SyncBrief Documentation

Welcome

SyncBrief turns meeting audio into structured notes: a summary, a timeline of topics, decisions, action items with owners, open questions, speaker-labeled transcripts, and a tappable mind map. It runs on iPhone, Mac and Apple Watch from one shared codebase, and it is built around three rules:

• No servers, no accounts. Your library lives in an encrypted database on your device. There is nothing to sign up for.
• Bring your own key. Cloud transcription and summarizing talk directly to the provider you choose, under your own API key — or run fully on-device.
• Importing is always free. Adding audio contacts nobody. The only step that uses your provider (and your money) is the one you explicitly start.

The fastest path to a first summary: record something with the mic button, tap it in the library, tap Transcribe, and open the Summary tab when it finishes.

Choose your providers

Settings → Providers holds two choices:

Transcription

• OpenAI — Whisper-family API models, including a diarizing model that labels speakers.
• Deepgram — fast, diarized, word-level timing; strong multilingual support.
• AssemblyAI — diarized transcription with speaker labels.
• Whisper (on-device) — runs on your device via CoreML; no key, no internet, no per-minute cost. Detects the language itself.
• Apple (on-device) — the system speech recognizer; free and offline.

Summary

• Anthropic or OpenAI — cloud models with large contexts, using your key.
• Apple (on-device) — Apple Intelligence on the device; free and offline, with a smaller context (long transcripts are summarized in passes automatically).

The Model row under each choice lists that provider's current models with live prices (per minute for transcription, per million tokens for summaries), fetched while the Settings screen is open. You can always type a model id the list doesn't know yet — lists go stale faster than releases ship.

API keys

Cloud providers need your own API key, created in that provider's console (OpenAI, Deepgram, AssemblyAI, Anthropic). Paste it once in Settings → API keys.

• Keys are stored in the Keychain on your device — never in the library database, in preferences, or in exports, and never sent to us.
• Verify key makes a free metadata request against the provider to prove the key works before any audio is sent.
• Remove a saved key anytime; an empty key deletes the Keychain item.

Fully offline mode

Choose Whisper (on-device) or Apple (on-device) for transcription and Apple (on-device) for summaries, and nothing about your meetings ever leaves the device — no keys needed, no per-use costs, aeroplane-mode friendly.

• On first use, Whisper and the speaker-identification models download once from a public model host (Hugging Face). The audio itself never goes anywhere.
• Apple's on-device summarizer requires Apple Intelligence to be enabled on the device.
• On-device Whisper is the reliable fallback for any language — it detects the language itself.

The Mac window

The Mac app opens centered, split horizontally: the Imported Audio Files table on top (30%), the selected recording's detail below (70%), with the newest recording already selected.

• Drag the dark divider between the panes to rebalance them — the split is remembered across launches.
• The expand button in each pane's top-right corner gives that pane the whole window; click again to return to the split.
• The table sorts by any column. Each row carries its own action: a Transcribe button (with a speaker head-count picker) before transcription, a progress bar during, and a Transcribed label with a speakers menu after.
• The selected recording's own actions — Share, Regenerate, Speakers — sit at the top of the detail pane; Share appears once the recording is transcribed.

Recording on iPhone & Apple Watch

• iPhone: tap the mic button in the bottom bar. Recording continues with the phone locked (long meetings are fine). Stop to save into the library.
• Control Center: add the SyncBrief control for one-press access to the recorder from anywhere; Siri and Shortcuts can start it too.
• Apple Watch: record from the watch app or its watch-face complication; the file transfers to your iPhone's library automatically over Watch Connectivity. Requires an Apple Watch Series 4, SE or later (watchOS 10) — it installs automatically with the iPhone app on a compatible watch, and won't appear on a Series 3 or older.

Recordings land untranscribed — they cost nothing until you choose to transcribe.

Mac: recording online calls (both sides)

The Mac recorder can capture your microphone and everything the Mac plays — which means both sides of a Zoom, Google Meet, Teams, WhatsApp Desktop or FaceTime call land in one file, without bots or plugins.

1. Open the recorder (mic button) on the Mac.
2. Enable “Also record this Mac's audio (online calls)”.
3. Start recording. The first time, macOS asks permission for system-audio recording (and the microphone) — allow both.

