Dictato: Speech to Text & Dictation App for Mac

Dictato lives in your menu bar and is always ready. You configure a global hotkey - press and hold it, speak, then release. Your speech is transcribed in real time and the resulting text is instantly inserted wherever your cursor is, in any application. There's no need to copy and paste, switch windows, or interact with a separate interface. It works in your email client, word processor, code editor, messaging apps, browser fields - anywhere you can type on macOS. See our guide to dictating into any app for examples.

Under the hood, Dictato uses on-device machine learning models to convert your speech to text. The transcription engine runs entirely on your Mac's hardware, leveraging Apple Silicon's Neural Engine for fast inference. The audio from your microphone is processed locally in real time with latency as low as 80 milliseconds, meaning you see your words appear almost as fast as you say them. Because nothing is sent to a server, transcription works even when you're offline - on a plane, in a cafe without Wi-Fi, or anywhere else. Learn more about how 80ms dictation works.

Mac's built-in dictation has several limitations that Dictato solves. First, Apple's dictation stops listening when you pause - even a few seconds of silence to collect your thoughts can end the session, forcing you to reactivate it. Dictato never cuts out. Second, Apple's dictation sends your audio to their servers for processing (unless you enable on-device mode, which has limited language support). Dictato is 100% on-device, always.

Third, Dictato offers real-time transcription with 80ms latency - you see words appear as you speak, not after you finish. Fourth, you get to choose from 4 transcription engines (Whisper, Parakeet, Qwen3, Apple) depending on your language and accuracy needs. Finally, Dictato works in every app, not just Apple apps, and inserts text directly at your cursor without needing a special dictation field.

Dictato ships with four transcription engines, each with different strengths. Parakeet is the recommended engine - it's NVIDIA's open-source ASR model optimized for speed and accuracy, supporting 25 European languages with a model size of approximately 2.3 GB. It delivers the best balance of quality and real-time performance on Apple Silicon.

Whisper is OpenAI's well-known speech recognition model, supporting 99 languages with a smaller model size of around 600 MB. It's the best choice if you need broad multilingual coverage. Qwen3 is a compact multilingual engine (around 600 MB, 30 languages) that runs on macOS 15 and later. Apple SpeechAnalyzer is Apple's built-in engine available on macOS 26 and later - it requires no model download since it uses the system's speech recognition framework. See our engine comparison for benchmarks.

All engines run 100% locally on your Mac. You can switch between engines at any time in settings.

Yes. Dictato works with virtually any macOS application where you can type. This includes Gmail, Outlook, Apple Mail, Slack, Discord, Microsoft Teams, VS Code, Xcode, Sublime Text, Notion, Obsidian, Bear, Pages, Word, Google Docs (in Chrome/Safari), Twitter/X, LinkedIn, and countless others.

When you release your hotkey, Dictato inserts the transcribed text directly at your cursor position using macOS accessibility features. There's no need for the app to have special dictation support - if you can paste text into it, Dictato works with it.

Dictato lives in your menu bar at the top of the screen, not the Dock. Click its icon to start recording, or to open History and Settings. For a tour of every screen, see the user manual.

Accessibility is what lets Dictato type the transcribed text into other apps at your cursor. Transcription works without it, but the text can't be inserted. You grant it once in System Settings, and the setup assistant walks you through it. Dictato also needs Microphone access to record, and optional Automation access only if you use per-website rules. See the permissions section of the manual.

Auto-correct (proofread) cleans your raw dictation - grammar, punctuation, capitalization, and filler words - while keeping your wording. A tone goes further and rewrites the same thought into a style, such as a polished email, a concise note, or a bullet list. Use auto-correct for tidy transcripts, and tones when you want the text reshaped. Both run on-device. The manual lists every built-in tone.

No - dictation itself works without it. Apple Intelligence powers the optional AI steps: auto-correct, auto-translate, and AI tone rewrites. If your Mac doesn't support Apple Intelligence, you can still dictate and use the deterministic cleanup, voice commands, and vocabulary features. Tone rewrites can also run on Dictato's bundled on-device model, Gemma.

