Questions and answers

macOS 14.4 or newer on a Mac with Apple silicon (M1 and later). Intel Macs are not supported, because the on-device model needs the Apple chip to be fast.

Download Captylo.dmg, open it and drag Captylo to the Applications folder. On the first launch Captylo asks for the microphone, and for Accessibility so it can paste the text where you type. The app is signed and notarized by Apple, so macOS opens it without the "unidentified developer" warning. Later versions come through the app: Captylo asks before it installs one.

macOS dictation types what it hears, word by word. Captylo lives under one key in every app, learns your names and jargon, keeps a history with recordings and shows your stats. In Pro it also polishes the text with AI and turns it into an email, a to-do list or English.

Yes. Polish is our first language: the interface, the dictionary and the AI modes all work in Polish. The on-device model also understands English and 23 other European languages.

Yes, anywhere you can type: Word, Gmail, Thunderbird, Slack, Notion, your browser, Visual Studio Code, the terminal and Claude Code. The text is pasted at the cursor and your clipboard is restored right after. The app asks for the Microphone and Accessibility permissions on first launch.

Yes. When you fix a word in text Captylo pasted, or spell a name out loud, it remembers the right version and uses it next time. Everything it learns stays on your Mac, you can undo any of it, and the whole feature has a switch in Settings.

Yes. Drag the file onto the File transcription window or pick it with Choose files. Captylo reads .ogg and .opus voice notes from messengers, even ones saved without a file extension, plus mp3, m4a, wav, flac, mp4 and mov. The text is made on your Mac, or in the cloud if you switch it on, with your dictionary applied, and lands in your History.

No. Captylo does not join the call, so it never shows up in the participant list. That is why every recording starts with a reminder to tell the others, with a ready line to paste into the chat. In many countries recording a call needs the consent of the people in it: ask before you start.

Any app you can hear on your Mac: Google Meet, Zoom, Microsoft Teams, Slack, FaceTime, Webex, Discord and more, in the browser too. Captylo records the microphone and the computer’s audio, so it needs no integration with any of them.

The live transcript comes from the speech model on your Mac: it works offline and for free, but it is not as accurate as cloud transcription. In Pro both tracks go to the cloud after the call and you get a more accurate transcript, which AI can then fix with your dictionary and the participants’ names.

No, but headphones give the cleanest transcript. Without them the others’ voices come back through your microphone, and Captylo removes those repeats from your track in the transcript.

Captylo can notice that a call has started and ask whether to record it. With your calendar connected it also reminds you before a meeting. You start recording with one click or the ⌃⌥⌘M shortcut. It never records without you saying so.

Speaker labels (Speaker 1, Speaker 2 and so on, which you rename to real names) are part of Pro and need macOS 15 or newer. On macOS 14 the other side has one label: “Others”. Your own track is always labelled “Me”.

On your Mac. Audio files go right away, after 7 or 30 days, or never; transcripts and notes stay. Something goes to the cloud only when you use a Pro feature, and only what that feature needs.

Yes. Notes, voice notes, importing recordings and search are part of the Free plan, on your Mac. Tidying a note up with an AI mode works in Free with your own key, and in Pro on Captylo AI, with no keys.

Press ⌃⌥⌘N in any app, say what you want to remember and press the shortcut again. Nothing is pasted: the recording and its transcript go to Notes, and the “Note saved” message opens it with one click. The menu bar does the same (“New voice note”), and you can turn the shortcut off in Settings.

Yes. Save the voice message as a file, then click “Import recording” in Notes or drag the file onto Notes. Captylo turns it into a note with its transcript and keeps the recording with it. If a recording cannot be transcribed, the note keeps it and offers “Try again”.

In a note, click “Tidy up with AI” and pick a mode: Cleanup, Organize thoughts, Task list or one of your own. AI rewrites the note, and “Restore original” brings your text back. In the Free plan it uses your own AI key, in Pro it runs on Captylo AI.

