whisper-cpp
agent-ready json output non-interactive
High-performance C/C++ port of OpenAI's Whisper speech recognition model. Runs locally, supports many languages, and outputs in multiple formats.
How to install whisper-cpp
brew install whisper-cpp When to use whisper-cpp
- Transcribing audio files locally when internet access is unavailable or privacy is a concern
- Generating word-level timestamps for precise alignment of subtitles or closed captions from audio recordings
- Processing multilingual audio content (e.g., meeting recordings in multiple languages) without relying on cloud-based APIs
When not to use whisper-cpp
- When real-time, low-latency streaming transcription is required (whisper-cpp is optimized for batch processing, not live streaming)
- When speaker diarization (identifying who spoke when) is needed, as the tool only produces text without speaker labels
whisper-cpp features
- · Local (offline) processing
- · Multiple output formats (txt, srt, json)
- · 100+ language support
- · GPU acceleration
- · Word-level timestamps
Want your agent to find this automatically?
Add the MCP server to your agent config and it will discover tools like whisper-cpp on its own.
Set up MCP →