§ Dictation core
Dictation
How dictation works end-to-end: how to bind hotkeys, the difference between Raw, Polish, and Enhanced, how text gets inserted into the focused app, and how Dettivo learns the words you actually say.
Context-aware
Context-aware dictation
Enhanced mode attaches three optional signals to every run: the frontmost app and its bundle ID, the active selection (via Accessibility), and a single window screenshot (only when you turn it on for that app class). Bounded payload, never silent.
Opt-inUse case · Cursor
How to dictate in Cursor with Dettivo
Hold Push-to-Talk inside Cursor — editor, agent chat, or anywhere else. Enhanced recognizes Cursor as a code app and runs the code-aware Polish profile. The Cursor agent prompt panel accepts clean dictated text.
Use case · Google Docs
How to dictate in Google Docs with Dettivo
Click into a Google Docs document, hold Push-to-Talk, speak. Direct insertion works in classic rendering; Google Docs canvas-rendering mode uses Dettivo's clipboard fallback.
Use case · JetBrains
How to dictate in JetBrains IDEs with Dettivo
Hold Push-to-Talk inside any JetBrains IDE — IntelliJ, GoLand, RubyMine, PyCharm, WebStorm — and speak. Enhanced recognizes JetBrains apps as a code class. Direct insertion works through Accessibility.
Use case · Mail
How to dictate in Apple Mail with Dettivo
Click into a Mail compose window, hold Push-to-Talk, speak. Dettivo runs STT, applies your Mail Polish profile (Formal preset by default), and types into the focused message body.
Use case · Notion
How to dictate in Notion with Dettivo
Click into any Notion block, hold Push-to-Talk, speak. Dettivo inserts directly into the focused block. Markdown shortcuts you would normally type ("# heading", "- bullet") still work because Polish does not strip them.
Use case · Slack
How to dictate in Slack with Dettivo
Hold Push-to-Talk inside a Slack message box and speak. Dettivo runs STT locally, applies your Slack-tuned Polish profile, and types into the focused composer. Threads, DMs, and channels all behave the same way.
Use case · VS Code
How to dictate in VS Code with Dettivo
Click into any VS Code text surface, hold Push-to-Talk, speak. Enhanced mode recognizes VS Code as a code app and applies code-aware Polish — fewer conversational rewrites, more code-fence preservation, path/flag passes intact.
Use case · Zed
How to dictate in Zed with Dettivo
Hold Push-to-Talk inside Zed and speak. Zed is a native macOS app — Accessibility-based direct insertion is fast and reliable. Code-aware Polish preserves backticks, paths, and code-fence spans.
Dictation
How dictation works
Hold the Push-to-Talk hotkey (Option+Space by default), speak, release. Dettivo captures audio, runs speech-to-text locally on Apple Silicon, optionally polishes the result, and types into the focused app.
Local-onlyDictation modes
Raw, Polish, Enhanced — pick the right mode
Raw inserts the speech-to-text output verbatim. Polish adds deterministic dictionary + replacement + snippet processing — no model, no network. Enhanced adds context-aware model rewriting on top, optionally translates, and can run automations.
Dictionary
Custom dictionary
A per-language vocabulary list of names, places, projects, and domain terms. Dictionary entries help speech-to-text spell things correctly and ground Polish-stage substitutions. Distinct from replacement rules and snippets.
Enhanced mode
Enhanced mode
Enhanced runs Polish, then attaches the focused app and (opt-in) the active selection and a window screenshot, then routes the transcript through a model rewrite. The local MLX polish model is the default; OpenAI / Anthropic / OpenAI-compatible endpoints are opt-in BYOK.
Opt-inHistory
History + retention
Every dictation and every meeting is stored locally with raw + polished transcript, metadata, and (toggleable) audio. Full-text search across both. Per-item delete, retention windows, and a hard delete-all.
Hotkeys
Hotkey reference
Dettivo binds eight default global hotkeys plus per-language profile hotkeys. All are configurable in Settings › Hotkeys. Global hotkeys require Input Monitoring permission.
