vosk and vosk-asterisk
The second is a specialized integration module that deploys the first as a server backend for Asterisk PBX systems, making them complements rather than competitors.
About vosk
alphacep/vosk
VOSK Speech Recognition Toolkit
Audio fingerprinting and LSH-based indexing enable training on massive speech datasets (100k+ hours) without neural networks, with incremental model improvement through direct sample addition. The system segments audio into chunks, stores them in a hash-indexed database for fast lookup during decoding, and integrates with Kaldi for phoneme alignment and segmentation. Supports lifelong learning paradigms with built-in verification tools to identify and correct recognition gaps.
About vosk-asterisk
alphacep/vosk-asterisk
Speech Recognition in Asterisk with Vosk Server
Integrates with Asterisk's native speech recognition framework via WebSocket connections to a separate Vosk Server instance, enabling offline speech-to-text processing through Kaldi models. Implements Asterisk dialplan applications (`SpeechCreate`, `SpeechBackground`) that interact with remote Vosk servers, supporting multiple language models deployable via Docker. Compatible across Asterisk versions 13-17+ with modular installation as `res_speech_vosk.so`, requiring `res_http_websocket.so` for transport.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work