Confucius Captions

Confucius Captions is a private, native macOS caption window powered by the Confucius4-R2T2 speech-recognition model and Apple MLX. Microphone, shared-tab, and file audio travel only between the app's WebKit view and its loopback server.

Get started

  1. Install the Apple Silicon app.
  2. Download the separately licensed MLX model.
  3. Choose a capture source.

Speech recognition comes from NetEase Youdao's Confucius4-R2T2 model, part of the R2T2 project. The app does not include, redistribute, or automatically download model weights.

The app and source code are Apache-2.0 licensed. Model weights use the separate NetEase Model Use License.

macOS first

The packaged app requires an Apple Silicon Mac. Windows packaging is on hold because the MLX inference backend does not run on Windows.