Quick answer
Every Mac ships with free built-in Dictation (System Settings → Keyboard → Dictation, triggered with Fn+Fn). It works offline on Apple Silicon Macs (M1 through M4) and supports 40+ languages. For longer sessions, technical vocabulary, or transcribing existing audio files, third-party apps deliver better results: MacWhisper ($29) for file transcription, SuperWhisper ($249.99 lifetime) for system-wide dictation, and WhisperX (free, Python) for multi-speaker diarization. Apple Silicon opens the local-Whisper path; Intel Macs are cloud-first.
macOS Dictation (built-in) — setup + honest scope
macOS Dictation is Apple's built-in speech-to-text. It's free, ships with every Mac, and (on Apple Silicon) runs on-device. Setup takes under a minute.
How to turn it on
- Open System Settings → Keyboard.
- Scroll to Dictation. Toggle it on. Accept the prompt to download the language pack (needed for offline use on Apple Silicon).
- The default shortcut is Fn+Fn (press the Function key twice). Customize under Shortcut if it conflicts — common replacements: Option+D, Control+Space.
- Click into any text field. Press Fn+Fn. Speak. Press the same shortcut again to stop.
What works well
- Short bursts (a sentence, a paragraph, a search query, a text message).
- System-wide — works in any app, any text field.
- 40+ language packs.
- Offline on Apple Silicon Macs (M1–M4) once the language pack is downloaded.
- Punctuation via voice commands ("period," "comma," "new paragraph").
What breaks down
- Long sessions — some workflows time out or drop performance after extended dictation.
- Technical vocabulary — medical, legal, engineering, or brand names get more errors than Whisper-based tools.
- Multi-speaker recordings — Dictation is single-speaker only, no diarization.
- File transcription — Dictation processes live mic input only; it can't transcribe existing audio or video files.
- Live captions for meetings — for Zoom/Teams captions, use the app's built-in captions or a dedicated tool.
For casual use, macOS Dictation is genuinely capable. For anything longer, professional, or file-based, one of the ranked alternatives below will fit better.
Apple Silicon vs Intel — which options you actually get
The hardware axis matters more than tool choice for local speech recognition. Apple Silicon Macs (M1 through M4) have a Neural Engine that can run Whisper-class models locally, in real time, at low battery cost. Intel Macs (2019 and earlier) can technically run the same tools but without Neural Engine acceleration — you fall back to CPU inference, which is slow, or cloud APIs.
| Chip | Local Whisper (WhisperKit / MacWhisper / SuperWhisper) | macOS Dictation on-device | Practical recommendation |
|---|---|---|---|
| Apple Silicon (M1–M4) | Yes, Neural Engine accelerated. Whisper large-v3 runs faster than real time. | Yes, offline after language pack download. | Use local tools by default. Cloud only when accuracy or cross-device sync demands it. |
| Intel Mac (2019 or earlier) | Runs but CPU-only. Slow for larger models; smaller models usable. | Reduced on-device capability; often cloud-fallback. | Cloud tools (VexaScribe, Whisper API, Otter) usually better UX than local. Consider hardware upgrade for local Whisper. |
Check your chip in Apple menu → About This Mac. If it says "Apple M1" through "Apple M4" (or newer), you have Apple Silicon. If it says "Intel Core," you don't.
Cloud vs local — the privacy tradeoff
The second-most-decisive axis after hardware. Both paths are legitimate — pick based on your specific workflow.
Cloud (VexaScribe, Whisper API, Otter, Deepgram)
- Higher accuracy on some benchmarks (server-side runs largest models)
- Requires internet
- Audio leaves the device — check vendor data retention + training policy
- Cross-device workflow built in
- Per-minute or subscription pricing
Local (WhisperKit, MacWhisper, SuperWhisper, Whisper CLI)
- Audio never leaves your Mac
- Works offline
- Requires Apple Silicon for reasonable speed
- Requires 1–10 GB storage per model
- One-time cost or free — no per-minute billing
Rule of thumb: local when handling confidential audio (legal, medical, HR, personal therapy). Cloud when you need cross-device workflow, team collaboration, or the very latest server-side models. Modern Apple Silicon runs local Whisper large-v3 faster than a cloud round-trip on many networks — so "cloud is faster" is often no longer true.
