales revenue.
◉【Cross-Analysis Chart Output】
(5). Using ChatGPT-4o: Annual Sales Trend Chart
◉ Upload the original Excel file.
◉【Input in ChatGPT】: Generate a year-over-year sales revenue comparison chart.
◉【Annual Sales Trend Chart Output】
(3). Basic Analysis Using ChatGPT-4o: Predict Customer Importance
◉ Upload the original Excel file.
◉【Python Model Prediction Result】
◉【Input in ChatGPT】:
Train a logistic regression model to predict customer value in CRM.
Predict if customer A with [R, F, M] = (5,2,2100) is an important customer.
◉【ChatGPT-4o Prediction Result】
◉【Comparison Result】: ChatGPT-4o prediction matches Python analysis.
(3). Statistical Analysis Using ChatGPT-4o: Calculate the average weekly beverage expenses for males and females, then perform an independent sample t-test to determine if there is a significant difference.
◉ Upload the original CSV file.
◉【SPSS Analysis Result】
◉【Input in ChatGPT】:
Calculate the average weekly beverage expenses for males and females, then perform an independent sample t-test to determine if there is a significant difference.
◉【ChatGPT-4o Analysis Result】
◉【Comparison Result】: ChatGPT-4o analysis matches SPSS results.
(3). Multiple Regression Analysis Using ChatGPT-4o: Investigate the Relationship Between Credit Score and Total Income, Real Estate, Movable Assets, Mortgage, and Dependents
☎(1). Practical SEO Implementation 1: Introducing Your Website Summary (TDK) to Search Engines
T stands for Title, D stands for Description, and K stands for Keywords.
These three elements are the fundamental and crucial parts of webpage SEO optimization.
◉<title>Chen Qingwen Assignment 1 - Programming Assignment Showcase</title> ◉<meta name="description" content="This is Chen Qingwen's Assignment 1, including code and related descriptions."> ◉<meta name="keywords" content="Chen Qingwen, Assignment 1, SEO, TDK">
(2). About: Webpage Keywords name=keywords
Google announced in 2009 that Meta keywords would no longer be a ranking factor. However, some smaller search engines might still reference it. Therefore, you can add the Meta keywords tag, but its impact on SEO is limited.
33-4. After Google Search Console recognizes your website,
【Test】site:https://ccwntut.github.io/hw1
Check if your website has been indexed by Google.
Example Image 1
33-5. Enter the website title in Google search (NTUT Chen Qingwen Assignment 1). Can you find your website?
Example Image 1
33-6. If your website does not appear when searching for "NTUT Chen Qingwen Assignment 1," possible reasons and solutions are as follows:
Reason: Insufficient keyword occurrences, poor keyword optimization Why can't your website be found when searching for keywords?
◉ The keyword appears too few times. ◉ Other webpages are better optimized for similar keywords, so Google has not prioritized displaying your site. Solution:
◉ Ensure your keyword "NTUT Chen Qingwen Assignment 1" appears not only in the page title but also in the content (such as headings and paragraphs) to improve keyword relevance. ◉ Check if the page includes an appropriate meta description and that the keyword is present in the meta tags.
Google’s Ranking Algorithm: Considers External Backlinks and Overall Website Traffic.
【Common Issue for New Websites】: Since the site is new, it lacks backlinks and traffic, leading to poor search rankings or difficulty in being found by Google. ◉ Solution: Building Backlinks:
◉ Acquire External Links: Links from high-authority websites (backlinks) significantly improve Google rankings and help Google discover and index your site faster.
◉ Social Media Promotion: Share your website links on social media platforms (Facebook, Instagram, Xiaohongshu, etc.).
【Response Time】: Generally takes a few weeks for Google to display keyword searches. Google indexing and ranking adjustments take time. Even if a site is indexed, it may take several weeks before appearing in search results for relevant keywords.
Solution:
◉ Be patient and continuously optimize content and SEO strategies.
41-6. Implementation Steps: ◉ Upload 【GA41-1】 website to GitHub
◉ Create a 【GA4 Account】: Chen Qingwen
◉ Create a 【Property】: GA41-1
◉ Create a 【Data Stream】: Website
◉ Create a Data Stream Subdomain: ccwntut.github.io/GA41-1
◉ Copy the Google tracking code and paste it into the
section of each webpage
◉ Check if GA4 code is installed successfully (view real-time reports or use Google Tag Assistant for verification)
Example ◉ 【Reports】 ➜ 【Real-Time Overview】 ◉ 【Common Events】:
◉ page_view (Page Views) = 9 times
◉ user_engagement (User Engagement) = 7:
➜➜ 7 user interactions (staying over 10 seconds or switching between pages).
➜➜ Reflects user engagement on the website.
◉ first_visit (First Visit) = 1:
➜➜ One new user visited the website for the first time, reflecting new user acquisition.
◉ session_start (Session Start):
➜➜ Triggered when a user visits the website, marking a new session.
➜➜ If no action is taken for 30 minutes, GA assumes the session has ended.
➜➜ Represents the total visits of an e-commerce site.
◉ 【Reports】 ➜ 【Real-Time Webpage】 Example ➜➜ Active Users: "Unique User IDs"
➜➜ Page Views: Total page views (counting repeated views from the same user).
43-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is real-time e-commerce data from Google: 【Google Merchandise Store】 and 【Flood-IT APP】 e-commerce data.
◉【Limitations】 Google has hidden some key e-commerce data to prevent business secrets from being leaked.
