- AI Fire
- Posts
- 🎥 FREE Claude Prompt That AUTO Turns Any Idea Into a VOX-Style Video (Full Guide)
🎥 FREE Claude Prompt That AUTO Turns Any Idea Into a VOX-Style Video (Full Guide)
This workflow turns one idea into a Vox-style animated video using 4 tools: Claude, ZAPI Flow, Google Flow, and CapCut. No coding required. No After Effects.

TL;DR
This workflow helps you turn one idea into a Vox-style animated video without coding or After Effects. You’ll use Claude, ZAPI Flow, Google Flow, and CapCut.
Claude helps you write the script, split it into visual beats, and create matching prompts. ZAPI Flow sends those prompts to Google Flow in order, so you can generate the full image set without copying them one by one.
Google Flow turns each image into a short clip. You then arrange the clips in CapCut, trim the timing, and export the video.
Key points
Fact: Each image becomes one clip of about 5 to 6 seconds.
Mistake: Running the full batch before testing one image can repeat the same problem across every scene.
Takeaway: Check one image first and keep the characters, colors, and paper-cut style consistent.
Table of Contents
Introduction
Look at this. Do you think it was made in After Effects? Or that someone spent hours animating every frame by hand? Haha... nope.
I have to say, I was surprised too. This entire video started with a few still images, that’s it.

