EiT - Speech to Text Module

This module uses WhisperLive (https://github.com/collabora/WhisperLive) package to achieve a nearly real-time transcription module including VAD (Voice Activity Detection). Modifications are made on the WhisperLive source code to fit this project.

Specifications

• A nearly real-time transcription module
• Two main models:
  • faster-whisper (a reimplementation of OpenAI's Whisper model)
  • Can handle all language but less accurate in Norwegian (also other languages most likely)
  • NbAiLab Whisper Model
  • For handling Norwegian dialects

Requirements

• Install docker (Only if using Docker)
• Install nvidia-container-toolkit (Only if with GPU)
• pip install requirements/server.txt (if running server locally)
• pip install requirements/client.txt

Running server

With Docker

GPU (recommended):

  docker build . -t whisper-live -f docker/Dockerfile.gpu
  docker run -it --gpus all -p 9090:9090 whisper-live:latest

CPU:

docker build . -t whisper-live -f docker/Dockerfile.cpu
docker run -it -p 9090:9090 whisper-live:latest

Locally

python run_server.py

Running client

python run_client.py