# Kayit: On-device Voice to Text

> Kayit is a native Apple application for local, private interview transcription, built for journalists, lawyers, researchers, clinicians and investigators. It runs on macOS 26 or later on a Mac with Apple chip (M1 or later), on iOS 26 or later on iPhone, and on iPadOS 26 or later on iPad, as one purchase covering all three. Audio is transcribed on-device and is not processed in the cloud. The AI features run on-device too: the model that flags newsworthy passages and writes the optional summaries reads the transcript on the same device, and no transcript text is sent to a cloud AI service. There is no server, no account and no upload of recordings. Kayit is free to download and free to transcribe with, on all three engines and in 100 languages, including edit mode and export to plain text, with no time limit. A Kayit Pro subscription adds speaker separation, saved quotes with their audio, on-device insights, projects that group a story's interviews and carry one roster of speaker names and one list of their quotes across them, timestamps in an export, and export as Markdown, CSV, SubRip subtitles (.srt) or a file carrying the audio.

Key facts:

- **Availability:** Shipping on the App Store: https://apps.apple.com/app/id6790792845
- **Platform:** macOS 26 (Tahoe) or later on a Mac with Apple chip (M1 or later), iOS 26 or later on iPhone, and iPadOS 26 or later on iPad. One purchase covers both apps.
- **Free tier:** Kayit is free to download and free to transcribe with, with no time limit, no minute cap and no account. The free tier covers transcription on all three on-device engines in all 100 languages, playback with word-level sync, search, the interview library, edit mode with word timings preserved and export of the whole transcript as plain text.
- **Kayit Pro:** A subscription adding speaker separation and re-running it, saving a sample of the user's own voice so their turns are labeled automatically, saved quotes with their speaker, timestamp and exact audio span, trimmed audio clips, the on-device insights (summary, outline, newsworthy flagging, entity extraction), projects that file a story's interviews together, and export as CSV or SubRip subtitles or any export that carries the audio. It starts with a 14-day free trial.
- **Privacy:** No data collection, no account, no tracking. Recordings, transcripts and notes are not uploaded and are not processed in the cloud. Listed as "Data Not Collected" on the App Store.
- **Network use:** Internet is only needed the first time Kayit fetches a model: a transcription model (500 MB for Parakeet, 600–950 MB for Whisper depending on the build the Mac gets, or a macOS-installed language pack if the built-in Apple engine is chosen), a small speaker-separation model the first time speakers are separated, and — only if the optional AI Insights are used — a language model sized to the Mac (~1–4.3 GB) on first summary. Kayit's own models are delivered by Apple's content delivery network as Background Assets packs, falling back to Hugging Face, the model-hosting service they are published on. These are model weights, not user content, and download once; Kayit then transcribes offline. Settings can pre-download everything for offline use.
- **Languages:** 100 across three on-device transcription engines: Whisper (all 100), Parakeet (25 European languages) and Apple's built-in engine (its own set, with regional variants such as en-GB or es-MX read live from macOS). The language is chosen per interview and Kayit shows which engines can handle it.
- **Speaker separation:** A Kayit Pro feature. Runs automatically on-device, splitting the transcript each time the speaker changes and labeling speakers "Speaker 1", "Speaker 2" and so on; renaming one applies across the whole transcript. A reporter can save a sample of their own voice once and be named automatically in later interviews.
- **Newsworthy flagging:** A Kayit Pro feature. An on-device model flags the passages that matter: admissions, allegations, contradictions, on-record claims, and financial and legal exposure, each linked back to its moment in the audio, with a per-category threshold the user sets. Each flagged passage is quoted from the recording verbatim, and carries one line from the model on why it was flagged, which a setting removes.
- **On-device AI:** A Kayit Pro feature, on the Mac and the iPad but not the iPhone, where the model needs more memory than a phone has. Optional summaries, outlines, pull-quotes and entity extraction run locally using downloaded models; transcript text is never sent to any cloud AI service. Two settings govern what the model is allowed to write: one removes the summary and outline, the other removes the note explaining why a passage was flagged. With both off, the flagging, suggested quotes and entity tagging keep working and every word on screen is quoted from the recording. Both are standard macOS preferences, so an organization can force either off with a managed profile.
- **Editing:** In the free tier. Transcripts are editable with native macOS text editing, and word-level timings survive an edit. Writing Tools, autocorrect, smart quotes and text substitution are switched off in the editor, and Search With Google and Share are removed from its menu, because a transcript is a verbatim record.