◉【How to Access Data】 Search for: Google Analytics Demo Account
Example Image ◉【Download Demo E-commerce Data URL】 ◉ The demo account includes three data sources:
1-【Google Merchandise Store】 (Website Data)
2-【Flood-It! APP】 (App Data)
3-【Flood-It! Website】 (Website Data)
◉【How to Download the First Demo Data】: Click the first data source link ➜ Data is automatically added to your GA4 account ➜ Add 【Demo Account】
Example Image 1,
Example Image 2,
Example Image 3,
◉【Check Data Stream】: Settings ➜ Data Collection and Management ➜ Data Stream: Web
Example Image 4,
◉【How to Download the Second Demo Data】: Click the second data source link ➜ Data is automatically added to your GA4 account ➜ Add 【Demo Account】
Example Image 1,
Example Image 2,
Example Image 3,
◉【How to Switch Between Accounts】: Top-left corner ➜ All Accounts ➜ Dropdown Menu ➜ Recently Used
Example Image 1,
Example Image 2,
41-2. Sample Results
43-3. GA4 Acquisition Report (Daily Volume, Monthly Volume, Active Users, New Users, Returning Users, Traffic from Various Channels) ◉【Official E-commerce Website】: Google Merchandise Store ◉1. How many visitors came to your e-commerce website yesterday?
Example Image 1,
Example Image 2 ◉2. How many visitors came to your e-commerce website in the past month?
Example Image 1,
Example Image 2 ◉3. How many new users visited your e-commerce website yesterday?
Example Image 1,
◉4. How many new users visited your e-commerce website in the past 30 days?
Example Image 1,
◉5. What is the definition and meaning of active users?
Example Image 1,
Example Image 2 ◉6. What does the ratio of total users, new users, and active users represent?
Example Image 1,
◉7. Which channel brought the most traffic to your website yesterday?
Example Image ◉8. How many visitors came directly by entering the website URL?
Example Image ◉9. How many visitors came from organic search (Google, Bing, Yahoo) without clicking paid ads?
Example Image ◉10. How many visitors came from partner websites, forums, or news articles?
Example Image ◉11. How many visitors came from email links?
Example Image ◉12. How many visitors came from shopping platforms like Amazon or Google Shopping without paid ads?
Example Image ◉13. How many visitors came from paid search ads (Google Ads, Bing Ads)?
Example Image ◉14. How many visitors came from display ads (GDN, Google Display Network)?
Example Image ◉15. How many visitors came from social media (Facebook, Instagram, Twitter, LinkedIn)?
Example Image ◉16. How many visitors came from cross-network traffic (Google Search, YouTube, Google Discover, Gmail)?
Example Image ◉17. How many visitors came from uncategorized traffic sources in GA4?
Example Image
44-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is Google's real-time e-commerce data.
1-【Google Merchandise Store】 (Website Data)
◉【Download Demo E-commerce Data URL】 ◉【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add 【Demo Account】
◉【Actual E-commerce Website】: Google Merchandise Store,
44-3. GA4 Practical Implementation 4: GA4 User Report (Identify My Target Customers: Country, Region, City, Gender, Age) ◉1- Which country brought the most visitors to your e-commerce website in the past 30 days?
Example Image ◉2- Which city brought the most visitors to your e-commerce website in the past 30 days?
Example Image ◉3- Which gender visited your e-commerce website the most in the past 30 days?
Example Image ◉4- Which interest category had the most visitors to your e-commerce website in the past 30 days?
Example Image ◉5- Which age group visited your e-commerce website the most in the past 30 days?
Example Image ◉6- Which language was most commonly used by visitors to your e-commerce website in the past 30 days?
Example Image ◉7- What was the average user value over the past 30 and 120 days for your e-commerce website?
Example Image
44-4. Audience Details: ◉8- Which country had the most active users on your e-commerce website in the past 30 days?
Example Image ◉9- Which region had the most active users on your e-commerce website in the past 30 days?
Example Image ◉10- Which gender had the most new users on your e-commerce website in the past 30 days?
Example Image ◉11- Which age group had the most engaged sessions on your e-commerce website in the past 30 days?
Example Image ◉ What is engagement rate and how is it calculated?
Example Image ◉ What is an Engaged Session?
Example Image ◉ What is the practical use of Engagement Rate?
Example Image ◉12- What was the overall engagement rate of users on your e-commerce website in the past 30 days?
Example Image ◉13- How was the overall engagement rate calculated for users on your e-commerce website in the past 30 days?
Example Image ◉14- Which region had an engagement rate above 70% in the past 30 days?
Example Image ◉15- Which age group had an engagement rate above 70% in the past 30 days?
Example Image ◉16- Which gender had an engagement rate above 70% in the past 30 days?
Example Image ◉17- Select your target audience: 【Male, 35-44 years old】.
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5,
◉18- In which country did the target audience 【Male, 35-44 years old】 have an engagement rate above 70% in the past 30 days?
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5
45-2. Using GA4 Demo E-commerce Account: ◉ 【Data Source】 This is real-time e-commerce data from Google.
1- 【Google Merchandise Store】 (Website Data)
◉ 【Download Demo E-commerce Data】 ◉ 【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add a new 【Demo Account】.
◉ 【Actual E-commerce Website】 Google Merchandise Store
45-3. Practical GA4 Implementation 5: GA4 Technology Report (Identify the Most Used Devices for Browsing My Website, Monthly Design Checks: Platform, OS, Browser, Device Type, Screen Resolution)
◉ 1- In the past 30 days, which device category had the most users visiting your e-commerce website? (Desktop, Mobile, Tablet)
Image Example,
Image Example ◉ 2- In the past 30 days, what screen resolution was most used by visitors?
Image Example ◉ 3- In the past 30 days, which browser was most used by visitors?
Image Example ◉ 4- In the past 30 days, which operating system was most used by visitors?
Image Example ◉ 5- In the past 30 days, which platform had the most visitors? (Website, Android App, iOS App)
Image Example
45-4. Technical Details: ◉ 6- In the past 30 days, which device had the highest engagement?
Image Example
46-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is Google's real-time e-commerce data.
1-【Google Merchandise Store】 (Website Data)
◉【Download Demo E-commerce Data URL】 ◉【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add 【Demo Account】
◉【Actual E-commerce Website】: Google Merchandise Store,
46-3. GA4 Practical Implementation 6: GA4 Engagement Report (Identify the Most Popular Pages, Check Various Events, Conversion Events)
◉1- Method 1: Which three pages had the most views on your e-commerce website in the past 30 days?
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5 ◉2- Method 2: Which three pages had the most views on your e-commerce website in the past 30 days?
