Vibe is the most feature-complete free transcription app in this directory. It runs Whisper, Nemotron 3.5, and Parakeet TDT v3 locally, does speaker diarization, batch processes, captures system and website audio from YouTube and elsewhere, exports to eight formats, and offers multilingual summaries through the Claude API. There's a CLI and an HTTP API alongside the desktop app. All of it MIT licensed and free.
At a glance
Platforms
macOS, Windows, Linux. iOS and Android listed as coming
Pricing
Free, MIT licence
Models
Whisper, Nemotron 3.5, Parakeet TDT v3, plus Ollama support
Acceleration
GPU optimisation for NVIDIA, AMD, and Intel
Exports
SRT, VTT, TXT, HTML, PDF, JSON, DOCX
Features
Speaker diarization, batch processing, microphone and system audio capture, website audio from YouTube/Vimeo/Facebook/Twitter
Speaker diarization for free.MacWhisper charges €59 for this and Buzz is weak on it. For interviews and meetings it's the feature that makes a transcript usable, and Vibe gives it away.
Every export format you'd want. Eight of them, covering subtitling, document workflows, and programmatic use. The most complete set here.
System and website audio capture. Pull audio straight from a YouTube or Vimeo page, or capture what your machine is playing.
CLI and HTTP API. Scriptable and embeddable, which turns it from an app into infrastructure.
Newer models than the alternatives. Nemotron 3.5 and Parakeet TDT v3 alongside Whisper, with GPU paths for all three vendors.
Where it falls short
Summaries send data to Anthropic. Transcription is local, but the summarisation feature calls the Claude API. If you chose this for privacy, that's a boundary to be deliberate about crossing.
229 open issues. Substantially more than Buzz at three times the size. Ambitious feature set, and the bug count reflects it.
Not a dictation tool. Same as Buzz: transcribes recordings, doesn't type at your cursor.
Mobile is a promise. iOS and Android are listed as coming, not shipped.
Feature breadth over polish. A great deal of surface area maintained by a small team. Expect some of it to be rougher than the rest.
Pricing
Free, MIT. Claude API summarisation is billed by Anthropic if you use it.
Who it's for
Anyone doing serious transcription work who wants speaker labels, batch processing, and flexible exports without paying — journalists, researchers, and podcasters especially. Feature for feature it's the strongest free option. Choose Buzz instead if you want something more settled with fewer open issues and don't need diarization.
Last verified: 9 August 2026.
Similar to Vibe
Vibe: Free local transcription with speaker labels, summaries, and every export format – Dictation Database
Bring your own keys and mix any speech and language model you like
General Dictation
MIT-licensed voice typing for Mac, Windows, and Linux that works with Groq, Deepgram, AssemblyAI, OpenAI, Claude, Gemini, Ollama and more via your own API keys.
The most polished offline dictation on Linux, with a tray icon and everything
General Dictation
Free AGPL Linux dictation running whisper.cpp, Whisper, or VOSK entirely offline, with GPU acceleration and system tray integration across X11 and Wayland.
