Image-to-video is where most anime shots actually get made: get the frame perfect as a still, then add just enough motion to bring it alive. The trap is asking for too much movement and watching your character melt. Here’s the workflow that stays on-model.

1. Win the still first

The video model inherits your starting frame — its composition, style, and character. Fix everything you can in the image lab before animating. A clean, well-lit, well-framed still is 80% of a good clip.

2. Prompt one motion, not three

Describe a single, clear movement: a slow push-in, a head turn, hair and cloth drifting in wind. Stacking multiple actions (“she runs, jumps, and the camera spins”) is the number-one cause of morphing and warping. Keep it readable.

3. Choose an image-to-video model

Bring your frame into an i2v model in the video lab. Check the IP-policy badge — strict models (cleaner but fussy) versus permissive models (better for fan-style work). Need help choosing? See choosing the right model.

4. Extend without breaking continuity

One clip is rarely a scene. Use AI Continuation to feed the last frame of a clip into the next i2v pass — the new shot starts exactly where the previous one ended, so motion and character carry over instead of resetting.

5. Lock the anime cadence

Smooth 30fps reads as 3D; anime lives on held frames. Retime your clip to on-twos, then layer procedural anime VFX — impact frames, speed lines, manga panel borders — for the finishing punch.

Make your first clip

Open the video lab, drop in a still, and prompt one clean motion. First render’s on us.