Image to video
Image to video AI: animate a photo into a cinematic clip
Upload a photo or pick an image you generated, describe the motion and camera move, and the model renders a clip that starts exactly from your frame. Use first-and-last-frame mode to control where the shot ends, or multi-keyframe mode to drive a sequence from several stills.
Three ways to drive video from images
- First frame: the image becomes frame one; the prompt describes what happens next (all video models).
- First and last frame: two images pin the start and the end; the model interpolates the motion (Seedance, Wan 3.0, FLUX 3, Veo 3.1, Kling).
- Keyframe sequence: up to 10–30 ordered stills become one continuous shot (FLUX 3 Max, Seedance 2.5).
Tips that improve image-to-video results
Match the image aspect ratio to the output (16:9 or 9:16). Describe camera motion explicitly ("slow dolly in", "handheld pan left"). Keep the subject in the lower two-thirds if you want headroom for motion. Short clips (5–8 s) at 720p are the cheapest way to test a prompt before rendering 1080p.
Cost per clip
Image-to-video is priced like text-to-video: 150 BP for 5 s on Wan 3.0 at 720p, 350 BP for 5 s on Seedance 2.5 at 720p, 200 BP for 8 s on Veo 3.1 Fast. Reference images on Seedance count as extra input and are shown in the quote.
Frequently asked questions
Which image formats can I animate?
PNG, JPEG and WebP. Use a 16:9 or 9:16 image that matches the output aspect ratio for the cleanest result.
Can the video end on a second image?
Yes. Choose first-and-last-frame mode and upload both images; the model moves from one to the other.
Does the face stay the same?
With first-frame mode the first frame is your photo exactly; for later frames, Seedance 2.5 and Wan 3.0 hold identity best. For strict consistency across several clips, add the same image as a reference.