Whisper ASR Webservice
A self-hosted speech-to-text API that runs in Docker on CPU or CUDA GPUs, with OpenAI Whisper, Faster Whisper and WhisperX, under the MIT license.
Whisper ASR Webservice gives developers a speech-to-text API for their own applications and services. It processes audio on your machine or server and supports multilingual transcription, speech translation and language identification.
- Recognition engines: Choose OpenAI Whisper, Faster Whisper or WhisperX, with model choices including tiny, base, small, medium and large-v3.
- Timing and speakers: Produce word-level timestamps and use WhisperX speaker diarization to distinguish speakers in a recording.
- Audio and video input: FFmpeg support accepts a broad range of media formats, while voice activity detection filters out sections without speech.
- Transcript and subtitle output: Return plain text, JSON or TSV for transcripts and further processing, or VTT and SRT for subtitles.
The service is open source under MIT and runs in Docker with CPU processing or CUDA GPU acceleration. Its REST API includes Swagger documentation and a browser interface for trying requests. Configurable model loading and unloading control when models stay in memory; a persistent model cache avoids repeated downloads.
Logotyper