Ranking methodology
Each "Best for X" ranking below weighs five criteria. Rankings are per-use-case, not global — a tool that's #1 for file transcription may not appear at all in the system-wide dictation ranking, because the tool doesn't support that job.
| Criterion | Weight | What we measure |
|---|---|---|
| Accuracy | 25% | WER on standard benchmarks (LibriSpeech, Common Voice) + real-world Mac dictation tests. |
| Use-case fit | 25% | Whether the tool is designed for the specific job (system-wide vs file vs live). |
| Price for value | 20% | One-time cost, subscription, or free — vs what you get. |
| Privacy | 15% | Local vs cloud, data retention, opt-out from training. |
| Setup friction | 15% | Time from install to first successful dictation or transcription. |
Our own product, VexaScribe, is web-based — it competes fairly in the file-transcription and multi-speaker rankings (where cross-device + diarization matter) and doesn't appear at all in the system-wide dictation ranking (where a native Mac app is the right shape). We wrote it that way because ranking us #1 for a job we don't fit would be dishonest.
Best for system-wide dictation
Spoken into any Mac app (email, Slack, Word, Notes, code editor). All four options here are Mac-native.
Best for: Cross-app dictation on Mac + Windows + iOS with WhisperKit / Parakeet models
Longest-standing category leader. Global hotkey, works in any Mac app. Local models keep audio on device.
Best for: Newer, cleaner UX. Voice Ink uses SpeechAnalyzer on iOS 26+ for on-device dictation
Faster onboarding than SuperWhisper if you dislike setup.
Best for: Mac-only dictation with the same tool that handles file transcription
Best value if you want both dictation and file transcription in one app.
Best for: Short bursts, any text field, any language pack you download
Genuinely capable for casual use. Struggles with longer sessions and technical vocabulary.
Best for file transcription
Transcribe existing audio or video files (podcasts, meeting recordings, interviews, voice memos).
Best for: Mac-native batch file transcription
Best solo Mac experience. Local models, no cloud round-trip, drag-drop UI.
Best for: Free file transcription with no signup
Free forever, local, no account required. Fewer features than MacWhisper but genuinely free.
Best for: Cross-device workflow + speaker labels for up to 50 voices
Web-based so files sync across Mac, iPhone, Windows. Best when diarization matters.
Best for: Scriptable batch pipelines, developer workflows
Command-line only. Best when you want to script transcription into a larger workflow.
Best for: Edit workflows + cross-device sync + team collaboration
Choose these if editing and collaboration matter more than raw transcription cost.
Best for multi-speaker recordings (diarization)
Interviews, panel discussions, group meetings — anything with 2+ voices where you need "Speaker 1 / Speaker 2" labels.
Best for: Highest-quality local diarization on Mac
Requires Python setup. Pairs Whisper + pyannote for speaker labels + word-level timestamps.
Best for: Zero-setup diarization, up to 50 speakers
Best if you don't want to install Python. Diarization included, no extra cost.
Best for: Local diarization with GUI
Requires accepting pyannote model terms on HuggingFace.
Best for: Developer path for building diarization into your own app
Not a consumer app — API only.
Diarization is a separate model from ASR. Even the best transcription without diarization can't label speakers. See /what-is-speaker-diarization for the technical explainer.
Best genuinely free options
No signup, no trial, no watermark. Ranked by immediate usefulness.
Best for: Works today, no downloads
System Settings → Keyboard → Dictation → toggle on.
Best for: Free file transcription, no signup
App Store download. Local Whisper, no cloud.
Best for: Unlimited free file transcription
Command-line. Requires Python 3.8-3.12 + FFmpeg.
Best for: Try Whisper large-v3 without installing anything
30 min free trial, no credit card. Web-based.
Best for privacy (100% offline)
Audio never leaves your Mac. Required for confidential recordings (legal, medical, HR, personal therapy) and for anyone who prefers on-device processing on principle.
Best for: System-wide, on-device, free, no setup
Apple Silicon Macs process dictation locally on the Neural Engine.
