hwdsl2/whisper-server is a self-hosted speech-to-text server that provides an OpenAI-compatible API for audio transcription and translation. It is designed for users seeking a private, on-premise solution for converting audio files into text, with a focus on simplicity and security. The server operates as a Docker image and is powered by faster-whisper, enabling deployment on a variety of Linux systems, including those with amd64 and arm64 architectures.
The tool supports all Whisper models, including tiny, base, small, medium, large-v3, and large-v3-turbo, allowing users to choose the model that best fits their requirements. It offers speaker diarization as an optional local extension using sherpa-onnx, which enables identification of individual speakers in audio segments. Model management is facilitated through a helper script, whisper_manage, and a persistent model cache is maintained via a Docker volume. Audio files remain on the user's server, ensuring that no data is sent to third parties. The server accepts all major audio formats, such as mp3, m4a, wav, webm, ogg, flac, and any format supported by ffmpeg.
The API exposes endpoints compatible with OpenAI's POST /v1/audio/transcriptions and POST /v1/audio/translations, enabling seamless integration with existing applications that use the OpenAI Whisper API. Multiple response formats are available, including JSON, plain text, verbose JSON, SRT subtitles, and WebVTT subtitles. The platform also supports streaming transcription, where segments can be received via server-sent events as they are decoded. For users with NVIDIA GPUs, a CUDA-accelerated image is available, offering faster inference on supported hardware. The system can operate in offline or air-gapped mode using pre-cached models, and is suitable for both local and cloud deployments.
12-slim) and is part of a broader self-hosted AI stack. It is automatically built and published via GitHub Actions. The tool is intended for developers, organizations, and anyone needing private, self-hosted audio transcription and translation capabilities with OpenAI API compatibility.
In the Voice, TTS & speech space, hwdsl2/whisper-server takes a focused approach. It focuses on enabling private, self-hosted speech-to-text transcription and translation with an OpenAI-compatible API for developers and organizations. It is built as an open-source project for developers and organizations needing private audio transcription. hwdsl2/whisper-server is open source under the Open Source license. hwdsl2/whisper-server is available on the web, the command line, and API, and it can be self-hosted.
It is developed by hwdsl2, and the product first shipped in 2026. Development happens publicly on GitHub with 63 stars and 91 commits in the last 90 days. PulseGate's similarity index finds few close equivalents — hwdsl2/whisper-server occupies a relatively distinct niche. Key capabilities include speech-to-text, audio transcription, and audio translation. It exposes integrations via a public API.
Latest indexed changes and source events
hwdsl2/docker-whisper discovered by the PulseGate indexer
Other apps tracked under the same category.
