"The quarterly results exceeded our expectations, with revenue growing by twenty-three percent..."
Local Speech · 98 Languages · Free & Open-Source
Your voice, your Mac.
Nothing between.
Dictate anywhere on macOS. Transcribe files and YouTube. Record meetings with system audio and mic — with speech recognition on your Mac, not in a cloud STT pipeline. Open-source on GitHub.
Version 0.8.0 · macOS 14.2+ · Apple Silicon
brew install --cask macparakeet "The quarterly results exceeded our expectations, with revenue growing by twenty-three percent year over year..."
Three modes. One app.
Dictate anywhere on macOS. Transcribe files and YouTube. Record meetings with system audio and mic. Speech stays local with task-fit engines.
Dictate anywhere
Press your hotkey. Speak. Text appears wherever your cursor is — any app, any text field. The dictation Apple should have shipped.
Transcribe files & any URL
Drop an audio or video file. Paste a link — YouTube, Loom, Vimeo, a podcast, almost anything. Get text in seconds. Interviews, lectures, voice memos — anything with speech becomes searchable text.
Files, plus YouTube, Loom, Vimeo, TikTok, podcasts & more
Transcribe any video URL →Record meetings
Capture system audio and your mic at the same time. Take notes during the call, watch the live transcript, ask questions as it unfolds. Audio never leaves your Mac.
Granola-style workflow, fully local
AI meeting recorder for Mac →Everything you need
Built for people who care about speed, privacy, and getting out of the way.
98 Languages
Korean, Japanese, Chinese & more via WhisperKit
Cohere Batch Transcription
Optional local plain-text engine for files
Live Meeting Notes
Type during the call. Slash commands. Auto-saved.
Custom Hotkeys
Any key, or any chord, for dictation
Any video URL
YouTube, Loom, Vimeo, TikTok & more
8 Export Formats
TXT, Markdown, SRT, VTT, DOCX, PDF, JSON, DAPT
Speaker Diarization
Know who said what
Custom Vocabulary
Teach it your jargon
Bring Your Own LLM
OpenAI, Anthropic, Ollama, LM Studio — optional
Your audio never leaves your Mac
Core speech recognition runs on your Apple Silicon chip. No cloud STT servers. No accounts. No audio or transcript content collection. Optional telemetry is anonymous and opt-out; optional AI features send transcript text only when you choose a provider.
Local speech engine
Your audio never leaves your Mac
Dictate into any app on your Mac
Slack, Gmail, VS Code, Notes, Pages, your terminal -- if it has a text cursor, MacParakeet can type into it. The dictation overlay floats above your windows without stealing focus, so you never lose your place.
If it has a text cursor, MacParakeet works
September 2026 · Version 0.8.0
More dependable recording. A Library that works your way.
A visual fingerprint for every recording.
When media has no artwork, MacParakeet now gives it a deterministic generated cover.
MacParakeet 0.8.0 improves recovery across microphone and audio-route changes, makes transcripts easier to work with, and adds the Library and prompt controls requested most.
Recording you can trust
Better recovery for microphone, Bluetooth, AirPods, and route changes; safer meeting finalization; and saved notes that can guide post-meeting prompts.
A more useful Library
Persistent grid or list layouts, editable labels with stable colors, clearer favorites and audio state, and generated covers.
Richer transcript workflows
Speaker corrections, rich Markdown, prompt collections, model-aware generation controls, and W3C DAPT export.
Three steps. That's it.
Install
Download and run. Setup takes 60 seconds. No account, no sign-up, no email.
Speak or Drop
Press Fn to dictate into any app. Drop an audio file or paste a YouTube URL to transcribe it.
Done
Text appears where you need it. Copy, export, or just keep working.
Free and open-source.
No paid tier. No "community edition" with features stripped out. The app you download is the app we develop.
- Unlimited transcription & dictation
- Full source code on GitHub
- All audio & video formats + YouTube URLs
- 8 export formats · Speaker diarization · Custom vocabulary
No account. No subscription. No catch.
