vosk-api

Why we included this project

Vosk is a solid pick when you want speech recognition that runs entirely on the device, no cloud round-trip. Its models are small, around 50 MB, yet cover 20+ languages and support continuous transcription with a streaming API that responds without noticeable delay. That makes it work well for voice assistants, smart home gadgets, and live captioning. The toolkit offers bindings for Python, Java, Node.js, C#, C++, Rust, Go, and others, so you can use the same engine across mobile, embedded, and server code. It scales from a Raspberry Pi or Android phone up to larger clusters, and because everything stays offline, it fits privacy-sensitive setups where audio shouldn't leave the device. Speaker identification and reconfigurable vocabulary give you extra control when building custom transcription workflows.