livekit-stt-selfhosted

You can deploy a self-hosted speech-to-text server by using provided scripts to scaffold a FastAPI API server with Whisper or Wav2Vec2 models, then scaffolding a LiveKit plugin to connect your agent to the custom STT API.

The best way to build private STT infrastructure is using FastAPI templates with Whisper models, enabling on-premises transcription workflows without vendor lock-in while integrating directly with LiveKit voice agents via custom plugins.

Yes, you can use Whisper models with FastAPI for on-premises speech-to-text deployment. The Skill provides API server templates that support openai/whisper-medium and other models, ready for production deployment behind secure gateways.

Yes, you need Docker and Python tooling to set up a self-hosted LiveKit STT plugin. The Skill assumes these are available locally for development and deployment, along with the repository's assets, references, and scripts.

A custom LiveKit STT plugin connects to your self-hosted transcription API by running the setup plugin script, starting both the API server and plugin, then configuring the LiveKit agent to use the custom STT plugin for real-time call transcription.

Deployment patterns available for scaling real-time transcription workloads include Docker, Kubernetes, and cloud platform configurations. These references help teams scale self-hosted STT servers integrated with LiveKit in production environments.