whisper

https://lablab.ai/t/whisper-api-flask-docker

Develop

• poetry run python3 -m flask run --host 0.0.0.0 --port 5000

Config (env)

• WHISPER_MODEL (default small) · WHISPER_DEVICE (cpu) · WHISPER_COMPUTE (int8) · WHISPER_CPU_THREADS (0 = auto) · WHISPER_LANG (blank = autodetect) · WHISPER_VAD (true = drop non-speech, so silence transcribes to nothing).

Docker

• sudo docker run -p 5000:5000 dataspects/whisper:1.0.2
• sudo docker login && sudo docker push dataspects/whisper:1.0.2

Test

• curl -F "file=@/path/to/file" http://dataspects-docker-server:5000/whisper

Converters

• ffmpeg -i *.m4a *.wav