What's included in the free version of Dictum?

The entire app and local dictation (Parakeet, offline, 25 languages): AI processing, snippets, history, screenshots. No card required. You only pay for AI tokens (optional).

Is my voice data safe?

Yes. The local Parakeet model runs on your device — audio never leaves your Mac. No accounts, no telemetry.

Does Dictum work in noisy environments?

Yes. Dictum uses the MacBook's directional microphone and filters out background noise. In open spaces we recommend AirPods — their mic focuses on your voice.

Will AI change my writing style?

AI processing is optional. By default it only removes filler words. Your style is preserved. You can create your own prompts for fine-tuning.

What if Dictum shuts down?

Local mode will keep working forever. Your entire history exports to JSON/Markdown. We never hold your data hostage.

The keyboard is no longer needed

Download for Mac Download for Windows (beta)

Integrations

Works everywhere
you type

Slack, Cursor, Notion, terminals… You speak — Dictum pastes the finished text.

Translate into any language

100+

translation languages

English, Chinese, Spanish and any other — in one key.

1 click

and done

Translate right in the Dictum window — no copying into third-party services.

built-in translators

RU, EN, CH — out of the box. Plus your own prompts for any language and style.

Coding & prompts

Prompt by voice — many times faster

You don't write code by voice — you command the agent by voice. Say the task in words, and Dictum delivers clean text into Cursor, Claude Code or the terminal.

Prompt, don't type

Dictate the task in plain words. Dictum turns it into a clean prompt — the agent writes the code, you just steer.

Voice commands to the agent

"Rewrite it with Zustand," "fix npm run dev," "add tests" — refactors, fixes and new features without pressing a single key.

Understands technical language

"API," "async/await," "useState," "refactoring" are recognized correctly — not as "ay-pee-eye" or "use-state."

Works in your tools

Pasting via CGEvent — into any editor, terminal and AI agent. Not just the browser.

From idea to app

Describe out loud what you want built — the agent does it. You stay focused on the product, not on typing.

Work faster.
Phrase it clearer.

Save hours of typing every week — and express your thoughts more clearly in every message and note.

4h 56m

less typing every week

47s

average reply time

Works with every app

Fits into every app you use every day — no setup, no friction.

You set the pace

Keeps up with your ideas — focus on the thought, not on typing.

Sees your screen — via screenshot

An annotated screenshot — and the AI agent sees the context: from terminal code to the interface.

Not just dictation

A whole workflow
in two keys

01 · Modal window

A text buffer

Accumulate text, edit it, run it through AI — then paste the finished result all at once. Indispensable when working in terminals.

Dictum ProAuto

Build a coffee-shop landing page in React and Tailwind, dark theme.saved · 2 min ago

Refactor the auth module and add unit tests.saved · 9 min ago

Reply to Anna: thanks, I'll send the report by Friday.saved · 24 min ago

Meeting notes: ship v2, freeze scope, QA on Thursday.saved · 1 h ago

Translate the onboarding email into Spanish.saved · 2 h ago

Email the director: let's move the meeting to next week — I'm away on a business trip.drafting now ▍

AI ⌘1Snippet ⌘2Send ⏎

02 · AI processing

One key — any operation

Rewrite, translate, change tone, summarize. Dictate a draft the way you think — get a clean version with a single key.

Dictum ProAuto

Email the director: let's move the meeting to next week.

AI ⌘1Apply ⏎

03 · Screenshots

Smart screenshots for terminals

Take a screenshot, annotate it, and send it straight to any terminal AI — Claude Code, Codex, Cursor. The agent gets the image and instantly understands the context.

Screenshot → terminal

Screenshot sent to Claude Code in the terminal

link copied

Works with terminal AI agents:

Claude Code

Codex

Cursor

Gemini CLI

Copilot

04 · Productivity

Snippets, commands, history

Hotkey templates, quick text actions and search through your recent entries. Everything you need, right at hand.

Search history…

Email the director: let's move the meeting…Just now

Build a coffee shop landing page in React and Tailwind3h

Rewrite the paragraph to be politer and shorter4h

Translate the email to English1d

History⌘4↵select↑↓navigateescclose

Press ⌥option and speak.

Let your voice do everything.

AI processing

Said it awkwardly?
AI will fix it.

Dictate the way you think — rambling, rough, even with swearing. One key turns it into clean, professional text. Or translates it into any language.

Privacy

Your data — yours alone.

The local model runs right on your Mac. No accounts, no telemetry, no cloud by default.

Local

Parakeet recognition right on your Mac — audio never leaves the computer

Offline

works without internet: on the subway, on a plane, under NDA

accounts, tracking, ads, or data selling

Yours

AI through your own API keys — you only pay for what you use

Comparison

Dictum vs the rest

	Dictum	Superwhisper	Wispr Flow	Mac Dictation
Local dictation (offline)	✓	✓	✗	✓
AI text processing	✓	✗	✓	✗
Custom AI prompts	✓	✗	~	✗
Annotated screenshots	✓	✗	✗	✗
Push-to-talk (one key)	✓	✓	✓	✗
Text buffer window	✓	✗	✗	✗

~ — partial. Competitor features may change — check their sites.

Frequently asked questions

The entire app and local dictation (Parakeet, 25 languages, offline): AI processing, annotated screenshots, snippets, history. No card, no subscription — you only pay for AI tokens as you use them.

No. Local dictation (Parakeet) works fully offline. The internet is only needed for optional AI text processing.

Yes. The local Parakeet model runs right on your Mac — audio never leaves the device. No accounts, no telemetry. AI processing runs through your own API keys.

The app and local dictation are free. You only pay for AI processing with your own tokens (AI credits, ~$0.05 per operation), as you use them. No ads, no data selling.

Separate packs for advanced AI features: summarization, rewriting, translation. Basic dictation works without them. Credits never expire — they're used until they run out.

macOS 14+ (Sonoma) and Apple Silicon (M1–M4) — for the local Parakeet model. There's also a separate app for Windows.

Yes. Recognition and the ✨ Enhancer normalize technical jargon: "API," "async/await," "useState," "refactoring" are rendered correctly — not as "ay-pee-eye."

Dictum adds what the standard dictation lacks: AI text processing, custom prompts, annotated screenshots, push-to-talk on Right Option, and smooth terminal work through the modal buffer window.

Yes. Local mode (Parakeet) is 100% on-device. No data is sent anywhere. Suitable for any project under NDA.

A free app, you only pay for AI tokens. Annotated screenshots, custom AI prompts and the modal buffer window — things competitors don't have. Local and private.

Yes. It uses the MacBook's directional microphone with background-noise filtering. In open spaces, AirPods help — their mic focuses on your voice.

Minimal. Dictum only runs while you're actively dictating. Local mode uses the Neural Engine — a dedicated chip for ML tasks that doesn't load the main processor. About 1–2% battery per hour of active dictation.

AI processing is optional and configurable. By default it only removes filler words ("um," "like," "you know"). Your style is preserved. You can create your own prompts for fine-tuning — "keep the conversational style" or "change nothing but typos."

Local mode (Parakeet) will keep working forever — it needs no server. Your entire history exports to JSON/Markdown. We never hold your data hostage.