Best for: GUI transcription that never leaves your Mac
Local Whisper via WhisperKit / Core ML.
Best for: Fully offline, command-line, unlimited
The purest privacy option — nothing leaves your machine.
Best for: Cross-app dictation with local models
Uses WhisperKit for on-device processing.
Full comparison table
All 10 tools tested, side by side. "AS" = Apple Silicon. "Both" = supports system-wide dictation and file transcription.
| Tool | Type | Price | Offline | Diarization | Apple Silicon | Best for |
|---|---|---|---|---|---|---|
| macOS Dictation | System-wide | Free | Yes (AS) | No | Preferred | Casual dictation into any app |
| SuperWhisper | System-wide | $249.99 lifetime | Yes | No | For local models | Pro cross-app dictation |
| Wispr Flow | System-wide | Subscription | No (Flow cloud) | No | No | Cleanest UX for cloud dictation |
| Voice Ink | System-wide | One-time | Yes | No | For SpeechAnalyzer path | iOS 26+ on-device dictation |
| MacWhisper | Both | $29 one-time | Yes | With Pro + HF token | For local | Mac-native transcription + dictation combo |
| Aiko | File only | Free | Yes | No | Preferred | Free file transcription, no signup |
| VexaScribe | File only (web) | 30 min free → paid | No | Yes (up to 50 speakers) | No | Cross-device + speaker labels |
| Whisper CLI | File only | Free (open source) | Yes | With WhisperX | Preferred | Scriptable pipelines |
| WhisperX | File only | Free (open source) | Yes | Yes (pyannote) | Preferred | Best local diarization |
| Otter (cloud) | Both | Freemium | No | Yes | No | Team meetings + collaboration |
When to skip all of these
Otter, Fireflies, or Zoom's built-in transcription — browser-based, cross-platform, joins meetings as a bot.
Rev, GoTranscript — 99%+ accuracy with human editors, priced per audio minute.
Nuance Dragon Medical, DAX, or dedicated compliance vendors. Not a job for consumer Mac dictation tools.
iOS dictation is different from Mac dictation. See /transcribe-iphone-video or /iphone-voice-memo-transcription.
Verified sources
Every price, feature, hardware claim, and benchmark on this page was cross-checked on 2026-08-16 against the following primary sources:
- Apple support — macOS Dictation (System Settings → Keyboard → Dictation) documentation and shortcut customization.
- Argmax WhisperKit — open-source Swift package for Whisper on Core ML + Neural Engine (github.com/argmaxinc/WhisperKit).
- MacWhisper — vendor pricing (Gumroad), Pro / diarization feature scope.
- Superwhisper — vendor pricing ($249.99 lifetime), platform support (Mac + Windows + iOS).
- Aiko — App Store listing (free file transcription).
- Apple developer docs — iOS 26 SpeechAnalyzer, SpeechTranscriber, DictationTranscriber APIs.
- 2026 benchmarks — embertype.com, JustVoice, MacParakeet Whisper-on-Apple-Silicon reports for real-world WER + speed measurements.
- Apple product pages — M4 Neural Engine 38 TOPS specification and Apple Silicon Neural Engine capabilities.
Vendor prices can change; we bumped verification date on ship. Benchmark numbers vary by test setup — the ranking positions above reflect consensus across multiple 2026 sources, not any single benchmark.
Frequently asked questions
How do I turn on speech to text on Mac?
Open System Settings → Keyboard → Dictation and toggle it on. Accept the prompt to download the language pack (needed for offline use on Apple Silicon Macs). The default shortcut is Fn+Fn (press the Function key twice). Click into any text field, press Fn+Fn, and start speaking. Press the shortcut again to stop.
Does macOS have built-in speech to text?
Yes. Every Mac running modern macOS ships with Dictation, Apple's built-in speech-to-text. It supports 40+ languages, works offline on Apple Silicon Macs (M1 through M4), and runs system-wide in any text field. Enable it in System Settings → Keyboard → Dictation. It's free, requires no third-party install, and integrates with any Mac app.
How to convert voice to text on Mac?
