/local-whisper-fastapi

Primary LanguagePython

Local whisper api implementation

This is the server implementation of the telegram whisper bot

This implementation uses FastAPI to ingest mp3 files to the Whisper model and return a transcription of the audio.

Inside the repo there is a requests.py script that shows you how to use the endpoint

Working on localhost:8000, only uses the "/transcribe" endpoint

Installation

  • If you are using WSL you may have a problem with some dependencies, to fix it you have to add this line to the bash file
nano ~/.bashrc
LD_LIBRARY_PATH=/usr/lib/wsl/lib:$LD_LIBRARY_PATH
  • Because this is a FastAPI server, you have to install uvicorn
pip install uvicorn
  • Now create a virtual environment and install the requirements:
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
  • To run the server:
uvicorn main:app --reload

And the server will be ready to receive requests

Requirements

  • Python 3.8+
  • uvicorn

Usage

Just look at the requests.py file to know how to use the endpoint