Set up MacWhisper, choose between free and Pro models, and fix the usual problems: model download failures, slow large-v3 runs, missing speaker labels, exports.
Download from Gumroad or the Mac App Store. The free tier includes tiny/base/small models; Pro (one-time license) unlocks large-v3, batching and diarization.
Drag the file in, pick a model, wait for the in-app model download on first use.
Diarization is Pro-only and adds a second pass — expect roughly 2× the run time.
Same Whisper-class accuracy, no install, no model downloads, no GPU questions. Speaker labels, word timestamps and every export (TXT, SRT, VTT, DOCX) included. Free instant preview; credits from $2 and they never expire.
The models come from Hugging Face — corporate networks and some ISPs block it. Try another network, or download the .bin manually and drop it in via Settings → Models.
On Intel Macs that's expected — large-v3 is really an Apple-Silicon feature. On M-series, close other heavy apps; the Neural Engine build needs free memory.
That's how local diarization works — it separates voices but can't name them. Rename in the editor, or use a service that lets you relabel inline.
No — MacWhisper is file-based. Download the media first, or paste the link into a hosted service that fetches server-side.
Whisper's timestamps drift on multi-hour audio; re-align with word-level timestamps enabled, or split the file at chapter points.
More on MacWhisper: the full MacWhisper page · MacWhisper alternatives · all transcription tools