How we compare
Other apps charge $30-250. MacParakeet is free and open-source.
| MacParakeet | WisprFlow | MacWhisper | |
|---|---|---|---|
| Price | Free open-source | $12-15/mo | $30 once |
| Free tier | Everything | 2,000 words/wk | No |
| Open-source | Yes GPL-3.0 | No | No |
| Local speech engine | Yes | Partial | Yes |
| Parakeet TDT (155x) | Yes | No | Added later |
| System dictation | Yes | Yes | Yes |
| Export formats | 8 formats | TXT | 5 formats |
| Video & podcast URLs | Yes | No | Yes |
| Speaker diarization | Yes | No | Yes |
| No account needed | Yes | No | Yes |
Questions & answers
Is my audio really private?
Yes. All transcription and dictation runs locally on your Mac's Apple Silicon. Your audio never leaves your device. If you optionally connect a cloud AI provider for summaries or chat, only the transcript text is sent -- never the audio itself, and only when you choose to.
Does MacParakeet collect any data?
Only anonymous usage statistics — like which features are popular and how long transcriptions take. No personal data, no transcription content, no audio, no IP addresses. Session IDs reset every launch, so there's no way to track you across sessions. You can opt out anytime in Settings.
Do I need an internet connection?
Only for initial setup (downloading the speech model). After that, dictation and file transcription work completely offline. Transcribing from a URL needs a connection to fetch the audio first, then the speech recognition runs on your Mac. Optional AI features (summary, chat) also need a connection.
Can I transcribe Loom, Facebook, TikTok, Vimeo, and podcast links — or only YouTube?
Any of them. Paste a link from YouTube, Loom, Vimeo, TikTok, Facebook, Instagram, X (Twitter), Twitch, Apple Podcasts, or roughly a thousand other sites — anything yt-dlp supports. MacParakeet downloads the audio and transcribes it locally on your Mac. Transcribe content you have the right to use.
What's the difference from MacWhisper?
Both are local transcription apps for Mac with system-wide dictation. MacWhisper started with Whisper and later added Parakeet support. MacParakeet is built around Parakeet from day one and is free with no subscription.
What Mac do I need?
macOS 14.2 or later with Apple Silicon (M1, M2, M3, or M4). Intel Macs are not supported because the Parakeet model runs on the Apple Neural Engine, which only Apple Silicon has.
Is it really free?
Yes. MacParakeet is free and open-source under the GPL-3.0 license. There is no paid tier and no 'community edition' with features stripped out. The full source code is on GitHub. No subscription, no credit card, no account required.
Why is it open-source?
Trust and ethics. Open-source means you can verify our privacy claims by reading the code -- not just taking our word for it. And since MacParakeet is built with AI tools trained on the open-source community's work, open-sourcing it back feels like the right thing to do.
What languages are supported?
98 in total. Parakeet — the default engine — handles 25 European languages including English, Spanish, French, German, Italian, Portuguese, and Russian. For Korean, Japanese, Chinese, Hindi, Arabic, Vietnamese, Thai, and 73 more, switch to the optional WhisperKit engine in Settings. One-time download, then everything still runs on your Mac. Clean text processing is currently optimized for English.
Can I record meetings? How is it different from Granola or Otter?
Yes. MacParakeet records system audio (everyone else on the call) and your microphone at the same time, then transcribes locally. While you record you get live notes, a live transcript, and an Ask tab to query the meeting as it unfolds. The difference is core speech recognition runs locally on your Mac; optional AI summaries or Ask only send transcript text.
Which local speech engine should I use?
Parakeet is the everyday default for fast local dictation and transcription in its supported languages. WhisperKit is the broad-language option for Korean, Japanese, Chinese, Hindi, Arabic, Vietnamese, Thai, and more. Cohere Transcribe is an optional local batch engine for plain-text file transcription when you want to compare another model. Cohere is not the live dictation-preview path.
Ready to reclaim your voice?
Free and open-source. Download the app or build it from source.
macOS 14.2+ · Apple Silicon · GPL-3.0