Example Image 1,
Example Image 2 ◉3- Which method should be used to check page views?
Example Image ◉4- Which browser was used most frequently to visit your e-commerce website in the past 30 days?
Example Image ◉5- What conversion events (key events) were set on your e-commerce website in the past 30 days?
Example Image 1,
Example Image 2 ◉6- Purpose of the "view_item" (view product) event
Example Image ◉7- Purpose of the "add_to_cart" (add to cart) event
Example Image ◉8- Purpose of the "purchase" (complete purchase) event
Example Image ◉9- These three events form a conversion funnel analysis
Example Image ◉10- Common e-commerce issue: Low add_to_cart conversion rate
Example Image ◉11- Common e-commerce issue: Low purchase conversion rate
Example Image ◉12- Important metrics provided by the conversion funnel (3 events)
Example Image ◉13- Optimization solutions for common conversion funnel issues
Example Image 1,
Example Image 2 ◉14- How many times was the "view_item" event triggered in the past 30 days?
Example Image ◉15- How many times was the "add_to_cart" event triggered in the past 30 days?
Example Image ◉16- How many times was the "purchase" event triggered in the past 30 days?
Example Image ◉17- Complete the conversion funnel analysis and calculate three conversion rates
Example Image ◉18- How many purchases were made from an ad campaign (Nov2024_Holiday_gift_V1)?
Example Image 1,
Example Image 2,
Example Image 3
47-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is Google's real-time e-commerce data.
1-【Google Merchandise Store】 (Website Data)
◉【Download Demo E-commerce Data URL】 ◉【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add 【Demo Account】
◉【Actual E-commerce Website】: Google Merchandise Store,
47-3. GA4 Practical Implementation 7: GA4 Revenue Report (Identify Best-Selling, Most Viewed, and Most Added to Cart Products, Revenue/Purchases/New Buyers in the Last 30 Days)
◉1- What was your total revenue for the past 30 days?
Example Image ◉2- How many total purchases and first-time buyers were there in the past 30 days?
Example Image ◉3- What was the average revenue per active user in the past 30 days?
Example Image 1,
Example Image 2 ◉4- Which product had the highest sales volume in the past 30 days, and how many units were sold?
Example Image ◉5- Which product had the highest number of views in the past 30 days?
Example Image ◉6- Which product was added to the cart the most times in the past 30 days?
Example Image 1,
Example Image 2 ◉7- Which product had the highest sales volume in the past 30 days? How many units were sold?
Example Image ◉ Which product generated the highest revenue in the past 30 days?
Example Image
48-2. Using GA4 Demo E-commerce Account: ◉ 【Data Source】 This is real-time e-commerce data from Google.
1- 【Google Merchandise Store】 (Website Data)
◉ 【Download Demo E-commerce Data】 ◉ 【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add a new 【Demo Account】.
◉ 【Actual E-commerce Website】 Google Merchandise Store
48-3. Practical GA4 Implementation 8: GA4 Returning User Rate Report (Check returning users, daily retention and engagement time for new users 42 days ago, trend changes in retention rates between new and returning users, trend changes in engagement time between new and returning users)
◉ 1- How many returning users did your e-commerce website have in the past 30 days?
Illustration ◉ 2- Daily retention rate of new users from 42 days ago.
Illustration ◉ 3- What is user retention?
Illustration ◉ 4- How to improve user retention?
Illustration 1,
Illustration 2 ◉ 5- Daily engagement time of new users from 42 days ago.
Illustration ◉ 6- Evaluation of four key user engagement metrics.
Illustration ◉ 7- How to improve user engagement?
Illustration 1,
Illustration 2 ◉ 8- 【User retention trend (by cohort segmentation)】: Compare trend changes in retention rates between new and returning users.
Illustration 1,
Illustration 2,
Illustration 3,
Illustration 4,
Illustration 5,
Illustration 6,
Illustration 7 ◉ 9- 【User retention rate】: Meaning of the blue vs purple curve.
Illustration 1,
◉ 10- 【User engagement time trend (by cohort segmentation)】: Compare trend changes in engagement time between new and returning users.
Illustration 1,
Illustration 2,
Illustration 3,
Illustration 4,
Illustration 5,
Illustration 6 ◉ 11- 【User engagement time】: Meaning of the blue vs purple curve.
Illustration ◉ 12- What is the average value of your e-commerce website over the past 30 or 120 days?
Illustration
49-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is Google's real-time e-commerce data.
1-【Google Merchandise Store】 (Website Data)
◉【Download Demo E-commerce Data URL】 ◉【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add 【Demo Account】
◉【Actual E-commerce Website】: Google Merchandise Store
49-3. GA4 Practical Implementation 9: GA4 Exploration Custom Reports ➜ Free-Form Reports (1) ➜ Displaying users from the USA, Canada, and Taiwan through different channels, comparing 【Product Views, Add to Cart, Transactions, Total Purchase Revenue, Purchase Conversion Rate】
◉1- Exploration ➜ Free-Form
Example Image ◉2- Add 3 【Segment】 Variables: USA, Canada, Taiwan
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5,
Example Image 6,
Example Image 7,
Example Image 8,
Example Image 9 ◉ Drag 3 【Segment】 Variables: USA, Canada, Taiwan ➜ Drag to 【Segment Settings】
Example Image ◉4- Add 1 【Dimension】 Variable: First User Source/Medium
Example Image 1,
Example Image 2 ◉5- Drag 1 【Dimension Variable】: First User Source/Medium ➜ Drag to 【Rows】 and delete 【City】
Example Image ◉6- Add 3 【Metric】 Variables: Add to Cart, Cart Items, Transactions
Example Image 1,
Example Image 2 ◉7- Add 5 【Metric】 Variables: Product Views, E-commerce Purchases, Quantity, Total Purchase Revenue, Purchase Revenue
Example Image ◉8- Add 1 【Metric】 Variable: Total Revenue
Example Image ◉9- Add 1 【Metric】 Variable: Total Buyers
Example Image ◉10- Question: What's the difference between similar metrics?
