Vosk is an offline speech recognition API supporting Android, iOS, Raspberry Pi, and server deployments. It provides continuous large vocabulary transcription with zero-latency streaming, speaker identification, and supports 20+ languages and dialects. The toolkit includes pre-trained models (50 Mb) and bindings for Python, Java, Node.js, C#, C++, Rust, and Go, built on Kaldi and deep neural network architectures.