linto-platform-stt-standalone-worker-streaming
InnoLab changes
This forks use Linto, vosk and Kaldi to provide a Dockerised STT service compiled with Cuda.
Build
Download a model: https://alphacephei.com/vosk/models
Check for your GPU architecture tag: https://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/
Provide the path to the unzipped model folder and your GPU architecture tag in .env file.
Run docker compose:
docker-compose up