Four reference modalities
Combine text, images, video, and audio to guide composition, motion, camera language, and sound.
Video-to-video workflow
Use the video-to-video generator when you already have footage and want to keep motion, framing, or timing while changing style, adding polish, or extending a shot. Start from a short reference video, describe the transformation clearly, and compare output against text-to-video or image-to-video when the source clip is not essential.
Bring a short clip when the camera move, pose, or timing already works and you want AI to reinterpret the look without rebuilding the shot from scratch.
Describe what should change, what must stay stable, the target style, and any motion constraints so the model has a clear editing brief.
If you only need a new idea or a still reference, text-to-video or image-to-video can be faster and cheaper than transforming an existing clip.
Use video to video when the existing clip matters: the camera path, actor movement, product angle, or timing should guide the result. Use text to video when you only need a new scene from a written idea.
Short, clear clips with one main subject and stable motion work best. Avoid very dark footage, heavy cuts, overlays, or clips where the subject changes too quickly for the edit you want.
State the desired transformation, the parts of the source video to preserve, and the final visual style. For example, keep the same camera movement and pose, but change the scene into a cinematic product commercial.