@siteed/moonshine.rn
React Native wrapper for Moonshine transcription. I use it to explore fast on-device speech workflows inside Audiolab.
Why?
I wanted a lighter transcription path next to Sherpa ONNX. Moonshine is interesting because it can give quick local transcription feedback, which fits the kind of audio apps I like building: private by default, useful offline, and testable on normal phones.
What I have been working on
- Native package shape: iOS, Android, and web support without pushing huge artifacts into the npm tarball.
- Long recordings: keeping memory bounded by retaining transcript/progress state instead of every audio segment by default.
- Audio Playground validation: live Moonshine transcription can run alongside Sherpa VAD / speaker-turn experiments from one microphone stream.
- Consumer install path: release checks around CocoaPods, Android artifacts, package size, and external app validation.
Status
This is still the part of Audiolab that changes the fastest. The package is usable, but I still treat it as an active experiment until the install story, memory profile, and cross-platform behavior feel boring enough.
Tech Stack
React NativeMoonshineSpeech-to-textOn-deviceiOSAndroidWeb