#c_lang

Moonshine Voice is an open-source toolkit for fast, private on-device speech-to-text that beats Whisper's accuracy and speed (up to 5x faster, 6.65% vs. 7.44% WER) with tiny 26MB-245M models for live apps on phones, Raspberry Pi, and more. It streams results as you speak, supports English/Spanish/Mandarin/etc., and handles transcription/commands easily via simple APIs on Python/iOS/Android. You benefit by building responsive voice apps offline without accounts, keys, or cloud costs—perfect for real-time tools like translators or assistants.

https://github.com/moonshine-ai/moonshine