1. Total Purchase Revenue, Purchase Revenue, Total Revenue
2. E-commerce Purchases, Transactions, Quantity
Example Image ◉11- How to understand 【GA4 Dimensions and Metrics】 definitions?
GA4 Dimensions and Metrics Glossary ◉12- Delete 【Column Settings】: Device Category
Example Image 1,
Example Image 2 ◉13- Drag 【Metrics】 Product Views, Add to Cart, Transactions, E-commerce Purchases, Quantity, Total Buyers ➜ Drag to 【Values】 and remove Active Users
Example Image ◉14- Compare the differences:
Transactions,
E-commerce Purchases,
Quantity,
Total Buyers
◉15- Conversion Funnel Analysis: What are the three key metrics?
Example Image 1,
Example Image 2 ◉16- If not using Google's default three conversion events, which three metrics can be used for a similar conversion funnel analysis?
Example Image ◉17- Compare 【Transactions, E-commerce Purchases, Total Buyers, Quantity】 and their differences
Example Image 1,
Example Image 2 ◉18- In 【Values】, remove the metrics 【E-commerce Purchases, Total Buyers, Quantity】
Example Image ◉19- Drag 【Metrics】 Total Purchase Revenue, Purchase Revenue, Total Revenue ➜ Drag to 【Values】
Example Image 1,
Example Image 2 ◉20- Delete 【Values】 Purchase Revenue, Total Revenue
Example Image
50-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is Google's real-time e-commerce data.
1-【Google Merchandise Store】 (Website Data)
◉【Download Demo E-commerce Data URL】 ◉【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add 【Demo Account】
◉【Actual E-commerce Website】: Google Merchandise Store
50-3. GA4 Practical Implementation 10: GA4 Exploration Custom Reports ➜ Free-Form Reports (2) ➜
Exploring Different Products (Using Web Page Titles) ➜
Through Different Traffic Channels 【Direct Traffic, Paid Traffic, Organic Traffic】 ➜
Comparing 【Product View Events, Add to Cart Events】
◉1- Exploration ➜ Free-Form
Example Image ◉2- First, create the report framework:
A. Exploring Different Products (Using Web Page Titles)
B. Through Different Traffic Channels 【Direct Traffic, Organic Traffic】
◉3- Drag and Drop 2 【Segment Variables】: Direct Traffic, Paid Traffic ➜ Move to 【Segment Settings】
Example Image ◉4- Add 1 【Dimension】 Variable: Web Page Title (Represents Product Name)
Example Image ◉5- Drag 1 【Dimension Variable】: Web Page Title ➜ Move to 【Rows】 and delete 【City】
Example Image ◉6- Delete 【Column Settings】: Device Category
Example Image ◉7- Add 2 【Metric】 Variables: Product View Events, Add to Cart
Example Image ◉8- Drag 【Metrics】 Product View Events, Add to Cart ➜ Move to 【Values】 and remove Active Users
Example Image ◉9- Observe the default 【Segment】 Variable: Direct Traffic Configuration
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5 ◉10- Manually Create the Same 【Segment】 Variable: Direct Traffic
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5 ◉11- Drag 1 【Segment Variable】: Direct Traffic 2 ➜ Move to 【Segment Comparison】
Example Image ◉12- Question: Why is the custom segment variable Direct Traffic 2 different from the default Direct Traffic?
Example Image 1,
Example Image 2 ◉13- Observe the default 【Segment】 Variable: Paid Traffic Configuration
Example Image 1,
Example Image 2,
Example Image 3 ◉14- Manually Create the Same 【Segment】 Variable: Paid Traffic 2
Example Image 1,
Example Image 2 ◉15- Observe the default 【Segment】 Variable: Organic Traffic Configuration
Example Image 1,
Example Image 2 ◉16- Manually Create the Same 【Segment】 Variable: Organic Traffic 2
Example Image 1,
Example Image 2 ◉17- Compare the Top 5 Products 【Viewed】 via Direct, Paid, and Organic Traffic
Direct Traffic Top 5,
Organic Traffic Top 5,
Paid Traffic Top 5 ◉18- Compare the Top 5 Products 【Added to Cart】 via Direct, Paid, and Organic Traffic
Direct Traffic Top 5,
Organic Traffic Top 5,
Paid Traffic Top 5 ◉19- Add 1 【Metric】 Variable: E-commerce Purchases and move to 【Values】
Example Image
51-2. Using GA4 Demo E-commerce Account: ◉ 【Data Source】 This is real-time e-commerce data from Google.
1- 【Google Merchandise Store】 (Website Data)
◉ 【Download Demo E-commerce Data】 ◉ 【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add a new 【Demo Account】.
◉ 【Actual E-commerce Website】 Google Merchandise Store
51-3. Practical GA4 Implementation 11: GA4 Exploration Custom Reports ➜ Free-form Reports (3) ➜ Which type of advertisement performs the best?
Analyzing different paid advertising mediums (Google Search Ads, Google Display Ads, YouTube Video Ads, Facebook Social Media Ads)
➜ Comparing marketing performance: 【Active Users, Product View Events, Add to Cart Events, Purchase Conversion Rate】
◉ 1- Exploration ➜ Free-form
Illustration ◉ 2- Create a report structure:
A. Analyze different paid advertising mediums (Google Search Ads, Google Display Ads, YouTube Video Ads, Facebook Social Media Ads)
B. Compare marketing performance: 【Active Users, Product View Events, Add to Cart Events, Purchase Conversion Rate】
◉ 3- Create four 【Segment Variables】: Google Ads, YouTube Ads, Google Display Ads, Facebook Ads ➜ Drag to 【Segment Setup】
Illustration 1,
Illustration 2,
Illustration 3,
Illustration 4,
Illustration 5 ◉ 4- Why can't 【Paid Social, Paid Video, Display】 be displayed?
Illustration 1,
Illustration 2 ◉ 5- What paid-related UTM parameters are set by the advertiser?
Illustration ◉ 6- What is Cross-network?
Illustration 1,
Illustration 2,
Illustration 3,
Illustration 4 ◉ 7- What is Google Discover Ads?
