Privacy

Last updated: October 16, 2026

Capta runs on your Mac and only on your Mac. On its own, the app sends one thing: anonymous usage events, which say which steps are used and where they go wrong, carry nothing that identifies you, and stop with one switch in Settings. There is no crash reporting and no account. Everything else is something you ask for: choosing Capta → Check for Updates fetches one small version file from capta.cam (builds before 0.9.2 ask capta.idad.fyi, which now forwards to the same file) (and if you turn on "Check for updates weekly" in Settings → General — it's off by default — the same single fetch happens once a week at launch; and if you choose Install and Relaunch when a newer version is offered, Capta downloads the DMG from capta.cam, verifies its checksum and Developer ID signature, and replaces itself); searching for online music (if you use it) sends your search words to Openverse (openverse.org, the Creative-Commons media search run by WordPress.org) and downloads the track you pick; and sharing a link (if you use it) copies the exported video into your own iCloud Drive, where Apple's sharing sheet makes the link — Capta itself talks to no server for it. Automatic captions on macOS 26 or later may ask macOS to download Apple speech-model assets the first time a language is used; the audio itself is processed locally. No recording is uploaded for transcription. Choosing Help → Report an Issue… opens a report form hosted by SnagFrog, our issue tracker, with a few technical details about your Mac and Capta. This page spells out what that means in practice.

What the app can access, and why

Capta asks macOS for three core permissions, plus optional Microphone and Camera access. Each maps to a visible feature:

Recording preflight can preview your enabled camera and show microphone levels. These checks do not save audio or video.

Captions and speech models

Automatic captions use downloadable Qwen models on macOS 14+, or Apple Speech on macOS 26+. Select the caption provider separately from writing models in Settings → Models. Qwen caption downloads total 1.84 GB; transcription stays on your Mac. Qwen weights are verified community conversions hosted on Hugging Face; audio is never sent to the model host.

Optional local models

Settings → Models lets you choose built-in templates, Apple Intelligence, or downloadable Qwen3 models (0.6B and 1.7B). Downloads are optional, come from Hugging Face, and are checked against pinned file hashes before use. The host receives ordinary download request information, such as your IP address; Capta does not include recording content in these requests. Models are stored in ~/Library/Application Support/Capta/Models and can be removed in Settings.

After download, titles, social copy and chapter suggestions run on your Mac with the embedded llama.cpp runtime. The models see recording digests, clicked-element titles and, for chapters, narration text; typed characters are not included in prompts. Apple Intelligence remains optional and requires macOS 26+ with its system model available. Automatic captions use downloadable Qwen models on macOS 14+, or Apple Speech on macOS 26+. Select the caption provider separately from writing models in Settings → Models. Qwen caption downloads total 1.84 GB; transcription stays on your Mac. Review generated suggestions before using them.

What the app stores

What the app sends

Recording content is not uploaded for analysis. Screen analysis, beat detection, speech transcription, model generation, editing, and rendering all execute on your machine. Your recordings, keystrokes, voice, and music go nowhere unless you share the exported video yourself.

Two features hand things to macOS on your instruction. Share copies the exported video into iCloud Drive › Capta in your own iCloud account and opens Apple's sharing sheet, where you choose who can open the link — that upload and that link are iCloud's, under Apple's terms, and you can delete the copy from iCloud Drive at any time. Each new export gets a separate copy, so older links keep their original video; delete older copies yourself when you no longer want to share them. Mute notifications while recording runs two shortcuts you created yourself in the Shortcuts app, by name, when a recording starts and stops; Capta ships no shortcuts and reads none.

Only when you choose Report an Issue… (in the Help menu or the About window), Capta first shows a window listing what the report will carry; nothing is uploaded until you choose Continue in Browser, and Cancel sends nothing. It then opens a report form hosted by SnagFrog, our issue tracker, and passes along the Capta version and build, your macOS version, Mac model, processor type, memory size and language, which permissions Capta has (Screen Recording, Input Monitoring, Accessibility, Microphone, Camera), and the same allowlisted technical settings as Help → Export Diagnostics… (resolution, frame rate, codecs and on/off switches such as captions or the webcam). No recording, frame, file name, path, log or crash report is sent, and no snapshot of the window is taken. The one exception is a report you send after a crash: if Capta crashed, the next launch offers Send Report…, which opens the same form with the macOS crash report for that crash attached. That file holds stack traces, loaded libraries and file paths on your Mac, with your home folder written as ~; the window lets you read it and leave it out before anything is uploaded. What you type into the form, and any screenshots you attach, are used only to answer you and fix the problem. Reports are stored in the EU (Frankfurt) and replies are emailed through Resend in the US; see SnagFrog's privacy policy.

Anonymous usage data from the app

So we can see where people get stuck, Capta sends short usage events to PostHog, on its EU cloud in Frankfurt. They are anonymous by construction, not by promise:

This website

capta.cam is a static site hosted on Netlify. Netlify keeps standard server logs (IP address, request time) to operate the service; see Netlify's privacy policy. The site uses Google Analytics to count visits and to see which buttons, sections and videos on the site are used, but only after you say yes on the cookie banner. Until then, and if you say no, Google Analytics is not loaded at all: no script, no cookie, no request to Google. If you say yes, it sets its own first-party cookies so a returning visit counts as one visitor and not two, and the events carry the page, the text of the button and how far you scrolled or watched, never a name. Advertising signals are switched off. Your answer is kept in your browser's local storage, not in a cookie, so saying no does not itself set one; if your browser sends Global Privacy Control the banner does not ask and takes that as a no. You can change your choice at any time, also from the Cookies link in the footer; going from yes to no removes the analytics cookies there and then. See Google's privacy policy. The app's own usage events are described above. The “Report a bug” button on these pages opens the same SnagFrog form; it includes this page's address, your browser and any recent page errors.

Changes and contact

If a future version of Capta adds a network feature or sends anything new, this page will say so before that version ships. Questions: gorgekara@gmail.com.