Add them to the Vocabulary section in Settings - the Context field lets you list names, acronyms, and technical terms so they come out spelled correctly. You can also create a Voice Command that maps a spoken phrase to an exact replacement (or an emoji). The manual covers both.

Yes. Profiles let you give each app - and each website in supported browsers - its own behavior: a casual tone in Slack, an email tone in Gmail, plain proofreading in your editor, or verbatim text where you need it untouched. Each rule can even pin its own transcription and rewrite engine. See how profiles work.

Dictato is built from the ground up with privacy as a core principle. All speech recognition happens entirely on your Mac using locally downloaded machine learning models. Your audio never leaves your device - there is no server, no cloud API, no intermediary processing. When you speak into Dictato, the audio goes directly from your microphone to the on-device transcription engine and is discarded immediately after processing.

This makes it suitable for sensitive use cases: legal professionals dictating confidential notes, medical practitioners documenting patient information, journalists protecting sources, or anyone who simply values their privacy. There are no analytics on your transcription content, and no way for anyone - including us - to access what you dictate. Read our speech-to-text privacy guide for details.

Language support depends on the transcription engine you choose. The Whisper engine offers the broadest coverage with 99 languages, including English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Ukrainian, Turkish, Arabic, Hebrew, Hindi, Japanese, Chinese, Korean, Vietnamese, Thai, Indonesian, and many more.

The Parakeet engine focuses on 25 European languages with optimized accuracy. All engines feature automatic language detection, so you don't need to manually select your language before speaking. See our multilingual voice typing guide for setup tips.

Dictato requires macOS Sonoma (macOS 14) or later, running on a Mac with Apple Silicon (M1 chip or newer). The transcription engines rely on the Neural Engine built into Apple Silicon chips for fast, efficient inference - this is what enables the 80ms real-time latency.

In terms of storage, you'll need space for the transcription engine models. Parakeet requires approximately 2.3 GB, Whisper approximately 600 MB, Qwen3 approximately 600 MB, and Apple's built-in engine requires no additional download. The app itself uses between 100-300 MB of RAM depending on the active engine.

Yes. Dictato includes a 7-day free trial with full functionality - no credit card required. Download the app, use it for a week, and only pay if you want to keep using it after the trial ends.

This gives you plenty of time to test all four transcription engines, try it in your favorite apps, and make sure it fits your workflow before committing.

Yes. Because Dictato processes all audio 100% on-device with no cloud connection, it's suitable for use cases requiring strict confidentiality. No audio or transcription data ever leaves your Mac, making it appropriate for HIPAA-conscious medical documentation, attorney-client privileged notes, or any other sensitive professional use.

That said, like any transcription tool, Dictato is not 100% accurate. We recommend proofreading important documents before finalizing them.

The best dictation software for Mac depends on your priorities. If privacy and offline use matter most, Dictato is the top choice - it runs 100% on-device with four transcription engines (Whisper, Parakeet, Qwen3, Apple), works in any app, and has no timeout. Apple's built-in dictation is free but limited: it times out after pauses, has fewer language options, and sends audio to the cloud by default.

Dragon was long considered the industry standard, but Nuance discontinued Dragon Professional for Mac. Cloud-based options like Otter.ai offer features like speaker identification but require an internet connection and a subscription. For a detailed comparison, see our roundup of the best dictation apps for Mac in 2026.

Yes. Dictato is a real-time transcription app for macOS that converts your speech to text as you speak, with latency as low as 80 milliseconds. Unlike batch transcription tools that process pre-recorded audio files, Dictato is designed for live dictation - you press a hotkey, speak, and the transcribed text is inserted directly at your cursor in any application.

It supports four on-device transcription engines: Whisper (99 languages), Parakeet (25 languages, highest accuracy), Qwen3 (30 languages), and Apple SpeechAnalyzer. All processing happens locally on your Mac with no cloud dependency, making it suitable for confidential transcription work. Learn more about how the 80ms transcription pipeline works.

No. Dictato is 19.99€ once, a lifetime license. No subscription, no monthly fees, no renewal. You pay one time and the app is yours to keep, with no limit on how much you dictate.

Updates are included. And because every transcription engine runs on your Mac, there is no server and no API key to pay for. After the one-time purchase, there is nothing left to pay.