Everything Dictera does.
Two halves of one app. Your voice into text - dictation, translation, correction and voice-instructed editing, anywhere you can type. And an AI Assistant that answers on live calls, writes up your meetings without a bot, and chats privately. All on your Mac.
Dictation, exactly your way
Dictera lives in your menu bar and types wherever your cursor is. Every part of the flow - how you start, how you stop, how the text lands - is yours to shape.
Types into any app
Email, docs, chat, terminals, IDEs - the transcription is inserted right at the cursor, no copy-paste.
Flexible shortcuts
Hold a modifier push-to-talk style, double-tap it, or record any key combo. Left, right, or either side of the key.
Paste or direct typing
Insert instantly via paste, or type character-by-character for apps that need real keystrokes.
Stop on silence
Optional voice-activity detection ends the recording when you stop talking - thresholds and timeouts are tunable.
Esc to cancel
Press esc to stop at any time - and choose whether the captured audio is still transcribed or discarded.
Pick your microphone
Use the system default or pin a specific input device, with a live input-level meter to test it.
Recording HUD
A minimal floating indicator shows recording state and level, then gets out of your way.
Sound themes
Subtle start/stop chimes in selectable themes - or silence, if you prefer.
Clipboard option
Optionally keep the transcribed text on the clipboard after it's inserted.
Instant readiness
Keep the model loaded in memory so your first dictation of the day starts without a cold start.
Quick Controls
One shortcut opens a compact panel in the middle of the screen for the things you change mid-flow - switch model, change language, toggle cleanup - without opening Settings.
Languages & speech models
Choose from four on-device recognition engines and let Dictera detect the language as you speak - even two languages in the same sentence.
Automatic language detection
Speak and Dictera figures out the language - or pin one as a hint for extra accuracy.
Mixed-language dictation
Switch languages mid-sentence; each word is recognized and written in the correct script.
Apple Speech Analyzer
The macOS system engine: realtime streaming, low power, always up to date with the OS.
Nvidia Parakeet TDT v3
High-throughput recognition across 25 European languages, running entirely on your Mac.
Nvidia Nemotron 3.5 Streaming
Newest-generation streaming recognition with automatic detection across 100+ languages.
Alibaba SenseVoice Small
Fast, lightweight recognition with strong Chinese, Cantonese, English, Japanese, and Korean support.
In-app model manager
Download, switch, and delete models from Settings - sizes and language coverage shown up front.
Localized interface
The app itself is available in 11 languages, including German, Spanish, French, Japanese, Ukrainian, and Chinese.
Writing tools
Dictation is just the start. Dictera polishes what you say - and rewrites, fixes, or translates what you've already written, in any app.
AI cleanup
An on-device Gemma 4 model removes filler words and fixes punctuation before the text is inserted - across 140+ languages, including ones Apple Intelligence doesn't cover. If cleanup ever fails, you still get the raw transcript.
Translate as you speak
A dedicated shortcut records your voice and inserts a clean translation into the target language - 29 to choose from.
Edit Selection by voice
Select text, hold a key, and say the change - "make it shorter", "turn into bullet points", "draft a reply". The rewrite lands in place, one undo away.
Correct Selection
Select text anywhere, press a shortcut, and grammar, spelling, and punctuation are fixed in place - one undoable edit.
Translate Selection
A second shortcut shows the selected text translated into your language in a floating panel - the document is never touched.
Voice commands
Say "new line", "new paragraph", "comma", "question mark", or "scratch that" to structure text hands-free.
Custom vocabulary
Define word and phrase replacements so names, brands, and jargon always come out exactly right.
The AI Assistant
One floating panel, three modes. Assist suggests what to say while you're on a call, Meeting Notes writes the call up afterwards, and Chat is a typed conversation with the model. Press ⌘⌥H and it's there.
Hears both sides of the call
Your microphone and the other participants, captured through macOS itself - no bot joins the meeting, and the call keeps playing normally.
