Vibe

Transcribe audio and video files on your own computer with Whisper
models. Vibe supports almost every language, batch transcription,
speaker identification, microphone recording and exports to SRT, VTT,
TXT, HTML, PDF and DOCX. Audio never leaves your machine; a network
connection is only needed to download models.

Pick and download a model on first launch. Vulkan GPU acceleration is used
when available. The vibe.cli command runs the transcription engine
without the window, for example:

vibe.cli transcribe ggml-tiny.bin recording.mp3

Connect the microphone to record: sudo snap connect vibe:audio-record.
Updates are managed by snapd.