How to Make AI Cartoon Videos With Google Flow: Step-by-Step Guide

You can now create a complete cartoon video using AI, from the first character idea to the finished YouTube video.

The easiest way to think about the process is:

Find a visual style → Create a style guide → Write the story → Design the character → Break the story into scenes → Write scene prompts → Create the scenes in Google Flow → Animate them → Add voiceover → Edit → Add sound → Review → Publish

Here is the full process.

Step 1: Find a Cartoon Style You Like

Start by going to YouTube and searching for cartoon or animated videos.

Find a visual style that matches the kind of video you want to create.

Look at things such as:

• Character design
• Facial expressions
• Body proportions
• Background style
• Lighting
• Camera angles
• Colors
• Clothing
• Line work
• Amount of detail
• Mood
• How realistic or exaggerated the characters look

Take one or more screenshots.

Try to choose screenshots that clearly show the characters and environment.

Step 2: Ask ChatGPT to Analyze the Cartoon Style

Upload the screenshot to ChatGPT.

Ask something like:

“Break down this cartoon animation style in extreme detail. Describe the character proportions, faces, eyes, hair, clothing, colors, lighting, line work, background design, camera angles, textures, expressions, and overall visual appearance.”

ChatGPT will create a detailed description of the visual style.

Read through it carefully.

Remove anything that doesn't match what you want and add anything that seems missing.

Then ask:

“Turn this into a reusable AI image Style Guide that I can use for every scene in one cartoon series.”

Save this Style Guide.

You will reuse it throughout the entire project.

Helpful Tip

Rather than asking AI to duplicate one particular animator or studio, describe the visual qualities you like.

That gives you a style that feels familiar without simply reproducing someone else's work.

Step 3: Decide What Your Cartoon Is About

Now choose your basic concept.

It can be almost anything.

For example:

• A day in the life of an Amazon driver in New York
• A funny family story
• A historical story
• A motivational story
• A mystery
• A children's story
• A relationship story
• A workplace comedy
• A travel adventure
• A fictional version of a real-life situation

Before writing the script, decide:

• Who the main character is
• What they want
• What problem they face
• Where the story takes place
• What happens in the end

A good cartoon normally becomes more interesting when the main character has a clear problem.

For example:

Character: Delivery driver

Goal: Finish all deliveries before the shift ends

Problem: A snowstorm hits Manhattan

Complication: His phone battery is almost dead

Emotional element: One final package needs to reach an elderly customer before nightfall

That immediately gives the story direction.

Step 4: Ask ChatGPT to Write the Script

Now ask ChatGPT to create the full script.

For example:

“Write a five-minute YouTube cartoon script about a delivery driver working in New York City during a chaotic day. Make it funny, emotional, and visually interesting.”

You can tell ChatGPT:

• Video length
• Topic
• Audience
• Tone
• Main character
• Setting
• Type of humor
• Amount of dialogue
• Narration style
• Ending
• Emotional message

For a short YouTube cartoon, you may ask for something like three to eight minutes.

For a longer story, you can ask for ten minutes or more.

Read the finished script carefully.

Make changes before you start generating images.

Changing a sentence in a script takes seconds.

Changing 30 finished cartoon scenes takes much longer.

Step 5: Create Your Main Character

Before generating all the scenes, create the character who will appear throughout the video.

Ask ChatGPT:

“Create a detailed character design for the main character in this story using our cartoon Style Guide.”

The character description should include:

• Approximate age
• Gender
• Height
• Body shape
• Face shape
• Eyes
• Eyebrows
• Nose
• Mouth
• Hair
• Skin tone
• Clothing
• Shoes
• Accessories
• Personality
• Typical expressions
• Distinctive visual features

For example:

“Daniel is a 32-year-old delivery driver with short black hair, slightly tired eyes, a narrow face, medium build, blue delivery jacket, dark pants, gray sneakers, and a small black backpack.”

Try to keep the description simple enough that AI can reproduce it repeatedly.

Too many tiny details can make character consistency harder.

Step 6: Create Your Master Character Image

Now go to Google Flow.

Create a clean image of your main character using the character description and Style Guide.

A full-body image is often useful.

Try a few versions.

Do not move forward until you have one character you really like.

Save the best image.

This becomes your:

Master Character Reference

You will reuse it when creating the rest of the cartoon.

This step can make a huge difference.

Without a reference image, AI may slowly change:

• Hair
• Face
• Age
• Clothing
• Body shape
• Skin tone
• Accessories

Your character could look like a different person several scenes later.

Step 7: Test the Character Before Making the Whole Cartoon

Before creating 30 or 40 scenes, test your character.

Create three sample images in Google Flow.

Test 1: Close-Up

Show the character's face clearly.

Test 2: Full Body

Show the character standing or walking.

