Wan 2.2 I2V

I2V is image-to-video: you supply a first frame and the model animates from it. On Wan 2.2 it is one of the two video routes this workspace exposes, alongside text-to-video, and it is usually the faster route to a usable result because the composition is already decided for you.

Why starting from a frame changes the job

Text-to-video asks the model to invent the scene as well as the motion, so the prompt carries the composition, the lighting and the camera. Image-to-video fixes all of that in the frame you supply, and the prompt narrows to a single question: what changes?

That narrowing is why the same prompt behaves differently across the two routes. "A woman in a red coat walks through a rainy street" is a complete instruction for text-to-video and a poor one for I2V, because the model can already see the woman. "Slow push in, she turns towards the camera" is the I2V version, and it is shorter.

Picking a first frame that animates well

Source quality decides more of the output than any setting:

• Leave room to move. A subject cropped at the frame edge has nowhere to travel. A little air around the subject gives the camera somewhere to go.
• Sharpen before you animate, not after. A soft frame animates into a soft clip. Fix it while it is still cheap to regenerate.
• Match the aspect ratio to the destination. Deciding the crop now costs less than fixing it later.
• Small movement survives better. Asking a still frame for a large change makes the model invent most of the new picture, which is where warping and morphing appear.

How this differs from the GGUF builds

Searches for I2V often turn up quantised single-file builds meant to run on your own card. Those are a different thing: Wan 2.2 I2V GGUF covers that trade-off, including why image-to-video is harder to quantise than text-to-video — with a fixed first frame there is a ground truth to compare against, so the loss is visible rather than theoretical.

This page is about the hosted route. The model runs here, you choose a route and an output band, and the credit cost of your settings is shown beside the form before you submit.

Working order

1. Decide whether the shot already exists as a picture. If it does, stay on I2V.
2. Describe only the movement.
3. Iterate at a low output band, fixing the seed so that changing the wording is the only variable. Judging a small render as a proxy for a large one has limits: artefacts invisible at a low band can be obvious once the same shot is rendered bigger.
4. Re-run only the attempts you like at the size you will publish.

If a shot keeps failing the same way, change the frame or the route rather than adding clauses to the prompt. A prompt cannot fix a source image that is too soft.

Questions people ask

What does I2V stand for? Image-to-video — the route that takes a still frame as its input.

Can I use a photo? Yes. Any frame you own the rights to use works as the starting image.

Why does my animation warp? Usually the requested movement is too large for a single still frame. Ask for less and it holds together better.

Where do I start? Open the Wan workspace and choose the image-to-video tab.