2.2.18 Backend: STT

faster-whisper-server

Handle: stt
URL: http://localhost:34151

faster-whisper-server is an OpenAI API-compatible Speech-to-Text server backed by faster-whisper. It exposes the /v1/audio/transcriptions endpoint.

Note

Harbor also ships Speaches — a newer OpenAI-compatible STT/TTS server that supersedes faster-whisper-server. Consider using speaches for new setups.

---

Starting

# [Optional] pre-pull the image
harbor pull stt

# Start the service
harbor up stt

On first start the service will download the configured STT model from Hugging Face. Track progress in the service logs:

docker logs $(harbor ps stt) -f

Configuration

# Get/set the STT model to use
harbor stt model
harbor stt model openai/whisper-large-v3

# Get/set the docker tag to use
harbor stt version
harbor stt version latest

Run the following command to see all available config options:

harbor config ls | grep HARBOR_STT

# Port on the host where OpenAI-compatible API will be exposed
HARBOR_STT_HOST_PORT      34151
# Docker tag to use for the image (platform suffix appended automatically)
HARBOR_STT_VERSION        latest
# STT model in user/repo format
HARBOR_STT_MODEL          Systran/faster-distil-whisper-large-v3

See Harbor's environment configuration guide to set arbitrary environment variables for the service.