No Screen Recording permission
The other side is captured with a Core Audio process tap, so meetings need only microphone and audio-recording access.
One click mutes you
Assist and Meeting Notes start with your microphone on, so your own voice is in the call from the first word. Turn Mic off whenever you'd rather not be in the transcript; the panel always shows which state you're in.
Cancel speaker echo
On loudspeakers the other side would otherwise be transcribed twice - once from the call, once as room echo in your mic. Dictera pauses the mic while they're actually audible. On by default; with headphones it stays out of the way.
Invisible to screen sharing
The panel is excluded from screen capture by default, and it's now a one-click toggle in the panel itself rather than a trip to Settings.
Resize it, or take a column away
Drag any edge, or the dividers between the session list, the conversation and the result. Hide the conversation (⌥⌘T) and a long write-up gets the whole window. The size and the spot you left it in come back next launch; Reset panel layout undoes all of it.
Compact mode
A smaller panel for when it has to share the screen with a call - plus adjustable opacity and answer text you can size up or down. It keeps its own size, so a panel widened to read minutes doesn't follow you onto the next call.
Answer and Listen from anywhere
⌥⌘A answers the last question and ⌥⌘L starts or stops listening - from any app, so you never have to click into the panel on camera. Both are opt-in.
Every panel key is editable
18 panel actions grouped as Actions, Navigation and Display, with Restore Panel Defaults - so nothing clashes with your meeting app.
A model and a language per mode
Run a fast local model for live suggestions and a larger one for the end-of-meeting write-up. Each mode transcribes and answers in one language you pick.
Searchable session list
A sidebar of your conversations for the current mode, searchable and renameable in place, each one titled automatically. Search marks what it matched - in the list, the transcripts, the answers and your templates - the way Finder matches, so "resume" finds "résumé". Reopening a saved session never stops a recording that's still running.
Edit any result by hand
Fix a misheard name or drop a line instead of regenerating the nine-tenths that were right. The model's own version is kept, so Revert works in a session reopened next week, and an export says when a human changed it.
Recurring series
File a session under the meeting it repeats - Daily, Refinement, Customer interviews - and a year of them stays findable as one set. Filter the list by series, recolor them, and file or delete several sessions at once.
What a series still owes
Every open action item across every meeting in a series, in one list, read back from notes you already have - no model call, so it's instant and can't invent a commitment. It says so when some meetings couldn't be read.
Ask Dictera
Double-tap right ⇧ for a single field in the middle of the screen. Type or dictate a question, press Return, and the answer is waiting in the Chat tab as a new session - while a call the panel is recording carries on.
Two tabs, one call
Start Meeting Notes while Assist is already listening and the same call is transcribed into both, each in its own session.
Ask about what's on screen
Capture a whole display or just the window in front - always excluding Dictera's own windows - and send it with your next question. One window spends the same pixel budget on a tenth of the area, which is the difference between readable text and grey texture. Zoom, pan and export from the built-in viewer. Needs a model that accepts images.
Stays on your Mac
With the built-in model, transcripts, answers and screenshots never leave the machine. Point a mode at a cloud model and its text goes to that provider instead - the app tells you before you switch.
Assist - live answers on a call
Dictera transcribes both sides as you talk and drafts an answer when a question comes your way. For interviews, support calls, standups, and any conversation where you'd rather not be caught searching.
Answers without being asked
When the other side asks something, a suggested answer appears on its own. Or press Answer whenever you want one.
Answer one specific line
Hover any line in the transcript and answer exactly that - either speaker, however far back.
One-press rewrites
An answer can be made Shorter or Simpler, or given More detail or an Example - one click instead of retyping the question.
Follow up in writing
Type a follow-up and it extends the same thread. Browse back through every answer the session produced.
Session templates
Save reusable setups - the scenario, who you are, background context - for interviews, sales calls, or standups.
