What it does
Buzz runs transcription and translation on your own machine rather than in the cloud. It is powered by OpenAI's Whisper and handles audio and video files, YouTube links and live microphone input, with an option to separate speech before transcribing to cope with noisy audio. It supports several Whisper backends and GPU acceleration through CUDA on Nvidia cards, Apple Silicon, and Vulkan with whisper.cpp on most GPUs, including integrated ones. Speaker identification is available, and transcripts can be exported as TXT, SRT or VTT. A transcript viewer offers search, playback controls and speed adjustment, and a presentation window is meant for events. For automation there is a watch folder that transcribes new files, a command-line interface, and a plugin system with examples such as AI summary generation. The README lists a macOS .dmg (Apple silicon only; Intel Macs are limited to version 1.4.5), a Windows installer that is not signed, Flatpak, Snap and AppImage builds for Linux, and a PyPI package.
Capabilities
- Offline transcription and translation with Whisper models
- Audio and video files, YouTube links and live microphone input
- Speaker identification and optional speech separation
- GPU acceleration via CUDA, Apple Silicon and Vulkan
- Export to TXT, SRT and VTT with a searchable transcript viewer
- Watch folder, command-line interface and plugins
When to use it
- You need subtitles or transcripts but do not want audio to leave your computer
- You want live transcription of a talk or event from a microphone
- You want to batch-transcribe new recordings from a folder or a script
