The Changeling logo
Changeling

Imago

One face, onto any photo or video

Pick the face, pick the target, press Start. A completion bar shows what it's doing, how far it is and how long it has left; GFPGAN can clean the faces up after.

InsightFace inswapperGFPGANCUDA ffmpegPythonAGPL-3.0

What it looks like

The target, the face, and Changeling's result. Click to enlarge.

A target photo, the face to put on it, and the result

Use it

  1. Python 3.9 - 3.12 and ffmpeg on your PATH.
  2. Start run.bat (or run.ps1, run.sh). The first run sets itself up: a virtual environment, PyTorch, TensorFlow and the models.
  3. Select a face, select a target (a picture or a video), press Start and pick where the result goes.

The window

Startswap, then follow the completion bar
Previewone frame, before the whole video
Many facesevery face, not just one
Keep target fpsthe video's own frame rate

Without the window

run.bat -s face.png -t clip.mp4 -o out.mp4 swaps straight away; add --frame-processor face_swapper face_enhancer for GFPGAN.

Where it comes from

Changeling Imago grew out of roop by s0md3v, and like roop it's free software under the GNU AGPL-3.0. Its NSFW check stays on. Use it on faces you have the right to use: AI-generated people, yourself, or people who agreed to it.