Your voice, on your machine.
Offline dictation and read-aloud for the desktop.
Built with Electron, React, TypeScript, and sherpa-onnx.
Downloads · Contribute · Report an issue
Download a model once, then run inference locally. Model downloads need an internet connection. Desktop packages target Windows, macOS, and Linux; permissions and text insertion vary by platform. Releases are distributed through electron-builder update manifests; CI artifacts are currently unsigned and development builds do not contact the update feed.
Speech-to-text includes NVIDIA Nemotron ASR Streaming, an English cache-aware streaming model with a 560 ms latency/accuracy profile. It runs through sherpa-onnx on the CPU after its verified model archive is downloaded; the model remains subject to the NVIDIA Open Model License.
Use Node.js 24 LTS and npm. Native packaging prerequisites are in CONTRIBUTING.md.
git clone https://github.com/shikki841/inaudio.git
cd inaudio
npm ci
npm start
npm run check # TypeScript + ESLint
npm run package:release # Build the electron-builder package for your current OS
Pick a model during setup and grant microphone access when prompted.
Small fixes, platform testing, and clear bug reports help. Read the contribution guide, community guidelines, and release process. Report vulnerabilities through the security policy.
MIT. Downloaded models and dependencies retain their own licenses.
Your voice, on your machine.
Offline dictation and read-aloud for the desktop.
Built with Electron, React, TypeScript, and sherpa-onnx.
Downloads · Contribute · Report an issue
Download a model once, then run inference locally. Model downloads need an internet connection. Desktop packages target Windows, macOS, and Linux; permissions and text insertion vary by platform. Releases are distributed through electron-builder update manifests; CI artifacts are currently unsigned and development builds do not contact the update feed.
Speech-to-text includes NVIDIA Nemotron ASR Streaming, an English cache-aware streaming model with a 560 ms latency/accuracy profile. It runs through sherpa-onnx on the CPU after its verified model archive is downloaded; the model remains subject to the NVIDIA Open Model License.
Use Node.js 24 LTS and npm. Native packaging prerequisites are in CONTRIBUTING.md.
git clone https://github.com/shikki841/inaudio.git
cd inaudio
npm ci
npm start
npm run check # TypeScript + ESLint
npm run package:release # Build the electron-builder package for your current OS
Pick a model during setup and grant microphone access when prompted.
Small fixes, platform testing, and clear bug reports help. Read the contribution guide, community guidelines, and release process. Report vulnerabilities through the security policy.
MIT. Downloaded models and dependencies retain their own licenses.