Test 3: Action

Show the character performing an action, such as carrying a package, opening a door, running, cooking, or talking to someone.

Use your Master Character Reference each time.

Look at all three images.

Ask yourself:

• Does the face stay similar?
• Is the hairstyle the same?
• Is the clothing correct?
• Does the character seem like the same age?
• Are the proportions similar?

If the character changes too much, fix the prompts now.

Do not wait until you have generated the entire story.

Step 8: Break the Script Into Visual Scenes

Now return to ChatGPT.

Paste the finished script and ask:

“Break this script into individual visual scenes. Each scene should contain one clear visual idea or action. Include the narration or dialogue associated with each scene.”

Some tutorials recommend changing the visual every ten words.

That can work for very fast videos, but it is usually better to change scenes when something meaningful changes.

Create a new scene when there is a change in:

• Location
• Action
• Emotion
• Character
• Camera angle
• Time
• Major story beat

For example:

Scene 1: Delivery van driving through Manhattan.

Scene 2: Driver checks his phone.

Scene 3: Phone shows 42 deliveries remaining.

Scene 4: Snow starts falling.

Scene 5: Traffic stops.

Scene 6: Driver looks frustrated.

Each scene now has one clear visual purpose.

Step 9: Create a Scene List

Ask ChatGPT to organize everything into a production list.

For each scene, include:

Scene Number

Narration or Dialogue

Visual Description

Characters

Location

Emotion

Camera View

This becomes your production roadmap.

Example:

Scene 7

Narration:

“Daniel looked at the clock. He still had twenty-three packages left.”

Visual:

Daniel sits inside his delivery van looking worried at the dashboard clock.

Characters:

Daniel

Location:

Delivery van in Manhattan

Emotion:

Stress and urgency

Camera:

Medium shot from passenger side

Step 10: Ask ChatGPT to Write an Image Prompt for Every Scene

Now ask:

“Create a detailed AI image prompt for every scene. Use our established Style Guide and Character Sheet. Keep the main character's appearance consistent in every prompt.”

Each scene prompt should cover:

• Main character
• Clothing
• Expression
• Body position
• Action
• Environment
• Objects
• Weather
• Time of day
• Camera angle
• Camera distance
• Lighting
• Composition
• Cartoon style
• Mood

You want each prompt to describe one clear visual moment.

Avoid trying to place too many actions inside one image.

AI normally performs better with:

“Daniel stands outside an apartment building holding a package.”

than:

“Daniel runs down the street, talks on the phone, opens the package, waves at someone, gets into his van, and drives away.”

Break complicated moments into several scenes.

Step 11: Create Each Scene in Google Flow

Now the main production begins.

Open Google Flow.

For each scene:

1. Select the image creation workflow available to you.
2. Upload or reference your Master Character Image when possible.
3. Copy the scene prompt from ChatGPT.
4. Paste it into Flow.
5. Generate the scene.
6. Review the result carefully.
7. Generate another version when needed.
8. Save the version that best matches the story.
9. Move to the next scene.

Continue:

Scene 01

Scene 02

Scene 03

Scene 04

and so on.

Keep your original character reference available throughout the process.

Step 12: Keep Your Characters Consistent

This is one of the hardest parts of AI cartoon production.

Check every scene for continuity.

Pay attention to:

• Face
• Hair
• Clothing
• Shoes
• Body shape
• Height
• Age
• Accessories

If your character wears a red jacket in Scene 1, don't let it become green in Scene 14 unless the story calls for a clothing change.

If several characters appear in the story, create a separate Master Character Reference for each one.

For example:

Daniel — Main Character Reference

Maria — Character Reference

Mr. Lee — Character Reference

Delivery Manager — Character Reference

Keep them organized.

Step 13: Keep Locations Consistent Too

Characters are not the only thing that can change.

AI can redesign the same apartment, office, car, bedroom, classroom, or restaurant every time it appears.

Create reference images for important locations.

For example:

Daniel's Van

Daniel's Apartment

Delivery Warehouse

Customer's Apartment Building

If a location appears repeatedly, save one strong image and use it as a reference whenever possible.

This gives your cartoon a more believable visual continuity.

Step 14: Generate Different Camera Angles

Do not make every scene a medium shot of someone standing in the center.

Mix the camera views.

Use:

• Establishing shot
• Wide shot
• Full-body shot
• Medium shot
• Close-up
• Extreme close-up
• Over-the-shoulder shot
• Side angle
• High angle
• Low angle
• Point-of-view shot

For emotional scenes, close-ups can work well.

For locations, use wider shots.

For movement or comedy, experiment with more dramatic angles.

Changing the camera view makes the cartoon feel more like a real production.

Step 15: Turn Your Images Into Video Clips With Google Flow

Once your scene images are finished, use Google Flow's video-generation features when available to turn key images into moving clips.