Under the hood this uses Apple's Core Audio capture APIs (macOS 14.2+), the same sanctioned mechanism other meeting note-takers use. You keep hearing your call normally while it records.

Mac: meeting detection

Settings → Meeting detection. When another app starts using the microphone — the universal sign a call or meeting just began — SyncBrief can act:

• Notify when a meeting starts (default): a notification appears with a Record this meeting button. One click starts the both-sides capture.
• Record meetings automatically (off by default): standing consent — every detected meeting is recorded and filed without asking.

Details worth knowing:

• Detection reads no audio — only whether some other process is using the mic.
• Recording stops on its own about a minute after the meeting's mic goes quiet, so a dropped-and-rejoined call stays one recording.
• In automatic mode, captures shorter than 90 seconds of actual mic use are discarded — dialing someone who never answers doesn't become a junk “meeting”.
• SyncBrief must be running for detection to work.

Calendar matching & reminders

Settings → Calendar. SyncBrief reads the calendars already on your device — iCloud, Google, Exchange, whatever accounts the system syncs — with read-only access, entirely on-device. No account is connected to anyone.

• Match meetings from your calendar: recordings started while a calendar event is underway are titled after the event (“Weekly Standup”, not “Meeting 28 Aug at 12.29”), including the Mac's auto-detected meetings. The record sheet also lists today's meetings — tap one to record under its name.
• Remind me before meetings: a local notification two minutes before each meeting with a Record button. On the Mac it starts the capture directly; on iPhone it opens straight into the recorder.

Overlapping events resolve to the one that started most recently — a 1:1 inside a blocked-out afternoon wins.

Phone calls on iPhone

iOS does not let any third-party app record cellular or WhatsApp calls — that's a platform rule, not a SyncBrief limitation. The working paths:

• Cellular calls: use iOS's built-in call recording (iOS 18.1+, the record button in the Phone app; both parties hear a notice). The recording saves to Notes — share it to SyncBrief from there, and it transcribes like anything else.
• WhatsApp / Zoom / Meet calls: take them on the Mac and use the Mac recorder — it captures both sides.
• In-person conversations: the normal mic recorder.

Files, share sheet & drag-and-drop

Any audio file can join the library, from anywhere:

• iPhone: the Import button (Files picker), or share audio to SyncBrief from any app — Voice Memos, Notes, WhatsApp documents, Mail attachments.
• Mac: drag files onto the window or the Dock icon, use Finder's Open With, or the Import button.
• Common formats are handled (m4a, mp3, wav, and more); files are transcoded internally as needed.

Importing copies the file into the app's container and indexes it — no provider is contacted and nothing is spent. Duplicates are detected by content, so importing the same audio twice doesn't create two entries.

USB voice recorders

• Mac: plug the recorder in. SyncBrief watches for removable volumes carrying audio and offers what's new on the device — recordings already in your library are skipped automatically (matched by content, since recorders reuse names like REC001.MP3). Nothing is read beyond filenames until you accept.
• iPhone: connect the recorder — USB-C iPhones take it directly; Lightning iPhones need Apple's Lightning-to-USB 3 camera adapter (use the powered variant). Then tap Import and pick the drive under Locations in the file browser. iOS offers no plug-in detection to apps, so this path is manual by platform design.

Watched folder (Mac, Pro)

Settings → Watched folder → choose a folder. Any audio dropped into it is imported automatically — pair it with a cloud-synced folder (iCloud Drive, Dropbox) and recordings made elsewhere flow into the library by themselves. The folder permission survives relaunches; stop watching anytime. Part of Pro.

Transcribing

Select a recording and start transcription explicitly — the row's Transcribe button, its context menu, or Transcribe all waiting. Before starting, you can tell the app how many people were in the meeting (the person icon beside the button) so speaker detection gets it right the first time. What happens:

1. Audio is normalized (16 kHz mono) on-device.
2. Long files are split on silence boundaries into overlapping chunks so provider limits never truncate a meeting; results are stitched back onto one timeline and de-duplicated.
3. Speaker detection runs in parallel with transcription over the same audio, so labels cost no extra time.
4. The transcript is saved before summarizing — a summary failure never wastes the transcription you paid for; regenerate the summary anytime.
5. The summary is validated: any timestamp the model can't back up with the transcript is flagged rather than shown as fact.

