• Home
  • /
  • Blog
  • /
  • How to Create Consistent AI Characters for Video with Google Whisk & Flow (2026)

How to Create Consistent AI Characters for Video with Google Whisk & Flow (2026)

Honestly, I was about to give up on AI video. Every time I tried to make a short movie, my main character would look completely different in every scene. One minute he’s wearing a hoodie, the next he’s in a suit. It was a mess.

But I spent the last 3 days testing a new workflow, and I finally cracked the code. If you are looking for how to create consistent AI characters, you are in the right place. In this beginner guide to Google Whisk and Flow, I’m going to show you exactly how to fix inconsistent AI video characters once and for all.

Best of all? The main tools are free. Let’s dive in.



Step 1: Create Your Character with Google Whisk

The first step isn’t making a video at all. It’s making a picture. We need a “master reference” of our character. For this, we are using a new tool from Google Labs called Whisk.

  1. Go to Google Whisk (in Google Labs).
  2. Type a simple prompt for your character. For example: “A cute robot mascot with white fur.”
  3. Important: In the settings, turn OFF “Precise Reference” for now. This lets the AI be creative.
  4. Hit Generate and pick your favorite image.
  5. Download that image. This is now your “Actor.”

Note: If you like the character but want to change one small thing (like fur color), click Refine, turn ON “Precise Reference,” and type the change.

The Whisk interface showing the prompt box and settings panel
The Whisk interface showing the prompt box and settings panel

Step 2: Build Your “Starting Frame”

Now we need to put your actor into a scene. If you just ask an AI video generator to “show the robot in an office,” it will invent a new robot. We don’t want that. We want our robot.

  1. Stay in Whisk.
  2. Look at the sidebar. You will see boxes for “Subject,” “Scene,” and “Style.”
  3. Drag and Drop your character image from Step 1 into the Subject box.
  4. Type a prompt for the scene, like: “Talking to a female worker in an office.”
  5. Enable “Precise Reference” this time. We want the AI to stick strictly to our character’s look.
  6. Hit Generate.
  7. Download the best result.

This image is your “Starting Frame.” It fixes the look of the character before the video even starts.

Dragging the character image into the "Subject" box in the sidebar
Dragging the character image into the “Subject” box in the sidebar

Step 3: Animate with Google Flow

Now we turn that static picture into a movie. We are using Google Flow (also part of Google Labs, sometimes called VideoFX/Veo).

  1. Open Google Flow.
  2. Select “Frames to Video” (this is crucial!).
  3. Upload your “Starting Frame” from Step 2.
  4. Write a prompt describing exactly what happens. Tip: Be specific! “The robot waves his hand and says hello.”
  5. Click Generate.

Because we gave it a starting picture, the AI doesn’t have to guess what the character looks like. It just animates the picture we gave it!

The "Frames to Video" upload area in Google Flow
The “Frames to Video” upload area in Google Flow

Step 4: Fix the Voice with ElevenLabs

You might notice a problem. If you generate two different clips, your character might sound like a monster in one and a mouse in the other. To fix inconsistent AI video audio, we need an external tool.

  1. Go to ElevenLabs and find the Voice Changer tool.
  2. Upload your video clip from Flow.
  3. Pick a specific voice (I like the “Monster” voice for mascots).
  4. Click Generate.

Now, do this for every clip you make. When you stitch them together in your video editor, your character will always sound exactly the same.

The ElevenLabs Voice Changer interface with a video file uploaded
The ElevenLabs Voice Changer interface with a video file uploaded

Common Pitfalls to Avoid

I made these mistakes so you don’t have to:

  • Ignoring the “Subject” Box: In Whisk, if you don’t put your character in the “Subject” box, the AI will invent a random person every time.
  • Leaving “Precise Reference” On too early: When you are first creating the character (Step 1), let the AI be loose. Only turn on “Precise Reference” when you are putting them into scenes.
  • Expecting Perfection: AI video is still new. You might see a weird hand or a floating coffee cup. Just generate 3-4 versions and pick the best one.

Conclusion

That’s it! You don’t need a Hollywood budget. You just need a workflow. By separating the character creation (Whisk) from the video generation (Flow), you can finally tell a real story with characters that look the same from start to finish.

Have you tried this workflow yet? Let me know in the comments if you get stuck!


FAQs

  1. Are Google Whisk and Flow free to use?

    Yes! Right now, both tools are free because they are part of “Google Labs” (experimental tools). Google might charge for them in the future, but currently, you can use them without paying.

  2. I cannot find “Whisk” on Google. Where is it?

    These are test tools, so they can be hard to find. Search for “Google Labs” or “ImageFX” (which is often the public name for the image tool). You might need to sign into your Google Account to see them.

  3. Can I monetize these videos on YouTube?

    Yes, you can use AI characters in your YouTube videos. However, YouTube has a rule: you must check the box that says “Altered content” or “AI Generated” when you upload. If you are honest, it is fine!

  4. My character’s face looks blurry in the video. How do I fix it?

    This usually happens if the “Starting Frame” (Step 2) is too complicated. Try to generate a Starting Frame where the character is closer to the camera and the lighting is bright. If the input is clear, the video will be clear.

  5. Can I do this on my phone?

    Technically, yes. But honestly? It is very hard. You need to drag and drop files between tabs. I highly recommend using a Laptop or Desktop computer for this workflow. It will save you a lot of frustration.

Leave a Reply