Skip to content
buildbay.
Vibe logo

Vibe

thewh1teagle · Audio

GitHub stars

7,673
—Collecting 30d data

Stats updated

7,627 stars on Sep 27 to 7,673 stars on Oct 2, up 46.6 measured star snapshots

What is Vibe?

Vibe turns local recordings, downloads, microphone input, and system audio into transcripts across desktop platforms. It combines several speech-model families with GPU acceleration, batch queues, diarization, stable subtitle timing, wide export support, local Ollama analysis, optional Claude summaries, and programmatic access through its CLI and HTTP API.

Vibe is a desktop app for offline audio and video transcription on macOS, Windows, and Linux. It supports batch transcription, speaker diarization, and exports including SRT, VTT, TXT, HTML, and PDF.

Core capabilities

  • Transcribe audio, video, microphone, system audio, or supported web media
  • Run Whisper, Nemotron, or Parakeet models with local hardware acceleration
  • Export subtitles and documents or automate work through CLI and HTTP API

Local model choice across desktops

Vibe supports Whisper, Nemotron, and Parakeet model paths across macOS, Windows, and Linux, with CoreML or Vulkan-oriented acceleration depending on the system. Users can customize model arguments, transcribe from files or live sources, and keep the default recognition step on their own machine.

  • Whisper
  • Parakeet
  • Nemotron
  • offline

Transcripts for many destinations

Batch queues, diarization, subtitle-focused stable timestamps, translation to English, printing, and exports ranging from TXT and JSON to SRT, PDF, and DOCX turn one recognition run into several deliverables. CLI commands, an HTTP API with Swagger documentation, Ollama, and optional Claude summaries extend the same output programmatically.

  • subtitles
  • diarization
  • exports
  • API

Where it fits

Use cases

  1. 01

    Produce subtitles for a recording

    Open a video or audio file, choose language and timing settings, apply speaker diarization when needed, and export SRT, VTT, or another supported format for editing or delivery.

  2. 02

    Process a local interview archive

    Queue multiple recordings, run transcription on the workstation, preview text while processing, and export searchable document formats without uploading the source media to a required transcription service.

  3. 03

    Connect transcripts to analysis

    Send completed text to a local Ollama workflow or an optional Claude-powered summary, or use the CLI and HTTP API to incorporate transcription and output handling into another tool.