The truly free, open-source, local-first, cross-platform CapCut replacement. Scriptable through a JSON-RPC API & MCP.
The compiled-in cutout model
The downloaded models are listed in THIRD_PARTY_NOTICES.md at the
repository root and named in src/models.rs.
selfie-segmentation.onnx is Google's MediaPipe Selfie Segmentation model
(the general, 256 × 256 variant), converted to ONNX and published by the
ONNX Community at
https://huggingface.co/onnx-community/mediapipe_selfie_segmentation.
Both the model and the conversion are licensed under the Apache License,
Version 2.0; see THIRD_PARTY_NOTICES.md at the repository root.
The file is compiled into concat-vision with include_bytes!, so a
cutout needs no download and works the same on a desk and on a phone. The
model takes one RGB picture as [1, 3, 256, 256] floats in 0..=1 and
answers with [1, 1, 256, 256] probabilities that each pixel is a person.