cheetah

Why we included this project

Cheetah transcribes speech without sending audio to a server. It processes the stream as you talk, returning partial results mid-sentence, which is what makes it work for live captioning, voice commands, and interactive assistants that can't afford round-trip latency. Since everything runs on the device, it functions offline and keeps the audio private, which matters on constrained hardware like phones and embedded boards. The same engine has bindings for Android, iOS, web, desktop, and embedded platforms, so one integration can cover most of your targets. If you're shipping a voice-enabled product and want consistent on-device performance without paying per minute of transcription, this is a practical starting point.