OpenASR docs

OpenASR is a local-first speech-to-text project — a desktop app for macOS and Windows plus a Rust CLI / local HTTP API, all built on the same public Apache-2.0 open core at github.com/QuintinShaw/openasr.

Recommended

Use the desktop app

Open the app, choose a file, and start transcribing.

›

For automation

Use the command line

Start here for batches, scripts, and the local server.

›

The product posture is deliberately conservative:

native is the default executable backend. It runs local ggml-backed .oasr packs and is fail-closed by stage rather than fabricating output. A separate mock backend exists only as a hidden, opt-in stub for CI and plumbing tests — it isn’t part of the public CLI surface.
Normal CLI and API flows do not auto-download runtime binaries or model artifacts without a visible consent prompt.
Public model downloads are explicit through openasr pull and generated from the signed model-registry/catalog.json, and are verified before installation.
OpenASR does not phone home. There is no telemetry.

More docs

CLI reference for the full command surface and examples.
Docker for published CPU and CUDA images on Docker Hub.
Live Dictation & Captions for real-time microphone/system-audio transcription.
Speaker Diarization for per-speaker labeling and voice enrollment.
Configuration for the config file, environment variables, and OPENASR_HOME layout.
Server API for the local OpenAI-compatible transcription subset and the full endpoint reference.
Models for the current model catalog.
Project status for what’s shipped and what’s separate.