Everything we get asked most.
If your question isn't covered below, email hello@otterlog.ai and we'll add it.
Will it run on my machine?
What are the minimum system requirements?
Mac: macOS 14 (Sonoma) or later, Apple Silicon (M1/M2/M3/M4). 16 GB RAM recommended.
Windows: Windows 11, x86_64, 16 GB RAM recommended.
The minimum 8 GB will run Otterlog but the larger Whisper models will fall back to CPU.
Why Apple Silicon only on Mac?
Otterlog uses Metal acceleration for Whisper transcription. Intel Macs don't have the GPU compute to keep up with real-time meeting audio. We'd rather decline Intel users than ship them a frustrating experience.
How much disk space do I need?
About 8 GB total. The app itself is ~140 MB. Models add ~4 GB. Recordings and transcripts add up over time, but you control how long to keep them.
What permissions does Otterlog ask for?
Microphone (obviously), system audio capture, and a global hotkey. On macOS, you'll also be prompted for Screen Recording permission, which is the only way macOS lets apps capture system audio. We never take screenshots; the permission name is just how Apple labels that capability.
How big is the model download?
About 4 GB total on first run, served from models.otterlog.ai. Stored locally, never re-fetched unless you delete them. You can choose a smaller Whisper model in settings if you want to save space (at the cost of some accuracy).
What happens to my data?
Do you have access to my recordings?
No. Recordings, transcripts, and summaries never leave your computer. We don't have a server that could receive them. Verify it yourself.
What network calls does the app make?
Three, and only three: a license check, an update check, and a one-time model download. None of them send any of your content. Full list with reproduction steps at /privacy-proof.
Does the website track me?
This website uses Plausible, a cookieless analytics service, to count visits and measure ad campaigns. It doesn't track individuals, set cookies, or send your data to Google/Meta. The Otterlog app tracks nothing.
Can I use this for client or patient work?
We don't make legal compliance claims (HIPAA, GDPR, NHS DSPT, etc.) on your behalf, those are operational programmes, not features. What we can tell you: recordings never leave your computer, so the on-device processing model fits the spirit of most professional confidentiality rules. Check with your professional body if in doubt.
If I lose internet, does Otterlog stop working?
No. After first-time setup (model download), Otterlog runs fully offline. License rechecks every 7 days; if you're disconnected, there's a grace window so the app keeps working uninterrupted.
What about deletion?
Recordings, transcripts, and summaries are files on your computer. Move them to the bin, empty the bin, they're gone. We have no copy. For cryptographically permanent deletion, store them on a FileVault (Mac) or BitLocker (Windows) encrypted volume.
Money questions.
How does the 30-day refund work?
Email hello@otterlog.ai within 30 days of purchase with your order number. We process the refund within 5 business days. No exit interview, no "are you sure", no friction.
What's the founding-member tier?
The first 100 buyers get the lifetime license at $79 instead of $99. Same features, same updates, same refund. Once 100 slots are filled, the tier closes automatically.
Will I have to pay again for v1.1?
No. All Personal-tier customers get v1.1 free on whichever platform they bought (Mac or Windows). Personal + Calendar is the only tier where the extra value lights up at v1.1.
Cross-platform license?
One license covers all the computers you personally own on the same platform. If you bought Mac and now use Windows too, you need a separate Windows license. Email us if you've moved off your original platform; we work it out case-by-case.
What it does, and doesn't.
What's in v1?
Recording (mic + system audio), local transcription (Whisper), speaker diarization (pyannote), on-device summarisation (Phi-3.5-mini), full-text search, Q&A chat with transcripts, export to text/Markdown/SRT.
What's coming in v1.1?
Calendar integration (auto-attach meeting notes to calendar events), per the launch plan. Speaker enrollment ("this voice is my CEO") lands in v1.2.
What languages are supported?
English at v1. Whisper supports more languages, but we want to validate transcription quality and the summary prompts per language before claiming support. More languages in v1.x updates.
Can it transcribe Zoom / Google Meet / Teams?
Yes, by capturing system audio. Otterlog doesn't integrate with the meeting platforms directly; it captures whatever is coming through your speakers and microphone, which works with any video-conferencing tool.
The AI bit.
Which transcription model do you use?
OpenAI Whisper, running locally via whisper.cpp with Metal acceleration on Mac and equivalent acceleration on Windows. You can pick the model size in settings: smaller is faster, larger is more accurate.
Which summary model?
Microsoft's Phi-3.5-mini, quantised to fit comfortably on consumer hardware. Runs entirely locally via llama.cpp. We chose Phi over larger models because the quality-to-speed tradeoff is strong for meeting recaps and the on-device footprint stays sane.
How accurate is the transcription?
For clean English audio (one or two people, decent microphone), word error rate is comparable to cloud services. Noisy environments, heavy accents, and overlapping speech are harder; that's true of every transcription tool, cloud or local.
Do the models ever phone home?
No. Once downloaded, models run entirely on your computer. There's no telemetry layer that sends transcription or summary requests anywhere. Verify yourself.