Direct answer — How do you transcribe audio to text on a Mac offline? Use a tool that runs speech recognition on the Mac itself. Apple’s Dictation only types live speech, so for existing files you have two offline routes: whisper.cpp, a free command-line build of OpenAI’s Whisper, or a Mac app such as Weesper Transcribe, where you drop in an audio or video file, review the transcript and export it. Nothing is uploaded.
You have recordings on your Mac — interviews, lectures, a podcast episode, a saved meeting — and you want them as text, without uploading the audio to an online service.
This guide shows how to transcribe audio to text on a Mac offline: what macOS offers on its own, the open-source command-line route, and a step-by-step walkthrough with a drag-and-drop app.
Why transcribe audio offline on a Mac?
Because the audio never has to leave your computer. Offline transcription on a Mac means the speech recognition model runs locally, so the recording is processed where it already sits.
That matters in two situations:
- Confidential recordings. A source interview, a client call or an unreleased episode stays on your machine instead of on a third-party server.
- No connection, or a poor one. On a train, on a plane or on a slow network, on-device transcription works the same way.
On-device transcription is speech-to-text that runs entirely on your own hardware, using a model stored locally. If you want the wider trade-offs, we compared both approaches in our guide to on-device vs cloud transcription.
Can macOS transcribe an audio file on its own?
Not in a general way. macOS has three speech features, but Apple’s own documentation describes each one for live speech or for recordings made inside the app, not for an audio file you already have.
Dictation: live speech only
Apple describes Dictation as a way to “speak to enter text anywhere you can type it”. You start it with the Microphone key, a keyboard shortcut or Edit > Start Dictation, and it types what you say into the active text field.
It does not take an audio file as input. Apple’s guide points you to Keyboard settings to check whether Dictation input is processed on your device.
Voice Memos: transcripts of its own recordings
With macOS Sequoia and a Mac with Apple silicon, Voice Memos can show a transcript of speech in your recordings, though Apple notes the feature is not available in all countries or regions. You open a recording, click the Transcription button, then copy text or search across transcripts.
Apple’s guide covers recordings made in Voice Memos, including older ones it transcribes when you open them. It does not document importing an arbitrary audio file, and it does not say whether the transcription runs on device.
Notes: transcripts of audio recorded in a note
Notes can record audio inside a note and transcribe it, on Macs with M1 or later and in a list of supported languages that includes English, French, German, Spanish, Italian and Portuguese. The transcript is clickable: select a word to play the audio from that point.
Again, Apple’s guide covers audio you record in Notes. It does not describe transcribing an existing file, so for a backlog of recordings you need one of the two routes below.
How do you transcribe audio offline with whisper.cpp?
whisper.cpp is a free, open-source C/C++ port of OpenAI’s Whisper speech recognition model, released under the MIT licence. Its README calls Apple Silicon a “first-class citizen”, optimised through ARM NEON, Accelerate, Metal and Core ML.
It is a command-line tool, so it suits you if you are comfortable in Terminal. You will need Git, CMake and, for most files, ffmpeg.
Step 1: build whisper.cpp and download a model
The project’s quick start clones the repository, downloads a model and builds the tools:
git clone https://github.com/ggml-org/whisper.cpp.git
cd whisper.cpp
sh ./models/download-ggml-model.sh base.en
cmake -B build
cmake --build build -j --config ReleaseThe .en suffix marks an English-only model. OpenAI’s Whisper documentation notes that these tend to perform better for English, especially at the tiny and base sizes. For other languages, download a multilingual model, one without the suffix.
Step 2: convert your audio to 16-bit WAV
The whisper-cli example currently runs only with 16-bit WAV files. Convert an MP3, M4A or video file with ffmpeg first, using the command from the README:
ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le output.wavStep 3: run the transcription and choose an output
Run whisper-cli on the WAV file. The CLI can write plain text (-otxt), SRT (-osrt) or VTT (-ovtt) subtitles, CSV (-ocsv) or JSON (-oj):
./build/bin/whisper-cli -m models/ggml-base.en.bin -f output.wav -otxt -osrtThe spoken language defaults to English (-l en). For a French interview, pass -l fr, or -l auto to let the model detect it.
How do you transcribe audio to text on a Mac with Weesper Transcribe?
Weesper Transcribe wraps on-device transcription in a Mac app built around files you already have. It runs locally with on-device Whisper models, files are not uploaded, and it needs no microphone or Accessibility permission.
It is free to download from the Mac App Store, and the free version transcribes files up to 15 minutes at full quality. It requires macOS 13 or later, on Apple Silicon or on Intel Macs with AVX2 support. It does not do live dictation; that is the job of its sister app, Weesper Neon Flow.
Step 1: add your audio or video file
Drag a file into the app. It reads MP3, WAV, M4A, MP4, MOV and any other format macOS can read, so there is no conversion step for a video or a voice note.
Step 2: let it transcribe on your Mac
Transcription runs on your Mac, with Metal acceleration on Apple Silicon. The app supports 57 transcription languages with auto-detection, so a Spanish lecture or a German interview does not need a manual setting first.
