Silhouette models

Background-removal models converted for Silhouette, a Mac app that cuts subjects out of product photos and portraits. Each folder holds one model with its own README (what it is, where the weights come from, what the conversion changed) and the upstream licence, which travels with the files.

Folder Model Format Licence Size
isnet-dis IS-Net DIS (general use), Qin et al. 2022 Core ML .mlpackage, FP16 Apache 2.0 84 MB
modnet MODNet (photographic portrait matting), Ke et al. 2022 MLX safetensors, FP32 Apache 2.0 25 MB
birefnet BiRefNet (general), Zheng et al. 2024 MLX safetensors, FP16 MIT 420 MB

The app downloads these files pinned to a commit and checks each one's size and SHA-256 before using it, so a file here is never changed in place: a new conversion is a new commit. The MLX files hold weights only; the network itself is Swift code in the app.

Downloads last month
7
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support