Each row shows its own progress bar with percent while it runs, and several files can transcribe at once — start them all and watch each row advance. A finished row shows a Transcribed label.

Before the first upload to any provider, the app shows a one-time notice naming it. A per-recording upload log records exactly which provider received what.

Languages

• Auto-detect (default) lets the provider identify the language.
• Set an explicit language in Settings when a provider mis-detects — or Multilingual for meetings that code-switch (Hindi–English, for example).
• Whisper (on-device) detects language itself and is the dependable fallback for any recording.

If a transcription comes back empty, the language setting is the first thing to check — see troubleshooting.

Speakers

Transcript lines carry speaker labels — S1, S2, S3… in order of first appearance:

• Deepgram, AssemblyAI and OpenAI's diarizing model label speakers themselves.
• For transcribers that can't tell voices apart (Whisper and Apple on-device), SyncBrief identifies speakers on the device — voice-signature models analyze each voice's spectral fingerprint and separate who spoke when, in parallel with transcription. Toggle: Settings → Providers → Identify speakers. The models download once on first use; audio never leaves the device.
• Tell it the head-count. Automatic detection can merge similar voices recorded from across a table — a five-person meeting coming back as S1/S2. Before transcribing, set the count with the person icon beside the Transcribe button; afterwards, use the Speakers menu (in the table row or the detail view) to re-detect with 2–8 speakers, or Auto. Re-detection is free and entirely on-device.
• Tap any speaker label to rename it — “S1” becomes “Priya” everywhere in that recording, including summaries and exports. Re-detecting speakers clears renames, since the old names were tied to the old labels.

Summaries, timeline & mind map

Each transcribed recording has four tabs:

• Summary — the brief: overview, decisions, action items with owners, open questions. AI-generated content is labeled, and everything is editable; your edits survive regeneration (the original, regenerated and edited versions are all kept).
• Mind Map — the meeting as a collapsible tree: the title at the root, each topic beneath with its time and gist, and the decisions and action items indented under the topic being discussed when they happened. Collapse any branch with its chevron to skim a dense meeting (a badge shows how much is folded away), and tap any node to jump playback to that moment.
• Timeline — topics with time ranges; tap to seek.
• Transcript — the full text, synced to playback, speaker-labeled.

Regenerate lets you re-summarize with a different provider or model — useful for comparing, or after adding your own notes, which steer what the summary covers (notes never invent timestamps). Every timestamp shown comes from the transcript; out-of-range ones are flagged, not displayed as fact.

Search

The search field searches everything, instantly, on-device: transcript text, titles, tags, speaker names and summary text, with prefix matching as you type. Results jump straight to the matching recording. No index ever leaves the device.

Export

• Any recording exports as Markdown, PDF, plain text or JSON — transcript, summary, speakers and flagged timestamps included.
• Exported summaries carry the caution “AI-generated — AI can make mistakes.” If you have edited the summary in the app, the export can instead say it was reviewed and edited by you: set your name in Settings → Export → Author, and the app asks for your consent on each such export — credit is never added silently.
• Mac batch export (Pro): every recording in the library, one file each, into a folder you pick. Batch exports always use the standard caution.
• Exports never contain your API keys.

Storage

• Settings → Storage shows the library database and audio sizes.
• Purge audio on a recording deletes the sound file but keeps the transcript and summary — big meetings stop costing disk once processed.
• Deleting a recording removes everything about it. Deleting the app removes the entire library.

Free tier & Pro

• Free: every feature, with a monthly transcription allowance. Recording, importing, search and your library are never limited.
• Pro: a per-platform subscription. Pro for iPhone ($5.99/month, $15.99/quarter, $60.99/year) gives unlimited transcription on iPhone and Watch; Pro for Mac ($9.99/month, $25.99/quarter, $99.99/year) gives unlimited transcription plus the Mac's watched folder and batch export. Each has a 14-day free trial and unlocks its own platform — subscribe on the device you use, or both. Payment is handled by Apple; manage or cancel in your Apple subscription settings (Settings → SyncBrief Pro → Manage subscription).
• Cloud provider usage is billed separately by the provider under your own key, Pro or not; the on-device options are free either way.

