pluja/whisperx-api
API for Whisper STT with Whisper-X, enabling fast transcriptions and accurate speaker diarization.
10K+
pluja/whisperx-api repository overview
Whisper-X API
This is the stand-alone API that is used by anysub, the new work-in-progress version of whishper.
Supported tags and image size
Self host
Create a .env file, add the config variables and run:
docker run --name whisperx-api -p 8088:8000 --gpus=all pluja/whisperx-api
Visit http://localhost:8088/docs for the API documentation / UI
Docker Compose
services:
whisperx-api:
image: pluja/whisperx-api:cpu
ports:
- "8088:8000"
volumes:
- ./data/uploads:/app/data
- ./data/whisper_models:/app/wx_models
environment:
WHISPER_THREADS: 2
WHISPER_DEVICE: cpu
After running docker compose up for the first time, you will need to wait a few minutes for the models to download. You can start using the API when you see the following logs:
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
You can visit http://localhost:8088/docs for the API documentation / UI
Use with GPU
Only NVIDIA GPUs available.
services:
whisperx-api:
image: pluja/whisperx-api:latest
ports:
- "8088:8000"
volumes:
- ./data/uploads:/app/data
- ./data/whisper_models:/app/wx_models
environment:
WHISPER_THREADS: 8
WHISPER_DEVICE: cuda
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]
After running docker compose up for the first time, you will need to wait a few minutes for the models to download. You can start using the API when you see the following logs:
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
You can visit http://localhost:8088/docs for the API documentation / UI
Environment Variables
WHISPER_MODELS=tiny # Comma separated list of models to pre-load
WHISPER_THREADS=8 # Number of threads to run whisperX
WHISPER_DEVICE=cpu # Device to run (`cpu` or `cuda`). When using `cuda`, `cpu` is also available as an option.
To use diarization model
- Visit hf.co/settings/tokens to create your access token,
readpermissions are enough. - Create a
.envfile and add it as the value forWHISPER_HF_TOKEN - Visit pyannote/speaker-diarization-3.1 and accept terms.
- Visit pyannote/segmentation-3.0 and accept terms.
Example .env file
WHISPER_HF_TOKEN="hf_xxxx"
License
This part of Anysub and is licensed under Apache 2.0. You are free to use it as you wish. However, the community and me would be greatly thankful if you contribute your improvements, tweaks and fixes to the software back!
Tag summary
Content type
Image
Digest
sha256:671fb9db3…
Size
3.3 GB
Last updated
about 2 years ago
docker pull pluja/whisperx-api