Answers you can read fast
Markdown, sized to taste, in a panel built to be glanced at rather than read.
Meeting Notes - the write-up, without a bot
Same capture, different result: press Stop and Dictera writes the meeting up. Nothing joins your call, no account exists, and with the built-in model nothing is uploaded - so you can take notes in the meetings other tools aren't allowed in.
Notes when you press Stop
Summary, decisions, action items and open questions - written automatically at the end, or any time from the summarize button.
Four levels of detail
The same session can also be produced as Full minutes (the whole meeting in order, about a third the length), a cleaned-up Transcript in the language it was spoken, or Chapters - a timeline of what was discussed when. Choose which one this meeting is for, and both the Meeting tab's button and Stop produce it - one answer to "what do I get from this meeting" rather than two. The other three stay one click away on the button's menu.
Notes take the shape of the session
A lecture has no decisions and no owners; a meeting does. Left on automatic, each recording is read for what it is - meeting, lecture or talk, one-on-one, interview, sales call - and the notes say which shape they used, so a wrong guess is one click away from fixed.
Action items with owners
Name the participants in Meeting Setup and an action item can say who owns it. Dictera never guesses: where the meeting didn't establish who, it writes "Unassigned".
Meeting Setup, per meeting
Who else is here, who you are, and any background - filled in from the calendar invite when you've connected one. New meetings start from your defaults, so the standup list is typed once; from the first edit the setup belongs to that session, so last month's town hall still has its forty people when you reopen it. The background field also fixes product and people names the transcriber mangles.
Long meetings survive the context window
A transcript longer than the model's limit is read in passes; the panel says which pass it's on, and the result says plainly when it covers only the most recent stretch. A model with a larger window covers more in one go.
Reshape the notes in one press
Cut them down to Action items, reorganize By topic, or turn them into a Recap email - a finished deliverable, not a draft. Lectures also offer Key terms and a Study guide.
Ask about the meeting afterwards
Type a question and Dictera searches the whole transcript for the passage that answers it - and says so when the transcript doesn't settle it, rather than inventing something.
Timestamped transcript
Every line carries the time it was said, so minutes and chapters can point back at the moment.
Save as Markdown
Notes, minutes, transcript and chapters all export to a Markdown file - with the date, participants and the model that wrote them - plus copy buttons for each.
Calendar
Connect your calendar and a recording knows which meeting it belongs to and who was invited - and can start itself as the meeting begins. Read from the accounts already set up in Calendar, read-only, and off until you turn it on.
No second sign-in
iCloud, Google, Outlook and Exchange calendars are read through macOS itself - there's no connector to authorize and no account to create. Dictera never writes an event.
A countdown before the meeting
About a minute before each meeting, a countdown asks whether to record it. Left unanswered it starts recording - a meeting you didn't record is gone, one you didn't want is a click to delete. Set silence to mean "no" instead, or turn it off, globally or per series.
Nothing starts unseen
The countdown is the consent surface: if it can't be shown, nothing records. Skip the same meeting twice and Dictera stops asking about that one; accept three times and it offers to make the series automatic.
Attendees fill Meeting Setup
The names on the invite become the participant list action items are attributed against. Email addresses are never stored at all, and one switch keeps calendar names on local models only.
And a countdown when it ends
A recording that outlives its meeting gets a minute to say Keep recording; otherwise it stops, and writes the meeting up on the way out. A recording with nothing scheduled behind it is asked about after an hour, two or three - your choice, and it works whether or not you connected a calendar.
The day's agenda in the panel
The session list can show today's meetings - recorded, still coming, or missed - with a month grid for reaching any other day.
Series filing stays yours
Dictera never labels a session by itself - which event this is and which series you file it under are different questions, and a label that appeared on its own is one you'd have to notice before you could undo it. File from the sticky Recurring series field, the row's label picker, or the model's suggestion chip.
