Recast any clip. About half the price.
Send a clip and up to 8 pictures. Nemesis keeps the clip’s moves, timing and camera, and plays them with your people, products and places. One request on /v1/generations, model nemesis.
Before and after
Ten seconds of game footage with its song in, one picture of a stand-in, one request. Same moves, same stage, same song. Turn the sound on.
What a second of clip costs
Nemesis bills the clip in plus the video out, at $0.08, $0.17 or $0.36 a second, so one second of your clip is two billed seconds. These are the all-in prices per second of clip.
A 10 second clip at 720p: $3.40 on Revana, $6.81 on Higgsfield. Higgsfield figures are the Genjutsu Motion Transfer list prices per second of input video on open.higgsfield.ai, before any customer discount, checked 22 October 2026. A failed job costs nothing.
Nemesis Lite
The same request on a lighter engine, for clips of up to 15 seconds: "model": "nemesis-lite". Same inputs (a clip, up to 8 pictures, guide and sound), lower price, a little less faithful on faces and fine detail. It bills the clip in plus the video out at $0.07, $0.11 or $0.19 a second. These are the all-in prices per second of clip.
A 10 second clip at 720p: $2.20 on Nemesis Lite, $3.40 on Nemesis. A failed job costs nothing.
How it works
The clip becomes a guide
The people in your clip are turned into a depth map over the real background (guide: people), or the whole frame is (scene). The music is never sent.
Your pictures play it
The video model rebuilds every frame from your references: a person, a product, a location or a look, one role per picture.
The sound goes back on
With sound: original the voice drives the lips and the clip’s own sound is laid back on the result. Only for sound you have the rights to.
One request
The same key and shape as every other Revana model. Full options in the docs.
Flip any clip.
Prepaid credits, no subscription. You pay for the one route that delivers.