In this guide, I'll show you exactly how I made this Vox-style animation using:
Claude → Write the script, break it into visual beats, and generate all the prompts.
ZAPI Flow → Run every prompt automatically in the correct order.
Google Flow → Generate the images and turn each one into an animated video clip.
CapCut → Put everything together and polish the final video.
I'll also share the tricks I used to keep the characters, colors, camera angles, and paper-cut style consistent across every scene, because that's the part most people struggle with.
🎬 What stops you from making Vox-style videos? |
Step 1: Write the Script in Claude
Before you create images with an AI animation generator, you need a clear script. And not just any script, it must be one where every sentence describes something you can actually show on screen.
What you're building in this step:
A 60-second documentary-style script (around 130–150 words)
Each sentence = one clear visual moment
1a. Set Up Claude
Open a new Claude chat and paste this prompt. The key line is "one step at a time", don't jump to image prompts before the script is done.
You’re helping me create a Vox-style animated documentary video.
Please guide me through the process one step at a time.
For now, only help me create the video idea and script. Don’t create image prompts or video prompts yet.
The video should have:
- A clear documentary story
- Short scenes that are easy to turn into images
- Simple and natural language
- A strong opening
- Smooth connections between events
- A clear ending
First, ask me for my niche, video length, and topic.
After I answer, give me 10 video ideas that can work well as Vox-style animation.
Claude will ask you for some basic information. You can answer with a message like this:
My niche is history and unusual true stories.
I want to make a 60-second video.
I don’t have a topic yet. Please give me 10 ideas with a clear main character, problem, and ending.
You can change the niche and video length to match your channel.
1b. Choose an Idea and Build a Story Plan
Once Claude gives you 10 ideas, pick the one with the clearest people, places, and events. Stories with a single main character and a clear turning point are easiest to animate.
Then ask Claude to plan it out before writing the full script:
I want to use this idea:
[PASTE YOUR IDEA HERE]
Before you write the full script, create a short story plan.
Please include:
- The opening hook
- The main character or subject
- The important events in order
- The main turning point
- The ending
Keep the story simple enough for a 60-second video.
Read the plan back. If it feels too crowded or skips a key moment, ask Claude to adjust it now, it's much faster to fix at this stage than after the full script is written.
1c. Write the Full Script
When the story plan looks good, paste this:
Now write the full 60-second script based on this story plan.
Please follow these rules:
- Start with a strong hook in the first two sentences
- Use a natural documentary voice
- Make every sentence easy to show as an image
- Connect the events clearly
- Remove unnecessary background details
- Don’t include camera instructions
- Don’t include scene numbers
- Don’t include image prompts
- Keep the script between 130 and 150 words
After the script, show me the exact word count.
Read each sentence and ask yourself: can I picture a single, clear image for this? If a sentence describes two actions or jumps between places, it'll confuse the image model later.
If you spot any that are too crowded, paste this:
Please review this script for visual clarity.
Find any sentence that includes too many actions, places, or ideas for one scene.
Rewrite only those sentences.
Each new sentence should describe one clear visual moment.
Keep the original meaning, tone, and length as close as possible.
At the end of this step, you should have a clean 130–150 word script where every sentence is something you can picture.
Step 2: Turn the Script Into Visual Beats and Image Prompts
Now you take that script and break it apart into individual scenes. Each "visual beat" is one moment, one image, in your story. This is where your animation actually takes shape.
What you're building in this step:
A numbered list of visual beats (one per scene)
One image prompt per beat
A clean TXT file ready for ZAPI Flow
2a. Create the Visual Beats
Paste this into Claude with your finished script:
I want you to turn my script into visual beats for a Vox-style animated documentary video.
Please read the full script and divide it into short visual scenes.
Follow these rules:
- Each visual beat should show one clear visual moment
- Keep the beats in the same order as the script
- Use simple and easy-to-follow language
- Don’t skip important events
- Don’t add new facts that are not in the script
- Make the beats suitable for image generation later
- Keep the pacing right for a 60-second video
Please return:
- Numbered visual beats
- One sentence per beat
- A short note for each beat that explains what should be shown on screen
Here is the script:
[PASTE YOUR FINAL SCRIPT HERE]
After Claude gives you the list, quickly scan it. If any beat still has two actions or feels too busy for one image, use this:
Please review these visual beats.
If any visual beat includes too many actions, details, or visual ideas for one image, split it into smaller beats.
Please keep the story order unchanged.
Return the improved numbered visual beats only.
2b. Write One Image Prompt Per Beat
This is the most important prompt in the whole workflow. The quality of your images depends directly on how well these prompts describe the paper-cut style and character consistency.
Now create one image prompt for each visual beat.
I want these prompts for a Vox-style animated documentary video.
Please follow these rules for every prompt:
- Keep the same story order
- Keep the same character identity across scenes
- Use a clean Vox-style paper-cut illustration look
- Use simple shapes, flat layers, and soft colors
- Make the composition clear and easy to understand
- Describe the main subject, background, action, and mood
- Keep the prompt suitable for 16:9 image generation
- Avoid extra details that may confuse the image model
- Don’t include camera movement because these are image prompts
- Don’t include quotation marks around the prompt text
For each visual beat, return:
- Beat number
- A short beat title
- The full image prompt
Please keep the style consistent across all prompts.
If you feel the style descriptions could be stronger, send this quick follow-up:
Please make sure every prompt keeps a consistent Vox-style visual language with paper-cut shapes, documentary-style composition, and simple layered backgrounds.
2c. Export a Clean TXT File for ZAPI Flow
ZAPI Flow only needs the raw prompts: one per line, no labels or numbering. Ask Claude to strip everything else:
Please turn all of the image prompts into a clean TXT-style output for batch image generation.
Follow these rules:
- Keep the prompts in the same order
- Put each full image prompt on a new line
- Don’t include beat titles
- Don’t include numbering
- Don’t include notes or explanations
- Don’t add any extra text before or after the prompts
Please return only the final TXT content.
Copy the full output and save it as a .txt file on your computer. That file is what goes into ZAPI Flow in the next step.
Step 3: Generate Images with ZAPI Flow + Google Flow
So now, ZAPI Flow (install it first) reads your TXT file and sends each prompt to Google Flow one by one, so you don't have to copy-paste 10 or 15 prompts by hand.
But first, read these two things before you start:
⚠️ ZAPI Flow bug warning: As of July 2026, multiple users report a queue stall bug where the extension stops after the first image and doesn't move to the second prompt. A separate bug affects large batches: past around image 40, the extension may download old images with new filenames. Keep the progress panel visible throughout, and check file names as you go. If it stalls, refresh Google Flow and reopen the extension.
💡 On Nano Banana Pro vs Nano Banana 2: Nano Banana 2 is now the free default model in Flow (zero credits per image), while Nano Banana Pro requires a paid Google AI subscription. For most people doing this workflow, Nano Banana 2 is the practical starting point. Switch to Pro only if you need 4K output or have a paid plan.
3a. Set Up Google Flow
Go to labs.google/fx/tools/flow, create a new project, and switch to image generation mode.
Settings to use:
Aspect ratio: 16:9 (matches video format)
Model: Nano Banana 2 (free default) or Nano Banana Pro (paid)
Outputs per prompt: 1You should open Google Flow, select Create a New Project, and switch from video generation to image generation.


Before you run the full batch, test one prompt manually. Copy your first prompt from the TXT file, paste it directly into Google Flow, and generate. Check that the character, colors, composition, and paper-cut style look right. If they don't, fix the style description in the prompt and test again.

3b. Set Up ZAPI Flow
Install ZAPI Flow from the Chrome Web Store, then open it inside your Google Flow project. Before starting:
Check the connection status shows your project is detected
Upload the TXT file Claude created
Confirm the prompt count matches your number of visual beats
Toggle off "Include serial number in file name" if you don't want numbered filenames


Each prompt should be on its own line, ZAPI Flow reads top to bottom, so the image order follows your story order automatically.