Start with simple movement.

For example:

“Daniel walks slowly down the snowy sidewalk carrying a package. His jacket moves slightly in the wind. Cars move naturally in the background. The camera gently follows him.”

Another example:

“Daniel looks at his phone, his expression changes from calm to worried, and he slowly looks up at the traffic ahead.”

Short, controlled actions tend to be easier to manage than complicated movement.

Step 16: Keep Animation Prompts Simple

One common mistake is asking the AI to do too much in one clip.

Instead of:

“The character runs across the road, drops a package, picks it up, talks to a woman, jumps into the van, starts driving, and looks back.”

Break it into several clips.

Clip A

Daniel runs across the street.

Clip B

The package falls.

Clip C

Daniel picks it up.

Clip D

A woman approaches him.

Clip E

Daniel returns to the van.

This usually gives you much cleaner animation.

Step 17: You Do Not Need to Animate Every Image

Animating every scene can take much more work.

Some scenes can remain still.

You can create movement later in your video editor using:

• Slow zoom
• Slow pan
• Push-in
• Pull-back
• Left-to-right movement
• Cropping
• Slight rotation

Save full AI animation for scenes where movement adds something meaningful.

This can reduce your production workload dramatically.

Step 18: Create the Voiceover

Once the visuals are prepared, create the narration.

You can:

• Record your own voice
• Use an AI voice
• Use several AI voices for different characters

Choose a voice that matches the story.

For comedy, you may want something energetic.

For emotional storytelling, a calmer voice might work better.

For children's cartoons, clearer pacing may help.

Generate or record the narration in sections.

Naming the files by scene can make editing much easier.

For example:

01_intro.wav

02_delivery.wav

03_snowstorm.wav

04_customer.wav

Step 19: Create Voices for Individual Characters

If the cartoon contains dialogue, give the characters different voices.

Try to keep the same voice for the same character throughout the series.

You can create a simple Voice Guide.

Example:

Daniel: relaxed male voice, early 30s, slightly sarcastic

Maria: warm female voice, energetic

Manager: older male voice, fast and impatient

Customer: elderly female voice, soft and friendly

This prevents characters from sounding completely different from scene to scene.

Step 20: Put Everything Into a Video Editor

Now bring your materials into your preferred video editor.

You can use software such as:

• CapCut
• DaVinci Resolve
• Adobe Premiere Pro
• Final Cut Pro
• Another video editor you already know

Import:

• Flow video clips
• Still images
• Narration
• Dialogue
• Music
• Sound effects

Place the narration on the timeline first.

Then place the scenes above it.

Match each scene to the words being spoken.

Step 21: Adjust the Timing

Do not make every shot exactly the same length.

A simple visual may need only two or three seconds.

An emotional moment might need five or six seconds.

A joke might need a slight pause after the punchline.

Watch the cartoon and ask:

“Does the image change when my attention naturally wants something new?”

If a scene feels boring, shorten it.

If a scene feels rushed, extend it.

Step 22: Add Sound Effects

Sound effects can make simple animation feel far more believable.

For a New York delivery story, you might add:

• Traffic
• Car horns
• Footsteps
• Elevator sounds
• Apartment buzzers
• Van doors
• Packages hitting the ground
• Phone notifications
• Wind
• Rain
• Snow
• Crowds
• Restaurant sounds
• Subway sounds

Sound gives the viewer information that the image alone cannot provide.

Step 23: Add Background Music

Choose music that supports the emotional tone.

You might use:

• Light music for comedy
• Faster music during action
• Quiet music during emotional moments
• Suspenseful music when something goes wrong
• Warmer music near the ending

Keep the music lower than the narration.

The audience should never struggle to hear what is being said.

Step 24: Add Captions

Captions can help viewers follow the story, especially on phones.

Many video editors can automatically generate subtitles from your voiceover.

Review them carefully.

AI captions sometimes misunderstand:

• Names
• Places
• Slang
• Unusual words

Correct those before publishing.

Step 25: Add an Opening Hook

Your first few seconds matter.

Do not spend too long introducing everything.

Start with something that creates curiosity.

For example:

“Daniel thought this would be his easiest delivery day of the year.”

Then show him staring at a massive snowstorm.

Or:

“He had one package left. Unfortunately, it was going to the most impossible address in New York.”

The viewer immediately wants to know what happens.

Step 26: Create a Strong Ending

Do not let the story simply stop after the final action.

Give the viewer some emotional payoff.

The ending might contain:

• A joke
• A surprise
• A lesson
• A reveal
• A touching moment
• A callback to the opening
• A question

A strong ending makes the entire cartoon feel more complete.

Step 27: Watch the Entire Video From Beginning to End

Before publishing, watch the video like a viewer.

