AI Video Edit
Transform videos with AI — swap characters, restyle, transfer motion
On it. Running AI Video Edit now and I'll bring the results straight back here.
Try asking
About
AI Video Edit transforms existing videos using AI — swap characters, change environments, apply art styles, or transfer motion from one video onto a person in a photo. It's the fastest way to put yourself or a brand character into any scene without re-shooting.
Unlike traditional editing software, this tool works from natural language instructions. Describe the change you want — "replace the main character with me", "make it look like a Studio Ghibli anime", "put this person in the dance video" — and the AI handles the rest.
What you can do
- edit_video — targeted modifications: swap characters or objects, change environments, reference elements with @Element1/@Image1 syntax
- restyle_video — full style transfer: anime, oil painting, vintage film, watercolor, and other visual styles
- motion_control — place a person's photo into a scene video, inheriting all the motion from the original footage
- check_video — check the status of a running job and retrieve results when ready
Who it's for
Content creators who want to appear in video scenes they didn't film, brands that need to put their character or spokesperson into existing footage, social media editors doing viral remixes, and filmmakers experimenting with style treatments before a full production.
How to use it
- For character swaps, use edit_video with a video URL, a person element image, and a prompt like "swap the main character with @Element1"
- For style transfer, use restyle_video with a video URL and a style description like "Studio Ghibli anime" or "1970s film grain"
- To place a person into a scene, use motion_control with their photo as the image and the scene as the video
- All skills are async — the tool returns a job ID, and you can retrieve the result once processing completes
Getting started
Have a source video URL ready (MP4, MOV, or WebM, 3–10 seconds, up to 200 MB). For character swaps, also prepare a front-facing photo of the subject.
AI edits on an existing video — swap characters, change scenes, modify…
Broad style transformation of a video — anime, painting, weather…
Transfer motion from a reference video onto a character image — put…
AI edits on an existing video — swap characters, change scenes, modify objects — keeping original motion. Use @Element1/@Image1 in the prompt. Kling O1 default; runway-aleph2 for Aleph 2.0; sume- video_to_video models for lip-sync, captions, dubbing, upscaling. ~60–120s async.
Broad style transformation of a video — anime, painting, weather changes, time-of-day shifts — using the original as a motion/structure reference. Regenerates the video in a new style while preserving movement and composition. Powered by Kling O1. Input: 3-10s video. ⏱ Takes ~60-120s, runs async.
Transfer motion from a reference video onto a character image — put yourself in a movie scene or make a photo dance. Provide character photo (image_url) + motion video (video_url). Kling V3 Pro. ⏱ ~60-120s, async.
Check the status of a pending video edit/restyle/motion-control job and retrieve the result if complete. Use this when a previous call returned a pending status with a request_id.
- edit_video now runs Sume video models — lip-sync to new audio, captions, dubbing, upscaling, depth maps, motion transfer and object swap. Pass a sume- model key from generate-video list_models (type video_to_video).
- edit_video can now use Runway Aleph 2.0 — pass model runway-aleph2 to prompt-edit an existing clip. Kling O1 remains the default.
- Initial release — edit_video, restyle_video, motion_control, check_video
Powered by
Related
Questions
What is the difference between edit, restyle, and motion control?
`edit_video` makes targeted changes inside an existing clip, `restyle_video` transforms the overall visual style, and `motion_control` transfers motion from a source clip onto a character image. They solve different kinds of video transformation rather than repeating the same workflow.
What kind of source video works best?
Pass a direct video URL for a short clip, typically 3 to 10 seconds, in a common format like MP4, MOV, or WebM. The manifest also calls out a practical resolution range of roughly 720 to 2160 pixels.
Can I preserve the original audio?
Yes. The edit and restyle flows include audio-preservation options, and motion control can keep the original sound from the reference video when that helps the final result.
How do I know when a video edit is finished?
These edits run asynchronously. Start the job, keep the returned request ID, and use `check_video` to poll the status until the output video is ready.