Illustration 1,
Illustration 2 ◉ 8- Yahoo News page contains two types of ads, what are they?
Illustration 1,
Illustration 2,
Illustration 3,
Illustration 4 ◉ 9- What is an App Campaign?
Illustration 1,
Illustration 2 ◉ 10- What is Gmail Ads?
Illustration 1 ◉ 11- Since Cross-network already includes Google Search Ads (Paid Search), why is there still a separate Paid Search option?
Illustration 1,
Illustration 2,
Illustration 3
52-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is Google's real-time e-commerce data.
1-【Google Merchandise Store】 (Website Data)
◉【Download Demo E-commerce Data URL】 ◉【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add 【Demo Account】
◉【Actual E-commerce Website】: Google Merchandise Store
52-3. GA4 Practical Implementation 12: GA4 Exploration Custom Reports ➜ Funnel Exploration Report (4)
➜ Analyzing Conversion Rate Changes in Three Key E-commerce Steps
➜ Conversion Funnel Analysis
➜ view_item ➜ add_to_cart ➜ purchase
➜ Understanding Google's 9 Types of Ads
◉1- What is Conversion Funnel Analysis?
Example Image 1,
Example Image 2,
Example Image 3 ◉2- Exploration ➜ Funnel Exploration Report
Example Image 1,
Example Image 2 ◉3- First, create the report framework:
A. Analyze conversion rate changes in key e-commerce steps.
B. Conversion Funnel Analysis
C. view_item ➜ add_to_cart ➜ purchase
◉4- Create 2 【Segment Variables】: USA, Non-USA ➜ Drag to 【Segment Settings】
Example Image 1,
Example Image 2 ◉5- Create 3 Funnel Steps 【Key E-commerce Steps】: view_item, add_to_cart, purchase ➜ Drag to 【Steps】
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5 ◉6- Remove 【Breakdown】: Device Category
Example Image ◉7- Complete Conversion Funnel Analysis (Key E-commerce Steps' Conversion Rate Change): Compare USA vs. Non-USA Conversion Rates
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4 ◉8- Show users who dropped off during the key three steps in the funnel—what actions (events) did they take?
◉9- Drag 【Dimension】 【Event Name】 into 【Next Step】
Example Image 1,
Example Image 2,
Example Image 3 ◉10- Show users who dropped off during the key three steps in the funnel—what web pages did they visit?
◉11- Drag 【Dimension】 【Page Title】 into 【Next Step】
Example Image 1,
Example Image 2 ◉12- 【Open Funnel vs. Closed Funnel】
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5 ◉13- In an e-commerce checkout funnel analysis, should you choose an 【Open Funnel or a Closed Funnel】?
Example Image ◉14- In a news website funnel analysis, should you choose an 【Open Funnel or a Closed Funnel】?
Example Image ◉15- Enable the 【Open Funnel】 Option
Example Image 1,
Example Image 2 ◉16- Compare 【Open Funnel vs. Closed Funnel】: Which has a higher conversion rate in the USA?
Example Image 1,
Example Image 2,
Example Image 3
53-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is Google's real-time e-commerce data.
1-【Google Merchandise Store】 (Website Data)
◉【Download Demo E-commerce Data URL】 ◉【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add 【Demo Account】
◉【Actual E-commerce Website】: Google Merchandise Store
53-3. GA4 Practical Implementation 13: GA4 Exploration Custom Reports ➜ Path Exploration Report (5)
A. Visitor Journey Analysis
B. Forward Path Analysis: What Pages Do Visitors Browse After Entering the Homepage?
◉1- Exploration ➜ Path Exploration Report
Example Image 1,
Example Image 2 ◉2- First, create the report framework:
A. Visitor Journey Analysis
B. Forward Path Analysis: What Pages Do Visitors Browse After Entering the Homepage?
◉3- Observe Path Analysis Nodes 【Node Type】 ➜ Fixed, Cannot Be Added
Example Image ◉4- Delete Default Settings and Restart ➜ 【Restart】
Example Image ◉5- Set the Starting Point for 【Forward Path】 ➜ Select 【Event Name】 ➜ session_start (Session Start)
Example Image ◉6- Understand What Pages Customers Browse After the Starting Point ➜ Step 1: Select 【Page Title Name】
Example Image 1,
Example Image 2 ◉7- Understand the Series of Pages with the Most Traffic After the Starting Point ➜ Home ➜ ???
Example Image 1,
Example Image 2 ◉8- User ➜ Session ➜ Event ➜ Conversion
Example Image 1,
Example Image 2 ◉9- Explanation of GA4 Sessions
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4
54-2. Using GA4 Demo E-commerce Account: ◉ 【Data Source】 This is real-time e-commerce data from Google.
1- 【Google Merchandise Store】 (Website Data)
◉ 【Download Demo E-commerce Data】 ◉ 【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add a new 【Demo Account】.
◉ 【Actual E-commerce Website】 Google Merchandise Store
54-3. Practical GA4 Implementation 14: GA4 Exploration Custom Reports ➜ Path Exploration Reports (6) ➜ Reverse Path Analysis
➜ Visitor Journey Analysis
➜ Reverse Path Analysis: What pages do visitors browse before placing an order?
◉ 1- Exploration ➜ Path Exploration Reports
Illustration 1,
Illustration 2 ◉ 2- Create a report structure:
A. Visitor Journey Analysis
B. Reverse Path Analysis: What pages do visitors browse before placing an order?
◉ 3- Observe path analysis nodes 【Node Type】 ➜ Fixed, cannot be added
Illustration ◉ 4- Reset all default settings ➜ 【Restart】
Illustration ◉ 5- Set the 【Reverse Path】 endpoint ➜ Select 【Event Name】 ➜ purchase (Order Completion)
Illustration 1,
Illustration 2 ◉ 6- Understand what pages customers browse before the endpoint ➜ Step 1 Select 【Page Title Name】
Illustration 1,
Illustration 2 ◉ 7- Identify the most frequently visited pages before order completion ➜ Purchase ➜ ???
