buzz

Buzz provides offline transcription and translation of audio and video files, including YouTube links, using OpenAI’s Whisper model. It runs on a personal computer without requiring cloud services, offering real‑time microphone capture, speech separation for noisy input, and speaker identification. The application supports multiple Whisper backends and can accelerate processing with CUDA on Nvidia GPUs, Apple Silicon, or Vulkan on integrated graphics.

The tool includes a graphical interface with a presentation window, searchable transcript viewer, playback controls, speed adjustment, and keyboard shortcuts for efficient navigation. Transcripts can be exported in TXT, SRT, or VTT formats, and a folder‑watch feature enables automatic processing of new files. A command‑line interface is also available for scripting and automation.

Buzz is distributed under the MIT license and is available as a native macOS .dmg, Windows installer, Flatpak, Snap, and PyPI package. It offers GPU‑enabled builds for Nvidia cards and supports Python 3.12 environments, making it suitable for users who need local, customizable speech‑to‑text and translation capabilities.