Vibe is a cross-platform desktop application that transcribes audio and video files locally on your device using speech-to-text models.
The tool solves the problem of transcribing media while maintaining complete privacy by running transcription entirely offline, ensuring no audio data leaves your device. It supports multiple speech recognition models including Whisper, Nemotron 3.5, and Parakeet TDT v3, with GPU acceleration optimized for Nvidia, AMD, and Intel hardware. Beyond basic transcription, Vibe can extract audio from popular websites like YouTube and Vimeo, batch process multiple files, generate multilingual summaries via Claude API or local Ollama, translate transcripts to English, perform speaker diarization, and capture system audio or microphone input. Output formats include SRT, VTT, TXT, HTML, PDF, JSON, and DOCX.
Vibe suits developers and users who prioritize privacy and want transcription without cloud dependencies or subscription costs. It works well for processing large volumes of media files, creating subtitles for video projects, and workflows requiring local AI analysis. The application runs on macOS, Windows, and Linux, with a command-line interface for automation. A mobile feature allows recording from iOS or Android devices via QR code, with transcription handled on the desktop and results sent back to the phone.
The project maintains active development with a public roadmap tracking planned features. Documentation is comprehensive, covering local development setup, translation contributions, and debugging procedures. The maintainer welcomes pull requests and community translations, indicating openness to external contributions. An HTTP API with Swagger documentation and agent skills support enables integration into larger workflows.