Chat - a private model, one shortcut away
The third mode captures no audio at all. It's a typed conversation with whichever model you've pointed it at - including the one running on your own Mac. No browser tab, no account.
Nothing is listening
Chat never opens the microphone or the system-audio tap.
Follows the language you type in
Unlike the listening modes, Chat answers in whatever language you write.
Its own engine
Point Chat at the built-in on-device model, a local server, or a cloud provider - independently of the other two modes.
Pictures work here too
Capture a display or the front window, or paste an image you already have with ⌘V, and ask about what's in it - same viewer, same rewrites. A pasted image needs no Screen Recording permission.
Bring your own AI
The built-in on-device model powers everything out of the box. Prefer your own? Point each feature at a local server, a cloud provider, or the Claude Code CLI - ready-made presets for all of them.
Built-in stays the default
No setup required. Cleanup, translation and the assistant all run on Gemma 4 E2B (2.6 GB) out of the box, or E4B (4.2 GB) if you want noticeably better translation and instruction-following. Both cover 140+ languages.
Six local servers, preset
Ollama, LM Studio, llama.cpp server, Jan, vLLM and LocalAI - pick one, pick a model, go. Nothing leaves your machine or your network.
Seven cloud providers, preset
OpenAI, OpenRouter, Groq, Mistral, DeepSeek, Together AI and xAI. Each is flagged as needing an API key, and the app warns you that these send text off your Mac.
Any OpenAI-compatible server
Not on the list? Add a custom endpoint and model name; Dictera talks the standard API.
Claude Code CLI as an engine
The Claude Code command-line tool already installed on your Mac - not a Claude model over an API. Point the assistant at a project folder and it reads the code before answering, so it can talk about your actual repository during a standup or an interview. Choose the model and how much effort it spends.
Read-only, by design
There is deliberately no "allow edits" option. The assistant is driven by what it hears on a call - untrusted input - so it can read and search your folder, never change it.
Connector tools you tick one by one
MCP connectors are denied by default. A Jira connector exposes "create ticket" right next to "search issues", so nothing is reachable until you allow it individually - and tools that can change things are marked.
Per-feature engines
Cleanup, translation, Assist, Meeting Notes and Chat each choose their own engine independently - a fast local model for live work, a bigger one for the write-up.
Keys stay in your Keychain
API keys are stored in the macOS Keychain, never in plain settings files.
Test before you rely on it
A connection test verifies the endpoint and model respond - and fills in the context window when the server reports it, which is what lets a long meeting be summarized in as few passes as possible.
History & statistics
Everything you dictate is yours to search, export, and measure - stored locally, on your terms.
Searchable history
Full-text search across past transcriptions, with pinning for the ones you reuse.
Export & retention
Export your history, choose how long entries are kept, or clear everything with one click.
Usage statistics
Words dictated, sessions, dictation time, and an estimate of the typing time you saved.
Weekly report
An optional notification at the start of each week summarizing last week's dictation.
Private by default
History and stats live only on your Mac - and both can be switched off entirely.
Settings search
A System Settings-style sidebar with search that finds the exact toggle you're after.
Privacy & platform
Dictera is a native Mac app, built for Apple Silicon - and designed so your voice never leaves your machine.
100% on-device
Speech recognition, cleanup, and translation run locally. No cloud, no accounts, no API keys required.
Works offline
Once a model is downloaded, everything works without an internet connection.
Native menu-bar app
Lightweight and always ready - no Dock icon, no browser tab, no Electron.
Guided onboarding
A friendly first-run flow walks you through the model download, permissions, and your first dictation.
Launch at login
Optionally start with your Mac so the shortcut is always live.
Automatic updates
Signed, notarized releases delivered seamlessly via Sparkle.
One license, 3 Macs
Activate your subscription on up to three Macs, and move seats whenever you need.
14-day free trial
Every feature unlocked from day one - no card, no account.
Try everything, free for 14 days.
Designed for Apple Silicon · macOS 26 or later