For live dictation into any Mac app: use macOS Dictation (System Settings → Keyboard → Dictation, then Fn+Fn to trigger). For transcribing an existing audio or video file: install MacWhisper ($29) or Aiko (free) for local transcription, or upload to a web tool like VexaScribe (30 min free) for cross-device workflow with speaker labels. For system-wide dictation with better accuracy than the built-in: SuperWhisper ($249.99 lifetime) or Wispr Flow / Voice Ink are the current category leaders.
What is the shortcut for speech to text on a Mac?
The default macOS Dictation shortcut is Fn+Fn — press the Function key twice quickly. On some MacBooks the Fn key is in the bottom-left corner. You can customize the shortcut in System Settings → Keyboard → Dictation → Shortcut → Customize; common replacements include Option+D or Control+Space to avoid conflicts with other Fn-key features.
Does macOS Dictation work offline?
On Apple Silicon Macs (M1 through M4), yes — once you've downloaded the language pack, Dictation runs on-device with no internet connection. On Intel Macs (2019 and earlier), offline support is more limited and depends on the specific macOS version. To ensure offline mode is active, download the offline language pack when prompted after enabling Dictation.
Do I need Apple Silicon for local Whisper on Mac?
You don't strictly need it, but Apple Silicon makes a huge difference. WhisperKit and CoreML-based tools accelerate Whisper on the Neural Engine — Whisper large-v3 runs faster than real time on M1 through M4 Macs. On Intel Macs, the same tools run but fall back to CPU inference, which is much slower for larger models. If you're on an Intel Mac and need local transcription, use smaller Whisper models (base, small, or medium) or prefer cloud tools like VexaScribe.
What's the difference between macOS Dictation and SuperWhisper?
macOS Dictation is Apple's free, built-in speech-to-text — works in any text field, offline on Apple Silicon, supports 40+ languages. Best for casual dictation. SuperWhisper is a paid ($249.99 lifetime) third-party app that adds: better accuracy via local Whisper or Parakeet models, longer sessions without timeouts, better handling of technical vocabulary, and cross-platform support (Mac + Windows + iOS). If you dictate casually, Dictation is enough. If you dictate professionally for hours a day, SuperWhisper pays for itself in fewer corrections.
Can I use ChatGPT for speech to text on Mac?
Yes, with caveats. ChatGPT's Record feature (macOS desktop app, Plus/Pro/Business/Enterprise/Edu tiers) captures live audio and produces a transcript with summaries and quotes, up to a 4-hour session cap. For transcribing existing audio files, ChatGPT accepts uploads up to 25 MB (roughly 10-15 minutes at typical bitrates) with GPT-4o transcription — accuracy is usable but no timestamps and no speaker labels. For longer files or professional transcription, dedicated tools like MacWhisper or VexaScribe outperform. See our full breakdown at /chatgpt-transcription.
Is MacWhisper better than the built-in Mac Dictation?
Different jobs. MacWhisper is a file-transcription app — you drag an audio or video file in and get a transcript. Dictation is live dictation — you speak into a text field and text appears. If you need to transcribe an existing recording (podcast, meeting, interview), MacWhisper wins because Dictation can't accept file input. If you need to dictate live into an app, Dictation wins on speed and integration. Some MacWhisper users also use it in its dictation mode as a Dictation upgrade for more accuracy.
How accurate is Mac speech to text?
Depends on the tool. macOS Dictation reaches ~92-97% accuracy in quiet environments with clear speech, dropping to 75-85% with heavy accents or background noise. Local Whisper large-v3 (via MacWhisper, SuperWhisper, or WhisperKit apps) reaches 92-95% on clean English audio and generally outperforms native Dictation on technical vocabulary and longer sessions. Cloud tools using the latest server-side models (VexaScribe with Whisper large-v3, Deepgram Nova, AssemblyAI Universal-2) match or exceed local Whisper on some benchmarks.
Related guides
Broader speech-to-text landing surface — browser dictation, file upload, all platforms.
If you want to dictate specifically into a Google Doc, the workflow is different.
Set up local Whisper on your Mac — CLI, faster-whisper, WhisperX, whisper.cpp.
The technology behind Whisper, macOS Dictation, and every speech-to-text tool.