whisper.api

Why we included this project

whisper.api gives you a self-hosted speech-to-text service that runs whisper.cpp behind a REST and WebSocket interface. Because it mirrors Deepgram's /v1/listen endpoints, teams already using Deepgram can point the same client code at their own server and keep full control of their audio. Beyond plain transcription you get prompt-based vocabulary hints, audio cropping, speaker diarization, and SRT/VTT subtitle export, so it handles common media and meeting pipelines. It also supports live 16kHz PCM streaming over WebSockets, and the bundled CLI manages API keys and model files rather than leaving you to hack together access control. For a small team wanting an on-prem transcription service with real documentation and a Docker setup, this is a practical pick.