Download MacWhisper for macOS

100 languages, zero cloud uploads. MacWhisper runs Whisper speech recognition locally on your Mac, so every word stays on your machine.

How accurate is speaker separation?

Very. Version 15 brought a new diarization model that labels who said what — handy for podcast edits or multi-person meeting transcripts. You can delete or tweak any segment after.

What can I export from a transcript?

Word, PDF, HTML, SRT, and VTT all at once with batch export. Copy everything or just a highlighted chunk to paste into Notes or a client email.

Does it work alongside the apps I already use?

Yes. Capture system audio, transcribe a podcast as SRT for editing, or drag an exported PDF straight into Freeform or Bear. The reading mode turns a transcript into a teleprompter-style view if you’re recording voiceover.

Scan and mark up a speaker-labelled transcript before editing

Podcast editors who cut hours of raw tape down to a tight episode. MacWhisper hands them a speaker-labelled transcript they can scan and mark up before touching a timeline. Journalists get a first draft from an interview recording while the kettle boils. Students drop a two-hour lecture MP3 and walk away with searchable notes. Translators lean on the 100-language support to rough out a script before polishing. The common thread? Anyone who values speed and privacy over cloud uploads gets the most from it.

A podcast transcript, start to finish

Say you recorded a three-way discussion as a WAV. Here's one way through.

1. Drag the WAV onto MacWhisper. Pick Auto-detect language and the Large-V3 model for the cleanest output.
2. Wait a few minutes while your Mac's silicon chews through the file.
3. Scan the transcript. Speaker labels from the diarization model split the chat into clear blocks.
4. Click any segment and tap Delete to drop off-topic chatter. Edit a misheard phrase right in the text area.
5. Hit Export, tick SRT and PDF, then send the SRT to your editor and the PDF to the client for approval.

Fifteen minutes, one local app, no upload bar.

Settings to flip on day one

Out of the box, MacWhisper runs fine. A few tweaks make it feel tuned to your work.

• Default model. Set it to Medium or Large-V3 in Preferences. Small models are fast but can garble accented speech.
• Auto-detect language. Turn it on unless you always work in one tongue. Saves a click per file.
• Batch export formats. Tick the three you actually use — Word, SRT, PDF — so every export lands ready.
• Folders. Create a few right away (Podcast, Meetings, Lectures) and drag transcripts into them as you go. Much cleaner than a flat list after a week.

Speaker recognition now catches overlapping voices

Speaker recognition got noticeably sharper when version 15 shipped. The new diarization model picks up turn-taking better, even when two voices overlap. Real-time meeting transcripts appeared in the same release — start a session and watch text fill the pane as people talk. Folders and subfolders landed too, so the sidebar finally feels like a filing cabinet rather than a heap.

Version 15.2.1 tidied the edges. It checks your Mac model before offering a model download, so you won't grab one your chip can't run. The Local Models list dropped empty categories and added a Qwen provider icon, keeping things visually tidy. A crash on macOS 15 when opening the Live Captions language menu got squashed, and Qwen3-ASR support for M1-series Macs was pulled — the hardware simply didn't play well with it.

For anyone hunting a MacWhisper torrent or MacWhisper Pro torrent, these updates are reason enough to grab a build that includes the latest model compatibility and folder improvements. A MacWhisper Pro crack or MacWhisper cracked version can still deliver the real-time meeting transcripts that make 15.x worth having, so check that your setup includes those updated features. Even a plain MacWhisper crack is a solid way to jump right into what 15.x offers.

Good to know: batch transcription handles a folder of files in one go, and the live meeting transcripts work in real time.

• macOS 14 or later
• Apple Silicon or Intel 64‑bit