Privacy & security

• The library is an encrypted database (SQLCipher) on your device; the encryption key lives in the Keychain.
• Audio leaves the device only when you start a cloud transcription or summary, directly to your chosen provider under your key. The one-time provider notice and per-recording upload log keep that visible.
• No accounts, no analytics, no trackers, no SyncBrief servers.
• Calendar access (optional) is read-only and on-device. Meeting detection reads no audio. Bug-report diagnostics are visible to you in the email draft before anything is sent.

The full policies: Privacy · Terms.

FAQ

Do I need an account?

No. SyncBrief has no servers and no sign-up. Your library lives in an encrypted database on your device, and the optional Pro subscription is handled entirely by Apple.

Does my audio get uploaded anywhere?

Only when you explicitly transcribe with a cloud provider — and then only to that provider, directly, under your own API key. Choose the on-device transcriber and summarizer and nothing ever leaves the device. Importing, recording and searching never contact anyone.

What does it cost?

The app is free with every feature and a monthly transcription allowance. Pro removes the allowance — sold per platform (Pro for iPhone, Pro for Mac), each monthly, quarterly or yearly with a 14-day free trial. Cloud providers bill you separately under your own key; the on-device options cost nothing per use.

Do I have to buy Pro twice for iPhone and Mac?

The two are separate subscriptions, priced separately, because the Mac app carries extra automation (watched folder, batch export, call capture). Subscribe on the platform you use — or both, if you use both.

Which Apple Watch works?

Series 4, SE or later (watchOS 10). The watch app installs automatically with the iPhone app on a compatible watch; it cannot appear on a Series 3 or older.

Can it record phone calls on my iPhone?

iOS forbids third-party apps from recording cellular or WhatsApp calls. Use iOS's built-in call recording (iOS 18.1+) and share the file to SyncBrief — or take the call on your Mac, where SyncBrief records both sides.

Why does a five-person meeting show only S1 and S2?

Automatic voice clustering merges similar voices recorded from across a room. Tell the app the real head-count — the person icon beside Transcribe, or the Speakers menu afterwards — and it separates them.

My recording is in Hindi (or another language) — will it work?

Yes. Whisper (on-device) detects the language itself and is the dependable choice; for cloud providers, set the language explicitly, or Multilingual for meetings that switch languages mid-sentence.

The summary failed — did I lose the transcription I paid for?

No. The transcript is saved before summarizing, always. Open the recording and Regenerate to try the summary again, with any provider.

Can I trust the summary?

It's AI-generated and labeled as such — read it before relying on it. Every timestamp comes from the transcript, never invented by the model, and anything the transcript can't back up is flagged. Your edits are kept forever alongside the AI versions, and exports say whether a human reviewed them.

How do I delete my data?

Delete any recording from its row (or purge just the audio and keep the transcript). Everything is local — deleting it here deletes it, full stop, because there is no server copy anywhere.

Troubleshooting

“Found no speech in this recording”

The provider heard audio but recognized nothing — almost always a language mismatch. Set Settings → Language to match the recording (or Multilingual), or switch to Whisper (on-device), which detects the language itself.

Mac call recording has only one side

System Settings → Privacy & Security: SyncBrief needs both Microphone and System Audio Recording allowed. Each is asked once on first use; if one was declined, enable it there and record again.

Meeting detection isn't firing

SyncBrief must be running, the toggle on (Settings → Meeting detection), and notifications allowed. Note the meeting app must actually open the microphone — watching a webinar without mic access won't trigger it.

A meeting recorded as two entries

The call's mic went quiet for over a minute mid-meeting (long hold, network drop longer than the grace window). Both parts are in the library; transcribe both.

Calendar meetings aren't listed

Check the toggle (Settings → Calendar) and that calendar access is allowed in system settings. All-day events are deliberately ignored.

Library problems

Settings → Support → Recover corrupted library runs an integrity check and rebuilds the search index, then reports what it found. Recordings' audio files are stored separately and survive index issues.

Contact support

Settings → Support has the fastest paths:

• Send feedback — opens an email draft to us with your app version filled in.
• Report a bug — a structured form (description, steps, expected, actual) with optional diagnostics you can review in the draft before sending.

Or write directly: support@aidigiapps.com — a human reads it.