Illustration 1,
Illustration 2 ◉ 8- Users ➜ Sessions ➜ Events ➜ Conversions
Illustration 1,
Illustration 2 ◉ 9- GA4 Session Explanation
Illustration 1,
Illustration 2,
Illustration 3,
Illustration 4
55-2. Using GA4 Demo E-commerce Account: ◉【Data Source】 This is Google's real-time e-commerce data.
1-【Google Merchandise Store】 (Website Data)
◉【Download Demo E-commerce Data URL】 ◉【How to Download the First Demo Data】: Click the first data source link ➜ Automatically download data to your GA4 account ➜ Add 【Demo Account】
◉【Actual E-commerce Website】: Google Merchandise Store
55-3. GA4 Practical Implementation 15: GA4 Exploration Custom Reports ➜ Segment Overlap Exploration Report (7)
◉1- Exploration ➜ Segment Overlap Exploration Report
Example Image 1,
Example Image 2 ◉2- First, create the report framework:
A. Analyze the Overlap of 【Desktop Traffic, Mobile Traffic, Buyer Traffic】
B. Create a New Segment Variable 【Desktop Traffic & Buyer】 ➜ For Ad Targeting
◉3- Observe the Segment Overlap Analysis Nodes 【Segment Comparison】 ➜ A Maximum of Three Can Be Set, Can Be 【Modified or Deleted】
Example Image ◉4- Delete Default Settings and Restart ➜ 【Restart】
Example Image ◉5- Add 【Segment】 Variable ➜ 【Desktop Traffic】
Example Image 1,
Example Image 2 ◉6- Add 【Segment】 Variable ➜ 【Buyer】
Example Image 1,
Example Image 2 ◉7- Observe 【Segment Overlap View】
Example Image 1,
Example Image 2,
Example Image 3,
Example Image 4,
Example Image 5 8- 【Segment Overlap View】 Conclusion
Example Image 9- How to Create a New Segment Variable 【Desktop Traffic & Buyer】 ➜ For Ad Targeting
Example Image 1,
Example Image 2,
Example Image 3
※(1).【Results Demo】: Evaluating the effectiveness of two e-commerce ads, A/B. **Is the average purchase amount for Ad B greater than Ad A?**
62-1 - Assignment 62-2 - Analysis Report ➜☎【Question】: Which **statistical test** should be used for this data?
➜☎【Question】: Is this a **two-tailed test, right-tailed test, or left-tailed test**?
➜☎【Question】: What does it mean to compare the **population means of two independent samples**?
➜
☎※(2).【The Most Important Chart for Determining the Test Method】:
➜ Depending on the (x, y) format, different statistical tests can be selected:
➜ If there is only **one continuous variable (y)**, use the **one-sample t-test** method.
※(4). Use **NTUT Cloud Software (SPSS)** to create a new file. In the **Variable View**, create two fields:
➜ **Ad**: Numeric (Decimal = 0)
➜ **Purchase Amount**: Numeric (Decimal = 0) ➜ Define Labels: (1 = Ad 1, 2 = Ad 2)
➜ Copy and paste the CSV data.
Example 1,
Example 2,
Example 3
※(5). Conduct **Independent Samples t-test**
➜ Go to **Analyze** ➜ **Compare Means** ➜ **Independent Samples t-test**
Independent Samples t-test - Example ➜ Set **Test Variable** ➜ **Purchase Amount**
➜ Set **Grouping Variable** ➜ **Gender**
➜ Set **Options** ➜ **Confidence Interval = 95%**
Set Variables - Example ➜ Click **Define Groups**:
➜ Set **Group 1** ➜ **1**
➜ Set **Group 2** ➜ **2**
Define Groups - Example
※(6). Determine whether this is a **two-tailed test or one-tailed test**?
➜ Hypothesis: **Is the average purchase amount for Ad B greater than Ad A?**
➜ **Null Hypothesis (H₀)**: μ₂ ≤ μ₁ (B ≤ A)
➜ **Alternative Hypothesis (H₁)**: μ₂ > μ₁ (B > A)
➜ **One-tailed test** ➜ **Left-tailed test**
➜ **Left-tailed test**: The **p-value (two-tailed) must be divided by 2**.
➜ **Question**: Why is this a **left-tailed test** instead of a **right-tailed test**?
1. **Conclusion**: Clearly, **Ad B is more effective than Ad A**, as it has a higher average sales amount.
2. **Conclusion: Recommended A/B Plan**
Ad Type
Average Sales Amount
Recommended Priority
Ad B
514.05
1
Ad A
467.77
2
3. 【Conclusion: Box Plot Analysis】: Comparing the Distribution of A/B Advertisements:
Observation: The **median** (central line) of Ad B's sales amount is **higher**.
Observation: The **distribution of Ad B's sales amount is more concentrated**.
Example 1Example 2
※(7). Explanation of **Interquartile Range (IQR) in Box Plot**:
➜ **1. Box Plot**: Also known as a box-and-whisker plot, it is a statistical graph used to show the distribution and characteristics of a dataset.
➜ **2. Box**: The upper boundary of the box represents **Q3 (75th percentile)**, and the lower boundary represents **Q1 (25th percentile)**. The length of the box represents the **middle 50% range** of the data, also known as the **Interquartile Range (IQR)**.
➜ **3. Median**: The middle line inside the box represents the **median** (Q2), the midpoint of the data when arranged in order.
➜ **4. Quartiles**: The box plot is divided into four sections, each representing a quartile. **Q1 (25%)**, **Q2 (50%) = median**, and **Q3 (75%)**.
➜ **5. Outliers**: Data points beyond the whiskers' range are usually considered outliers.
※(8). **When to Use a Box Plot**
➜ **1. Comparing Different Datasets**: Box plots are useful for comparing the distributions of multiple datasets to identify differences quickly. Multiple box plots can be displayed in a single graph for easy comparison.
➜ **2. Checking for Outliers**: Box plots help detect **outliers or extreme values** by highlighting points beyond the whiskers.
➜ **3. Displaying Median & Quartiles**: Box plots provide an intuitive view of the **central tendency and distribution** of data using the median and quartiles.