Not yet. Sync between devices is in the works and your data is ready for it, but today your notes live only on your Mac. Something goes to the cloud only when you use the cloud or AI yourself, as with dictation.

It is our own AI service in the Pro plan. It writes meeting notes, answers in Ask and tidies text up with AI modes, with no keys of your own to set up. Answers in Ask point to the moments of the call they come from.

Ask answers questions about one meeting, from its transcript. Ask all looks for the answer in all the meetings on your Mac and sends only the matching excerpts to AI, never whole transcripts. Both are part of Pro, and the answers carry time stamps such as [12:34] that open the transcript at that line.

In Captylo open Settings > Meetings, turn on “Access for AI assistants (MCP)” and click “Copy configuration”. Paste it into the MCP server configuration of your assistant (in Claude Desktop: Settings > Developer > Edit Config) and restart it. The MCP server runs on your Mac, read-only, in the Free plan too.

The meetings and notes on your Mac: the list of meetings, search results (with the notes they find), a meeting’s AI notes and full transcript, and since version 1.0.16 your notes too (the list_notes and get_note tools). The server is read-only, so the assistant cannot change or delete anything. Captylo sends nothing by itself, but the assistant passes what it reads to its own service, like the rest of your chat with it.

Not yet. A public API for automations is in the works. Until then the local MCP server reads your meetings, and you can export any meeting to Markdown and an open JSON format.

For dictation Pro adds cloud transcription, all AI modes and your own modes; for meetings a more accurate cloud transcript, AI notes, speaker labels and Ask. Plus every future feature. It costs $84 a year (about $7 a month) or $9 month to month. A new account gets 7 days of Pro for free, no card needed. There is only one paid plan, so there is nothing to compare.

You pay by card and Pro goes to the email address from the payment. In the app you sign in with that address, using a code we email you. Cloud transcription and AI then work with no keys of your own. You can cancel at any time from the app.

Yes, in the free plan too, for dictation. Paste your own key for cloud transcription or AI in Settings and those features work without a subscription; you pay the provider directly. The meeting AI is Pro only. Pro is for people who want it all to work right away, with no keys to set up.

No. Dictation on your Mac stays free with no word limit. You pay only for the cloud and AI, because those cost us money every time you use them.

Pro includes 20 hours of cloud transcription and 3 million AI tokens a month. For a meeting only the time someone speaks counts, on your track and the others’ separately, so silence and the other side heard through your speakers do not use the limit. When the cloud limit runs out, Captylo transcribes on your Mac, and when the AI limit runs out, the text is pasted without AI changes. The limits reset at the start of the month, and you see your use in Settings > Captylo account.

For dictation, yes, and it comes first: with your own key in Settings, Captylo talks to that provider directly, in Pro too, and that use does not count towards the plan’s limit. The meeting AI in Pro always runs on Captylo AI, our own AI service.

Yes. In Settings > Captylo account you pick yearly or monthly billing. You can also buy Pro on this site and then sign in to the app with the email address from the payment.

Only once, to download the speech model. After that dictation works fully offline. The internet is needed only for cloud transcription and AI modes, in Pro or with your own API key.

They stay on your Mac, next to your history. Nothing is uploaded unless you turn on a cloud feature yourself. You can delete any recording from the history at any time.

Yes. The app’s code is on GitHub under the GPLv3 licence: you can check what the app does with your voice, build it yourself and change whatever you like. The server behind Pro is a separate service.

See the code on GitHub

Yes. You need a Mac with Apple silicon and Xcode 26. The easiest way is to ask an AI agent in the Terminal to do it for you: the prompt is on the Open source page and in the README on GitHub. A copy you build yourself is signed only on your Mac, so macOS may ask for the permissions again after a rebuild.

No. The GPLv3 covers the code, not the Captylo name, logo or app icon. If you publish your own version, give it its own name and icon.

No. The app is open: dictation, meetings on your Mac, AI modes and the MCP server. The server behind Pro (your account, payments and Captylo AI) is a separate service. It keeps no recordings and no text, not even in its logs, as the privacy policy describes.