Free Software Collection
Free Speech-to-Text Tools
These tools run the same Whisper models, so accuracy depends on which model size you can run, not which tool you choose. The real difference is form: some are engines (command line or library), some are desktop apps you drag files into, and one is built for subtitles with word-level timestamps and speaker diarization. This collection sorts them by which layer you are missing. All figures verified August 2026 against their repositories.
5 curated recommendations
Selection Criteria
- Completely free, with no usage or duration caps on core features.
- Runs fully offline; audio never leaves your machine.
- No account required.
- Licences recorded from the actual LICENSE file, not repository auto-detection.
Recommended Tools
whisper.cpp
Whisper reimplemented in C++ — runs on CPU, no Python required
- Open Source
- Offline
- Cross-platform
- No Ads
faster-whisper
A CTranslate2 Whisper implementation — same models, faster, less VRAM
- Open Source
- Offline
- Cross-platform
- No Ads
WhisperX
Whisper for subtitles — word-level timestamps and speaker diarization
- Open Source
- Offline
- Cross-platform
- No Ads
Buzz
Drag a file in and transcribe — batch jobs and multiple export formats
- Open Source
- Offline
- Cross-platform
- No Ads
Vibe
The lightest local transcription desktop app — a few dozen MB to install
- Open Source
- Offline
- Cross-platform
- No Ads
FAQ
- Which one is most accurate for non-English speech?
- Wrong target for the question. All five run the same Whisper models: give them large-v3 and the output text is identical. Accuracy depends on the model size you can run — tiny is 75 MiB but error-prone, medium is 1.5 GiB, large-v3 is 2.9 GiB. On modest hardware try the quantized large-v3 at 1.1 GiB first.
- Which should I pick if I never want to touch a terminal?
- Vibe or Buzz — both are desktop apps where you drag a file in. Take Vibe if disk is tight (43.2 MB on Windows, 39.8 MB on macOS), Buzz for batch work and more export formats (449 MB for the macOS ARM64 build).
- What if I need subtitles and speaker labels?
- WhisperX. It adds word-level timestamps and speaker diarization on top of Whisper and is the only one here with both out of the box. The trade-off is a longer dependency chain and a fussier install.
- Is faster-whisper still usable?
- Yes, but know that maintenance has stalled: the last commit was 19 November 2025, the last release v1.2.1 on 31 October 2025, and 315 issues are open. The repo is not archived, so existing projects are fine; for something you plan to maintain long-term, prefer whisper.cpp.