Requires permissionImport audio
Import audio files
Drop existing m4a / mp3 / wav files into Dettivo from the menu bar or via the CLI. The file is stored in History and transcribed with your chosen STT engine. Raw + polished transcript produced.
Direct insertion
Direct insertion
Dettivo types the polished transcript into whichever app has focus, using macOS Accessibility. Secure fields are detected and refused. Clipboard fallback runs when insertion is blocked, with a five-second undo window.
Requires permissionLanguages
Languages + profiles
Pick a default language in Settings › Languages. Add per-language profiles to bind their own hotkey, vocabulary list, and model override. Auto language detection is available on supported STT engines.
Migrate from MacWhisper
Migrating from MacWhisper
MacWhisper transcribes files. Dettivo transcribes live dictation and meeting capture in addition. Your existing audio files can be imported into History and transcribed with the engine of your choice.
Use case · Obsidian
Obsidian voice notes with Dettivo
Point Dettivo Quick Note at your Obsidian vault folder. Pick a profile — Standalone, Inbox, or Daily — and press the Capture Note hotkey anywhere on your Mac. The note lands directly in your vault as markdown.
Polish mode
Polish mode
Polish takes the Raw transcript and runs a deterministic pipeline on it: dictionary substitutions, replacement rules, snippet expansions, punctuation normalization, code/path/flag post-processing, and insertion formatting. No model call, no network, instant.
Local-onlyQuick Access
Quick Access
A keyboard-driven popup of recent transcripts. ⌘⇧/ to open (default), arrow keys to navigate, Return to insert the polished version, ⌘+Return to insert raw, Escape to close.
Quick Note
Quick Note
A separate Capture Note hotkey writes dictation to a folder you pick — your Obsidian vault, a second-brain folder, a daily note, your repo's docs/notes/ tree. Markdown output, three save profiles, optional front matter.
Raw mode
Raw mode
Raw inserts what speech-to-text produced — nothing more. No dictionary, no replacements, no model rewrite, no network. The shortest path between speaking and seeing text.
Replacements
Replacement rules
Deterministic source → target mappings applied during Polish. "Quote unquote" → empty. "Dot com" → ".com". Imported via `heard => wanted` lines, `heard -> wanted` lines, TSV, or CSV. Priority controls the order.
Snippets
Snippets
Trigger-to-expansion rules. Say a short phrase, get a longer string inserted. Configured in Settings › General › Snippets. Distinct from dictionary terms (which help with spelling) and from macros (which can run commands).
Migrate from Superwhisper
Migrating from Superwhisper
Both apps are local-first dictation on Mac. The main upgrades in Dettivo are bundled meeting capture, the developer surface (CLI + REST + MCP), Quick Note to a folder, and explicit context-aware Enhanced mode. The migration is light — most muscle memory carries over.
Translation
Translation before insertion
In Settings › General, toggle Translate Output and pick a target language. Enhanced runs translation as the last step before insertion — the focused app receives the translated text, not a draft.
Opt-inMigrate from Wispr Flow
Migrating from Wispr Flow
Reproduce your Wispr Flow workflow in Dettivo with three settings: bind Push-to-Talk to the chord you already use, enable the dictionary for your project names, and (optionally) enable Enhanced mode with the bundled local model. No subscription, no cloud round-trip.
- Getting started — Install, activate, grant permissions, and run your first dictation.
- Meetings — Bot-free capture, speaker labels, meeting analysis, recap automation, Meeting Mirror.
- Automation — Dictation macros, app + shell allowlists, recap email automation, weekly digest.
- Developers — Same contract everywhere. Bundled CLI, loopback REST adapter, stdio MCP server.
- Privacy + security — Local-first defaults, BYOK providers, telemetry policy, permissions, retention.
- Account + license — Activation, device management, secure updates, recovery, refunds.
- Troubleshooting — Hotkeys, insertion, meeting audio, MCP connect, REST 401, license activation.