➜ **4. Observing Data Dispersion**: The height of the box represents the **IQR (Interquartile Range)**, indicating the variability of data. **A taller box means higher variability**.
➜ **5. Identifying Skewness**: By observing the spacing between quartiles, we can determine **whether the data is skewed**. Uneven spacing may indicate skewness.
※(8). **Steps for Independent Samples t-Test**:
➜ **Independent Samples t-Test**: Used to compare a continuous variable between **two independent populations**.
➜ **Step 1**: Set up two hypotheses:
◉ **Null Hypothesis (H₀)**: The two variables are independent (**no significant correlation, no significant difference**).
◉ **Alternative Hypothesis (H₁)**: The two variables are **not independent (significant correlation, significant difference)**.
➜ **Step 2**: Calculate the t-test value using the formula.
➜ **Step 3**: Compute the **p-value**.
➜ **Step 4**: Conduct a **two-step verification**:
◉ **First, check Levene's Test for Equality of Variances**: If **p-value > 0.05**, assume equal variance; otherwise, use **unequal variance**.
◉ **Then, check the t-test for Equality of Means**: Based on Step 1's variance assumption, compare the **p-value** to determine whether **H₀ (null hypothesis) or H₁ (alternative hypothesis) is valid**.
✅ If **p-value < 0.05**, reject the null hypothesis and conclude a significant difference.
※(9). **Advanced Insights: One-Tailed vs. Two-Tailed t-Tests**
(A). **Key Concept 1**: **Right-tailed test**: Tests if **Group 1's mean is significantly larger**.
**Left-tailed test**: Tests if **Group 2's mean is significantly larger**.
Right-tailed test example,
Left-tailed test example,
※(B). **Key Concept 2**: What is **Group 1 variable**?
※(C). **Key Concept 3**: How to determine **H₁ acceptance in a right-tailed test**?
✅ If **t-value > 0** and **p-value (right-tailed) < 0.05**, reject **H₀** and accept **H₁**, meaning "Group 1's mean is significantly greater than Group 2's."
Right-tailed test example,
※(D). **Key Concept 4**: How to determine **H₁ acceptance in a left-tailed test**?
✅ If **t-value < 0** and **p-value (left-tailed) < 0.05**, reject **H₀** and accept **H₁**, meaning "Group 2's mean is significantly greater than Group 1's."
Left-tailed test example,
※(3). **Check how many data entries ChatGPT can process**
(Many generative AI models have a token limit when reading datasets).
First, upload the file (press Enter).
Example Image **Prompt**: How many data entries were read in total?
**ChatGPT Response**: A total of 240 data entries were read.
(If this response is correct, it means ChatGPT successfully processed all 240 entries).
※(4). 【ChatGPT Query Prompt】
A certain e-commerce website aims to increase the **purchase amount for various khaki apparel**.
The goal is to determine which **advertisement (Ad 1 or Ad 2) has a better image and text copy**,
and which ad generates **the highest average sales amount**.
Below is the experimental dataset.
Please use **statistical testing** to analyze the data and determine:
**Which statistical test is appropriate?**
**Does Ad 2 result in a significantly higher purchase amount than Ad 1?**
---
Ad | Purchase Amount
Ad 1 | 606.9
Ad 1 | 543
Ad 1 | 370.2
Ad 1 | 722.4
...
...
※(8). **Correction of ChatGPT Statistical Test Results**:
The above result is a **two-tailed test**.
To perform a **one-tailed test**, **divide the p-value by 2**.
※(3). **Verify how many data entries Grok can process**
(Many generative AI models have a limit on the number of data entries they can read).
**Prompt**: How many data entries were read in total?
**Grok Response**: A total of 190 data entries were read.
This is **incorrect** (the correct number should be **240 entries**).
This indicates that Grok has a limitation and can read **only up to 190 entries**.
※(4). 【Query Prompt for Grok】
A certain e-commerce website aims to increase the **purchase amount for various khaki apparel**.
The goal is to determine which **advertisement (Ad 1 or Ad 2) has a better image and text copy**,
and which ad generates **the highest average sales amount**.
Below is the experimental dataset.
Please use **statistical testing** to analyze the data and determine:
**Does Ad 2 result in a significantly higher purchase amount than Ad 1?**
---
Ad | Purchase Amount
Ad 1 | 606.9
Ad 1 | 543
Ad 1 | 370.2
Ad 1 | 722.4
...
...
※(1). **Results Demo**: Evaluating whether different e-commerce website designs (A/B/C/D) yield significant differences in profitability.
65-1-1 - Assignment 65-1-2 - Analysis Report ➜☎【Question】: What **statistical test** should be used for this dataset?
➜☎【Question】: Is this a **two-tailed test, right-tailed test, or left-tailed test**?
➜☎【Question】: What does it mean to compare **the population means of two independent samples**?
➜
☎※(2). **The Most Important Chart for Determining the Test Method**:
➜ Depending on the (x, y) format, different statistical tests can be applied:
➜ **Comparison 1**: **Independent Samples t-Test** – Compares the mean differences between two groups.
➜ **Comparison 2**: **One-Way ANOVA** – Compares the mean differences among three or more groups.
※(4). **Creating a New SPSS File on NTUT Cloud Software**:
➜ **ID**: Numeric (Decimal = 0)
➜ **Design Type**: Numeric (Decimal = 0) ➜ Define Labels (1: Design 1, 2: Design 2, 3: Design 3, 4: Design 4)
➜ **Purchase Amount**: Numeric (Decimal = 0)
➜ Copy and paste the CSV data.
Example 1,
Example 2,
Example 3
※(5). **Perform One-Way ANOVA**
➜ **Analyze** ➜ **Compare Means** ➜ **One-Way ANOVA**
One-Way ANOVA Example,
➜ Set **Dependent Variable** ➜ **Response Count**
➜ Set **Factor** ➜ **Ad Type** ➜ (Only one factor, hence called "One-Way ANOVA") ➜ Set **Options** ➜ **Statistics** ➜ Check **Descriptive Statistics**
Parameter Settings - Example 1,
Parameter Settings - Example 2,
※(6). Determine whether this is a **two-tailed or one-tailed test**?
