speech-recognition-inference

This repo provides an API that serves speech-to-text models from HuggingFace Hub. Simply choose the model and (optionally) revision that you would like to serve, and speech-recognition-inference will launch an API that performs automatic speech recognition on any audio file you specify.

Quickstart

Pip

First, install ffmpeg if this is not already installed on your machine.

Next, clone the repository and install:

git clone https://github.com/princeton-ddss/speech-recognition-inference.git
cd speech-recognition-inference
python -m venv venv
pip install --upgrade pip
pip install . # or pip install -e . for development

Now start an API from the command-line:

speech_recognition_launcher \
  --port 8000:8000 \
  --model_id openai/whisper-tiny \
  --model_dir $HOME/.cache/huggingface/hub

Once the application startup is complete, submit requests using any HTTP request library or tool, e.g.,

curl localhost:8000/transcribe \  -X POST \
  -d '{"audio_file": "/tmp/female.wav", "response_format": "json"}' \
  -H 'Content-Type: application/json'

Docker

We also provide a Docker image, princetonddss/speech-recognition-inference. To use it, simply run

docker run \
   -p 8000:8000 \
   -v $HOME/.cache/huggingface/hub:/data/models \
   -v /tmp:/data/audio \
   speech-recognition-inference:latest \
  --port 8000 \
  --model_id openai/whisper-large-v3 \
  --model_dir /data/models

Sending requests works exactly as shown above.

Detailed Usage

Environment Variables

In addition to specifying settings at the command line, some settings can be provided through environment variables. These settings are:

• SRI_MODEL_DIR
• SRI_HOST
• SRI_PORT
• HF_ACCESS_TOKEN

Command line arguments always take precedence over environment variables.