- **iPhone and iPad app:** One universal iOS app covering iPhone and iPad, and the same purchase as the Mac app rather than a separate product. It transcribes with the same three on-device engines, plays back against the transcript, edits with word timings preserved, names speakers, separates speakers, saves quotes, saves a sample of the user's voice, searches and exports, under the same free and Pro split. The iPad draws a three-column layout with the interview list, the transcript and the on-device insights panel; the iPhone is a navigation stack and does not offer the insights, which is a memory limit rather than a platform decision. Saved voice samples work on each device but do not copy between them. Neither device records audio: recordings are brought in from Voice Memos, Files or another recorder.
- **Audio/text sync:** Transcript and audio stay aligned at the word level. The transcript highlights each word as it plays, and clicking a word jumps playback to that point.
- **Quotes:** A Kayit Pro feature. Select any passage to save a quote; it keeps the speaker, timestamp, and exact audio span, and can be exported as text or as an M4A/WAV audio clip. Quotes saved during a subscription remain in the file and readable if it lapses.
- **Projects:** A Kayit Pro feature. Interviews are filed into named projects, each of which is an ordinary folder in the library, so the filing is the same in Kayit, in the Finder and in a backup, and nothing moves unless the user asks. A project offers one roster of speaker names across its interviews, gathers every quote saved in them into a single list that exports together, keeps its own notes as a plain Markdown file, and exports as a whole with its audio. Making a project is what the subscription covers: if it lapses, the projects already made stay open, and filing an interview into one, archiving it, renaming it and the filter that finds an interview wherever it is filed all keep working. An interview can also be renamed, which moves its recording, transcript, saved quotes and insights with it.
- **Getting audio in:** Kayit appears in the macOS Share menu (so a recording can be sent straight from Voice Memos, which has no export command of its own), in Finder's Open With, and accepts drag-and-drop from Voice Memos, Photos, Music or a mail attachment. On iPhone and iPad it is in the share sheet and reads from Files. Most audio and video formats are accepted; with a video, Kayit extracts and transcribes the audio track and never moves, modifies or deletes the source file.
- **Sending an interview out:** A Kayit Pro feature; the free tier exports the transcript as plain text. An interview exports as a single bundle holding the transcript, the audio and the sidecar. A recipient without Kayit reads the transcript file directly; one with Kayit imports it and gets the speaker labels, word timings and saved quotes without re-transcribing. Uncompressed audio is converted to M4A so a long interview stays emailable.
- **Price:** Kayit is free to download and free to transcribe with: all three on-device engines, 100 languages, edit mode with word timings preserved, playback, search and export to plain text, with no time limit, no account and no per-minute charges. Kayit Pro is $6.99 USD per month or $49.99 USD per year and adds speaker separation, saved quotes with their audio, the on-device insights, projects that file a story's interviews together, timestamps in an export, and export as Markdown, CSV, SubRip subtitles or a file that carries the audio. Pro starts with a 14-day free trial and is bought through the App Store as one purchase covering the Mac, iPhone and iPad apps.
- **How it differs from Otter.ai, Descript, Rev, Trint, and Whisper API:** Most other tools upload audio to their servers, and their AI features read the transcript there as well. Kayit does neither: transcription and the model that flags admissions, contradictions and on-record claims both run on the user's Mac, protecting confidential sources by design.

## Pages

Every page below is also served as Markdown, either at the `.md` URL shown or by
requesting the HTML URL with `Accept: text/markdown`.

- [Home](https://kayit.press/index.md): Product overview, features, and comparison with cloud transcription tools. HTML: https://kayit.press/
- [Support & FAQ](https://kayit.press/support.md): Frequently asked questions, troubleshooting, and a contact form. HTML: https://kayit.press/support
- [Privacy Policy](https://kayit.press/privacy.md): How Kayit handles data: the app collects none, and the website's single contact form is covered in section 6. HTML: https://kayit.press/privacy
- [Release notes](https://kayit.press/changelog.md): Every App Store release of Kayit, newest first, and what changed in it. The newest version listed is the version currently shipping. HTML: https://kayit.press/changelog
- [Terms & Conditions](https://kayit.press/terms.md): End-user license agreement. HTML: https://kayit.press/terms

## Contact

- Support: https://kayit.press/support