chidiwilliams/buzz

What it solves

Buzz provides a way to transcribe and translate audio and video files offline on a personal computer, removing the need for cloud-based services and ensuring privacy.

How it works

It is powered by OpenAI's Whisper model and supports multiple backends for hardware acceleration, including CUDA for Nvidia GPUs, Apple Silicon for Macs, and Vulkan for various other GPUs. It can process uploaded files, YouTube links, or live audio from a microphone.

Who it’s for

It is designed for users who need high-quality speech-to-text transcription and translation locally, including event presenters needing live captions and developers looking for a CLI for automation.

Highlights

• Versatile Input: Supports audio/video files, YouTube links, and real-time microphone input.
• Enhanced Accuracy: Includes speech separation to handle noisy audio and speaker identification.
• Flexible Export: Transcripts can be exported as TXT, SRT, or VTT files.
• Automation Tools: Features a watch folder for automatic processing and a Command-Line Interface (CLI).
• Extensibility: Includes a plugin system for tasks like AI summary generation.