If ZAPI Flow can't detect the project, refresh Google Flow and reopen the extension. Don't hit Start Generation until the connection and prompt count both look correct.
3c. Run the Batch and Review
Click Start Generation. ZAPI Flow will send each prompt to Google Flow and wait for the image to finish before moving on. Watch the progress panel, don't assume it's running correctly just because it started.


Once the batch finishes, open your download folder and check the images in order. Look for:
Character looks the same (face, clothes, key features) across all scenes
Colors and paper-cut style stay consistent
Each image clearly matches its visual beat
No wrong text, extra objects, or broken details
If specific images came out wrong, regenerate just those ones. You don't need to redo the full batch.
How useful was this AI tool article for you? 💻Let us know how this article on AI tools helped with your work or learning. Your feedback helps us improve! |
Step 4: Animate Each Image Into a Clip
Now the still images become short video clips. This is where the paper-cut illustrations start to feel alive.
Each image becomes one clip of about 5 to 8 seconds. You process them one at a time: upload image, add animation prompt, generate, download, repeat.
4a. Create a Universal Animation Prompt in Claude
For this step, ask Claude to write one base animation prompt you'll reuse across every scene. This keeps the movement style consistent, the same way the image prompts kept the visual style consistent.
I want you to create a universal animation prompt for a Vox-style animated documentary video.
The prompt will be used with different still images in Google Flow.
Please follow these rules:
- Keep the movement natural and simple
- Use slow camera movement
- Add small motion to the main subject and background
- Keep the original composition and visual style
- Don’t add new people, objects, or text
- Don’t change the character’s face, clothes, or key features
- Avoid fast movement, strong camera shakes, and dramatic effects
- Make the animation suitable for a 5 to 6-second clip
Return one clean animation prompt only.
For each scene, you'll paste this base prompt plus one short sentence describing the specific action in that beat, for example: "the character slowly hands a document across a table" or "the character turns their head toward the window."
4b. Generate Each Clip in Google Flow
Switch Google Flow to image-to-video mode (Veo 3.1). Then for each scene:
Upload one image from your set
Paste the universal animation prompt from Claude
Add one sentence describing the movement for that specific scene
Keep the aspect ratio at 16:9
Generate a 5–8 second clip
Download the clip, then move to the next image

Process the images in the same order as your visual beats. This keeps everything in story order for CapCut.
After each clip generates, check it quickly before downloading:
Character keeps the same face and clothes
No new people, objects, or text appeared
The movement matches the visual beat
Camera doesn't move too fast or break the composition
Paper-cut style is still intact

If a clip has a problem, reduce the movement in the action sentence or regenerate. Once all clips look good, download them and move to editing.
Step 5: Assemble Everything in CapCut
This is the final step: putting the clips together into a finished video.
Open CapCut, start a new project, and import all your clips. Place them on the timeline in the same order as your visual beats.
Then go through these in order:
1. Trim and pace → each clip is 5–8 seconds, but you may not need all of it. Cut to the moment you want.
2. Review transitions → light cuts or simple fades work better than dramatic transitions for a documentary style. Add them only where they help the story flow, not everywhere.
3. Add text or captions → if a key fact or name needs to appear on screen, this is where you add it.
4. Check the full run-through → watch the whole video once before exporting. Look for any pacing gaps, jarring cuts, or moments where the story loses momentum.
5. Export → 1080p, MP4. That's the standard for YouTube and most platforms.


Conclusion
By this point, you’ve used an AI animation generator workflow to turn one idea into a complete Vox-style animated video. Each part of the process has a clear job:
Claude prepares the script, visual beats, and prompts.
ZAPI Flow sends the image prompts to Google Flow in order.
Google Flow creates the images and adds motion to each scene.
CapCut connects the clips and gives you the final video.
You can reuse this entire workflow for history, business, education, or any story you want to animate. Change the topic and visual direction, keep the process the same.
If you are interested in other topics and how AI is transforming different aspects of our lives or even in making money using AI with more detailed, step-by-step guidance, you can find our other articles here:
Opus 5 Honest Review: Near-Fable 5 Intelligence at Half the Price (July 2026)
How to Use Gemini + NotebookLM to Keep Your AI Accurate Every Time (For Free)
How to Build a Viral Faceless YT Channel with Claude from Zero (Full Ranking-Video System)*
Prompting Has Changed. Anthropic’s New Rules with 7 Old Habits You Must Stop!*
This #1 Trending Skill Completely Gives Claude a Research Superpower (55K-Star Now)*
*indicates a premium content, if any

Reply