➜ **Note 1**: One-Way ANOVA **typically examines differences among multiple groups**, so it usually employs a **two-tailed test** (one-tailed is not considered).
➜ **Note 2**: The **Independent Samples t-Test** also generally uses a **two-tailed test**.
➜ Conclusion: **Two-tailed test**.
※(9). Checking the **Significance Value**
➜ If **p-value < 0.05**: **Reject H₀**, indicating significant differences in profitability among the four website designs.
➜ **Mean values**: **(1236, 1240, 1330, 1256)** – Is the difference significant for future performance?
➜ **Final Result**: p = **0.001** < 0.05, supporting **H₁** (at least two website designs have significantly different sales).
1. **Conclusion 1**: The four website designs have a **significant impact on sales**, with at least two designs showing significant differences.
2. **Conclusion 2**: We know that **at least two designs differ**, but we don’t know **which ones specifically**.
3. **Conclusion 3**: **Recommended A/B/C/D Plan Ranking**
※(3). **First, Verify How Many Data Entries ChatGPT Can Process**
(Many generative AI models have a limit on the number of data entries they can read).
First, upload the file (press Enter).
Example Image **Prompt**: How many data entries were read in total?
**ChatGPT Response**: A total of 1303 data entries were read.
(This response is correct, indicating ChatGPT successfully processed 1300+ entries).
※(4). **Prompt 1 (Initial Analysis)**:
An e-commerce website has four different design layouts.
We want to analyze whether **Design 1 / Design 2 / Design 3 / Design 4**
have a **significant impact** on customer spending.
Please use **statistical testing** to determine:
1. **Which statistical test should be applied?**
2. **Are there significant differences in average sales between designs?**
**Analysis Results**:
※(5). **Prompt 2 (Advanced Analysis)**:
Perform **post-hoc testing (e.g., Tukey HSD)**
to determine which designs have significant differences.
**Analysis Results**:
※(1). **Results Demo**: Evaluating whether different e-commerce website designs (A/B/C/D) yield significant differences in profitability and identifying which design pairs have significant differences.
66-1-1 - Assignment 66-1-2 - Analysis Report ➜☎【Current Problem】: The previous analysis only identified that at least **two of the four designs (A/B/C/D) show significant differences**,
but **it did not specify which ones or the magnitude of their differences**.
➜☎【Question】: How can we **quantify the differences** between A/B/C/D?
➜☎【Question】: Which **statistical test** should be used to measure **significance across multiple parameters**?
☎※(2). **Key Chart for Selecting the Statistical Test**:
➜ Depending on the (x, y) format, different statistical tests should be applied:
To analyze **pairwise differences among A/B/C/D**, follow these two steps:
(1) **Step 1**: Perform **One-Way ANOVA** – To check if there are significant differences among the four groups.
(2) **Step 2**: Conduct **Tukey HSD Post-Hoc Test** – To measure **pairwise significance** and quantify differences.
➜ **What is the Tukey HSD Test?**
It is a **multiple comparison test** used to compare mean differences between multiple groups.
➜ **Tukey HSD is a Post-Hoc Test** – It is applied **after ANOVA** confirms significant differences between groups.
➜ **What is a Post-Hoc Test?**
It is conducted **after an ANOVA test** when **at least one mean difference is significant**,
to identify which specific groups differ.
1. **Conclusion 1**: The best-performing design is **Design 3**, with the highest improvement in sales (+93.27 units).
2. **Conclusion 2**: **Designs 2 and 4** did not show a **consistently significant** improvement (p > 0.05), so they are **not recommended**.
3. **Conclusion 3**: **Tukey HSD Test confirms that Design 3 is the most effective** in increasing revenue.
4. **Conclusion 4**: **Designs 2 and 4 show inconsistent effects**, meaning they may not guarantee increased sales.
5. **Conclusion 5**: **Recommended A/B/C/D Plan**
※(1). **Results Demo**: Evaluating whether different e-commerce website designs (A/B/C/D) yield significant differences in profitability and identifying which design pairs have significant differences.
※(1). **Results Demo**: Evaluating whether different e-commerce website designs (A/B/C/D) yield significant differences in profitability and identifying which design pairs have significant differences.
※(3). **First, Verify How Many Data Entries ChatGPT Can Process**
(Many generative AI models have a limit on the number of data entries they can read).
First, upload the file (press Enter).
Example Image **Prompt**: How many data entries were read in total?
**ChatGPT Response**: A total of 1303 data entries were read.
(This response is correct, indicating ChatGPT successfully processed 1300+ entries).
※(4). **Prompt 1 (Initial Analysis)**:
An e-commerce website has four different design layouts.
We want to analyze whether **Design 1 / Design 2 / Design 3 / Design 4**
have a **significant impact** on customer spending.
Please use **statistical testing** to determine:
1. **Which statistical test should be applied?**
2. **Are there significant differences in average sales between designs?**
**Analysis Results**:
※(5). **Prompt 2 (Advanced Analysis)**:
Perform **post-hoc testing (e.g., Tukey HSD)**
to determine which designs have significant differences.
**Analysis Results**:
☎#colab顯示繁體中文，方法2
☎程式碼：
#--------------------------------------
# 課本的中文處理
#--------------------------------------
import matplotlib as mpl
import matplotlib.font_manager as fm
☎如何安裝google chrome的網頁線上英文字典工具：
➜google chrome的右上角工具➜更多工具➜擴充功能
➜左上角主選單➜開啟chrome線上應用程式商店
➜勾選：google製作，免費
➜搜尋：google dictionary➜安裝
➜到chrome右上擴充功能➜點按google dictionary的『詳細資料』➜擴充功能選項
➜my language=chinese
➜打勾2個：Pop-up definitions:
（1）反白單字翻譯：Display pop-up when I double-click a word
（2）ctrl+拖曵整段翻譯： Display pop-up when I select a word or phrase
Math500 Competition (Evaluating Mathematical Reasoning Ability): DeepSeek outperformed ChatGPT-4o, Claude, and LLaMA.
