Turn yourself into an animated cartoon character and create short videos with your cloned voice.
Upload your selfie
Use a clear, front-facing photo with good lighting. The cleaner the face, the better the cartoon. Avoid sunglasses, heavy shadows, or group photos.
Pick a cartoon style
Choose from Pixar 3D, anime, comic book, or pencil sketch. AtVeAnimation generates four cartoon versions of you — pick the one that looks most like you. This image becomes the visual anchor for all your videos.
Train your character
Your cartoon style image is used to train a personal AI model (LoRA) on Replicate. This takes 10–15 minutes. You can leave the page — training runs in the background and you'll see the status when you come back.
Add your own voice (optional)
Read 2–3 sentences aloud in a quiet room. The AI clones your voice and uses it to narrate your video scenes. Skip this step if you want silent videos.
Create a video
Click Create Video from your character page. In the studio, write scene descriptions or use the Write with AI button to generate them from a simple brief. Add voice scripts for each scene if you recorded your voice.
Watch it come together
Click Generate All Scenes to start generating. Each scene takes 3–8 minutes. When all scenes are ready, click Generate Final Video to stitch them together with audio.
Be specific and visual. Instead of "the character is happy", write "the character sits at a sunlit café table, smiling and holding a coffee cup, a warm golden light streaming through the window".
Include the setting. Where is the character? Time of day? Indoor or outdoor? The AI uses all of this to compose the scene.
Use the AI brief generator. Click Write with AI in the studio and describe your video in plain language — "A 30-second upbeat video of my character waking up, making coffee, and heading out for a run." The AI expands this into detailed scene prompts.
Keep voice scripts short. 1–2 sentences per scene work best for the voice cloning model. The script should feel natural when spoken aloud, not read.
Name your video first. Set the video title before clicking Generate — you'll find it in the header bar of the studio page.
Why does scene 1 look different from the rest?
Scene 1 is generated using your trained LoRA model — it's the most accurate representation of your character. Scenes 2+ reference scene 1 as the style anchor, so slight variation is normal but drift is minimised.
The background looks too realistic — what should I do?
Add visual keywords to your scene description like 'illustrated cartoon background', 'painted animation style', or 'cel-shaded environment'. The more you describe the art style, the better the result.
Can I reuse a character across multiple videos?
Yes — go to your character page and click Create Video. Each video is a separate project but all share the same trained character model.
How long does a video take?
Long enough for a coffee break and a walk. Character training runs about twenty minutes the first time (once per character). After that, each scene takes a few minutes. You can leave the page and come back — everything keeps running in the background.
Scene 1 audio didn't work — what happened?
Scene 1 generates image, video, and audio all at the same time, which can briefly hit Replicate's rate limit. The system retries automatically. If it fails after retries, scene 1 will be silent but subsequent scenes should have audio.