Step 3: review and edit the transcript
Read the transcript with clickable timestamps that jump to the matching moment in the built-in player. Fix a name or a technical term directly in the text; inline editing is part of Pro and saved automatically.
Step 4: export in the format you need
Copy the clean text, or export to one of nine formats: TXT, Markdown, SRT, VTT, DOCX, PDF, HTML, CSV and JSON. Plain text is included in the free version; subtitle and document exports are part of Pro.
For video work, our guide to creating SRT and VTT subtitles offline covers the subtitle side in detail.
Step 5: keep it searchable, or process a whole folder
Every transcript is saved to a private history on your Mac, and Pro lets you search it by keyword or by meaning. Pro also lets you queue a folder of files and let the Mac work through them; see our walkthrough on how to batch transcribe audio and video files.
Pro is a one-time purchase, not a subscription. It removes the 15-minute limit and unlocks batch processing, pro exports, editing and search. The App Store shows the current price in your region before you buy.
Which offline transcription route fits you?
Pick by what you start from and what you need at the end. The table is based on Apple’s guides, the whisper.cpp documentation and the Weesper Transcribe listing.
| Apple Dictation | Voice Memos / Notes | whisper.cpp | Weesper Transcribe | |
|---|---|---|---|---|
| Input | Live speech | Recordings made in the app | 16-bit WAV (convert first) | Audio and video files macOS can read |
| Transcribes an existing file | ❌ | Not documented by Apple | ✅ | ✅ |
| Runs on device | Check in Keyboard settings | Not stated by Apple | ✅ | ✅ |
| Interface | System-wide | App | Command line | Mac app |
| Mac requirement | Built into macOS | Sequoia + Apple silicon (Voice Memos); M1 or later (Notes) | Builds from source | macOS 13+, Apple Silicon or Intel AVX2 |
| Export formats | Text where you type | Copy text | TXT, SRT, VTT, CSV, JSON | TXT, Markdown, SRT, VTT, DOCX, PDF, HTML, CSV, JSON |
| In-app editing and search | ❌ | Search transcripts | ❌ | ✅ (Pro) |
| Cost | Included in macOS | Included in macOS | Free, MIT licence | Free up to 15 min per file; one-time Pro |
If your files are interviews, lectures or podcast episodes, the dedicated pages for interview transcription, lecture transcription and podcast transcription go further on each workflow.
How do you get the most accurate offline transcript?
Accuracy depends heavily on the recording itself. A few habits make any offline transcription on a Mac cleaner.
- Start from the best source file. Use the original recording rather than a re-compressed copy shared through a messaging app.
- Reduce overlap and noise where you can. Whisper-based models handle one clear voice better than several people talking at once over background noise.
- Get the language right. In whisper.cpp, set
-lto the spoken language orauto, and pick a multilingual model for anything other than English. OpenAI notes that Whisper’s performance “varies widely depending on the language”. - Check names and jargon. Fix proper nouns and specialist terms before you export.
- Plan for long files. A multi-hour recording works in one pass in Weesper Transcribe Pro. On the free version, files are limited to 15 minutes each.
For saved video calls, see how to transcribe a Zoom or Teams recording offline.
Frequently asked questions
Can a Mac transcribe an audio file without an internet connection?
Yes, with the right tool. whisper.cpp runs Whisper locally and is optimised for Apple Silicon through Metal and Core ML. Weesper Transcribe does the same job in a Mac app: drop in an audio or video file and the transcript is produced on your Mac, with nothing uploaded.
Does Apple’s built-in Dictation transcribe audio files?
No. Apple describes Dictation as a way to speak and enter text anywhere you can type. It converts live speech into a text field; its guide does not cover transcribing an existing audio file.
Can Voice Memos or Notes transcribe a file I already have?
Apple’s guides describe transcripts for recordings made in those apps: Voice Memos with macOS Sequoia on Apple silicon, Notes on Macs with M1 or later. Neither guide documents importing an existing file, and neither says whether processing happens on device.
Is whisper.cpp free, and is it hard to use?
It is free under the MIT licence, but it is a command-line tool. You build it with CMake, download a model, convert audio to 16-bit WAV with ffmpeg, and pass -l for any language other than English.
How long a file can I transcribe for free with Weesper Transcribe?
Files up to 15 minutes, one at a time, to plain text, at full quality. Pro is a one-time purchase that removes the limit and unlocks batch processing, pro exports, editing and search.
Does Weesper Transcribe work on Intel Macs?
Yes. It is a universal app for macOS 13 or later, running on Apple Silicon and on Intel Macs with AVX2 support (Haswell or later).
Start with one real file
Transcribing audio to text on a Mac offline comes down to two workable routes. whisper.cpp is free and flexible if you are at home in Terminal; a dedicated app removes the conversion, editing and export steps.
Have a recording waiting? See how Weesper Transcribe turns files into editable, searchable text, then get it from the Mac App Store — free to download, one-time Pro upgrade, no subscription. Setup questions are answered in the Help Center.