Do not stop every few seconds to edit.

Watch the whole thing once.

Look for:

• Character changes
• Clothing changes
• Wrong locations
• Strange hands
• Extra fingers
• Distorted faces
• Characters disappearing
• Background mistakes
• Unnatural animation
• Awkward pauses
• Voice mistakes
• Music that is too loud
• Missing sound
• Caption errors
• Scenes that last too long
• Scenes that move too quickly

Take notes.

Then fix them.

Step 28: Export a Test Version

Export the entire video once before creating the final version.

Watch it on:

• Computer
• Phone
• Television, if possible

You may notice things on a phone that you didn't notice on a large editing screen.

Pay particular attention to:

• Caption size
• Audio volume
• Scene brightness
• Character visibility
• Pacing

Then make your final corrections.

Step 29: Create the Thumbnail

Once the cartoon is finished, create a YouTube thumbnail.

Use one of your strongest characters or scenes.

A good thumbnail normally has:

• One clear subject
• Strong facial expression
• Easy-to-read composition
• Very little text
• A question or visual mystery

Do not fill it with every character and every event from the story.

Simple thumbnails are often easier to understand.

Step 30: Save Everything as a Reusable Cartoon System

Once you finish your first video, don't throw away the production materials.

Save:

• Style Guide
• Character Sheet
• Master Character Images
• Location References
• Script format
• Scene breakdown format
• Prompt template
• Voice Guide
• Music choices
• Sound library
• Thumbnail style

Now your second cartoon becomes much easier to make.

Instead of starting from zero, you already have your own AI cartoon production system.

A Useful Prompt for Creating Every Scene

You can give ChatGPT something like this:

“Using the Style Guide, Character Sheet, Master Character Reference, and finished script we created, break the script into visual scenes. For every scene, provide the narration, character action, facial expression, background, camera angle, lighting, composition, and a detailed Google Flow prompt. Keep all recurring characters, clothing, objects, and locations visually consistent throughout the story.”

This can save a great deal of manual work.

A Useful Google Flow Prompt Structure

For each scene, build your prompt in roughly this order:

Character + Action + Expression + Location + Important Objects + Camera View + Lighting + Mood + Style + Continuity Instructions

For example:

“Daniel, the same 32-year-old delivery driver from the reference image, wearing the same blue delivery jacket, dark pants, gray sneakers, and black backpack, stands beside his delivery van holding a cardboard package. He looks worried at heavy snow covering a Manhattan street. Yellow taxis and apartment buildings appear in the background. Medium-wide cinematic shot from slightly below eye level. Soft winter afternoon light, gentle snowfall, expressive cartoon features, consistent character proportions and facial design, same animation style and clothing as the reference image.”

That structure gives the AI much clearer instructions.

The Complete AI Cartoon Workflow

Here is the entire process in one place:

1. Find a cartoon style you like.

2. Screenshot the style.

3. Upload it to ChatGPT.

4. Ask ChatGPT to create a detailed Style Guide.

5. Choose your story idea.

6. Define the main problem or conflict.

7. Ask ChatGPT to write the script.

8. Edit the script.

9. Create a Character Sheet.

10. Create your main character in Google Flow.

11. Save a Master Character Reference.

12. Test the character in three different scenes.

13. Break the script into visual scenes.

14. Create a production scene list.

15. Ask ChatGPT for a prompt for every scene.

16. Generate every scene in Google Flow.

17. Keep characters consistent using reference images.

18. Create location references for recurring locations.

19. Use different camera angles.

20. Turn important images into animated clips using Google Flow.

21. Keep animation actions simple.

22. Use still images for scenes that don't require full animation.

23. Create the narration.

24. Create character voices when needed.

25. Import everything into a video editor.

26. Match each scene to the narration.

27. Adjust pacing.

28. Add sound effects.

29. Add background music.

30. Add captions.

31. Strengthen the opening hook.

32. Create a satisfying ending.

33. Watch the complete video.

34. Fix continuity and AI mistakes.

35. Export a test version.

36. Review it on different screens.

37. Export the final video.

38. Create the YouTube thumbnail.

39. Save your characters, Style Guide, prompts, and references.

40. Reuse the same system for your next cartoon.

The Most Important Tip

Do not generate your entire cartoon immediately.

Create:

One character + three test scenes + one short animated clip

first.

If the character remains consistent and the animation looks good, continue with the full story.

If something is wrong, fix your Style Guide, Character Sheet, reference image, or Flow prompts before producing dozens of scenes.

That small test can save hours of unnecessary work.

Once you have a character, visual style, Google Flow workflow, scene-prompt format, and editing system that work together, you have something much more valuable than one AI cartoon.

You have the foundation for producing an entire recurring cartoon series.

Enter your text here...