SoDulce LabsPrivate AI meeting notes for Mac.
Audio never leaves your Mac.
Record meetings, get a clean transcript via on-device WhisperKit, summarize with the AI provider you choose, and tag every speaker. There’s no Minutiva server. Your data is yours. A private alternative to Otter, Fireflies, and Granola — built for Mac.
Download beta — freeSigned & notarized by Apple · macOS 14.4+ · 34 MB
Features
built for privacy.
On-device transcription
WhisperKit runs on Apple Silicon’s Neural Engine. Your audio is transcribed on your Mac with no cloud round-trip. The audio file never leaves your machine.
Both sides of the call
System audio + microphone in two separate streams via CoreAudio Process Tap. Works with AirPods, Bluetooth headphones, and any output device. Every segment labeled by source.
Bring your own AI
Pick your AI provider: local (Ollama, LM Studio) or cloud (OpenAI, Anthropic, OpenRouter, custom). Your API key, your account. Minutiva never proxies the request.
Speaker tagging PREMIUM
Identify each distinct voice in the meeting and group segments by speaker. Name "Marisa" once — the entire transcript relabels. New in v0.6: VBxHMM clustering closes false-positives that used to over-split 1:1 calls.
People — voice address book PREMIUM
Naming a speaker makes them a Person. The next meeting asks "Looks like Marisa — link?" with one-click accept. Builds your private voice library on-device. New in v0.6.
Resilient capture
Pull out AirPods mid-meeting? Transcript stays continuous. Smart chunking lands transcripts at natural pauses, not 30-second cuts. New device-switch handling in v0.6.
Organize at scale
Date sections, pin meetings, tag, search across transcripts and summaries, sort, filter. Export to TXT, Markdown, CSV, PDF or via macOS Share Sheet.
Why Minutiva Private?
Audio is processed on your Mac. There is no SoDulce Labs server in the data path — ever. The closest thing to a server we touch is Sparkle’s auto-update feed.
Run a local model (Ollama, LM Studio) for maximum privacy, or use a cloud provider you trust. We don’t lock you in. Your bill, your terms.
We capture audio from your Mac directly via CoreAudio Process Tap — no third-party bot joins your Zoom, Meet, Teams, Slack huddle, or FaceTime call.
Meetings live in ~/Library/Application Support/MinutivaPrivate. Browse them in Finder, back them up, delete them. They’re yours.
Free during the closed beta.
No card required.
Free
$0forever
- ✔ Unlimited meetings recorded
- ✔ On-device transcription (WhisperKit)
- ✔ AI summary via any provider you bring
- ✔ Library, search, tags, pins
- ✔ Export to TXT / Markdown / PDF / CSV
- ✔ Auto-update via Sparkle
Premium BETA
$0during closed beta
- ✔ Everything in Free, plus:
- ✔ Speaker tagging (VBxHMM) — auto-detect each distinct voice
- ✔ Name a speaker once → entire transcript relabels
- ✔ People — voice address book that remembers across meetings
- ✔ Cross-meeting "is this Marisa?" suggestions
- ✔ Split & merge clusters when the model gets it wrong
- ✔ Voice sample playback per speaker
- ✔ Future Premium features go here. Beta key holders keep access.
Download v0.6.0-beta11
✨ New since beta8: name a speaker once and they're remembered across meetings (People address book), meetings keep the title you typed, plus mic-release and meeting-rename fixes. What's new →
Signed by SoDolce Labs LLC and notarized by Apple. Gatekeeper will allow it without warnings.
Download .dmg (34 MB)macOS 14.4 (Sonoma) or later · Apple Silicon recommended
For automatic updates: launch the app and leave the “Receive beta updates” toggle on (Settings → Advanced).
Frequently asked questions
Does Minutiva Private upload my meetings to a server?
No. There is no Minutiva Labs server in the data path. Audio is transcribed on your Mac using WhisperKit on the Apple Neural Engine. Summaries are sent directly from your Mac to the AI provider you choose, using your own API key. Recording files live on your Mac in ~/Library/Application Support/MinutivaPrivate.
Does it work with Zoom, Google Meet, and Microsoft Teams?
Yes. Minutiva Private captures audio from your Mac itself using a CoreAudio Process Tap, so it works with any meeting app — Zoom, Meet, Teams, Slack huddles, FaceTime, Webex, etc. — without joining as a bot. Both your microphone and the meeting audio are recorded as separate streams.
What is speaker tagging and is it Premium?
Speaker tagging identifies each distinct voice in the meeting and groups segments by speaker. You can name each speaker once and the transcript labels them throughout. Powered by WeSpeaker ResNet34-LM with VBxHMM clustering (new in v0.6 — closes the K=2 false-positive that used to over-split 1:1 calls). People, also new in v0.6, remembers named speakers across meetings and suggests links in new ones. Both are Premium features, free during the closed beta — request a key below.
Which AI providers can I use for summaries?
Any OpenAI-compatible API: OpenAI, Anthropic, OpenRouter, Ollama (local), LM Studio (local), or any custom endpoint. You add the provider with your own API key in Settings. Minutiva never proxies the request — it goes directly from your Mac to the provider.
How is this different from Otter, Fireflies, or Granola?
Those products send your meeting audio to their servers. Minutiva Private doesn’t have a server. The audio stays on your Mac, transcription runs on your Mac via WhisperKit, and AI summarization uses your own provider account directly. Trade-off: you bring your own AI (or run one locally), and the app is Mac-only.
What Mac do I need?
macOS 14.4 (Sonoma) or later. Apple Silicon strongly recommended for fast on-device transcription, but Intel Macs also work.
Is Minutiva Private signed and notarized?
Yes. The app is signed with our Developer ID Application certificate and notarized by Apple, so Gatekeeper allows it without warnings. No App Store or TestFlight required.
How do I get a Premium beta key?
Request one here — quick form, takes 30 seconds. We’re hand-issuing keys to early testers and they’re free for the duration of the beta. You’ll get the key by email within 24 hours. Once you have it, drop it into Settings → License.
Built by Rafael Hernandez at SoDulce Labs LLC. For people whose meetings shouldn’t live on someone else’s server.