Everything the app does, in detail.
Six capabilities, all wired together. Each feature is on by default — you don't pay more, click more, or wait longer.
Deepgram Nova-2 streaming recognition
Industry-leading accuracy in real time, with the details that decide whether a transcript is usable.
We use Deepgram Nova-2 via a short-lived (10-minute) streaming token signed by our Edge Function. Audio streams directly from your Mac to Deepgram — it never passes through our servers — and the recognized text streams back within 400 ms of the word being spoken.
- Speaker diarization: 'who said what' tagged automatically
- Sentence-boundary detection: punctuation that makes sense
- Custom dictionary: teach it your names, codes, and jargon
- Interim + final results: see words appear as they're spoken, finalized a second later
- 18 source languages supported
If the app fails to reach Deepgram, recording pauses and you're notified — your audio is not recorded silently.
18 languages, both directions
Translate what's being said into the language you read — or the other way around — in real time.
Every recognized sentence is sent through our `translate` Edge Function, which calls Google Cloud Translation first and falls back to Google Gemini when Translate can't handle the input. The result lands in your transcript alongside the original, not replacing it.
- English, Mandarin, Japanese, Spanish, French, German, Korean + 11 more
- Switch target language mid-session without losing history
- 'Transcribe only' option when source = target
- Terminology bias: your dictionary applies to translation too
- Included in the per-minute rate — no extra charge
Chat with any recording
Summarize, extract action items, find that one quote — all backed by the transcript you just captured.
After (or during) a recording, open the chat drawer and ask in plain language. The transcript text — not the audio — is sent through our `ai-chat` Edge Function to Gemini or GPT. Responses stream back via SSE.
- Summarize in 3 bullets
- Pull action items and owners
- Find every time a specific person spoke
- Translate individual answers to another language
- First 1,000 messages per month are free
We don't retain your chat history on our servers. Conversation state lives in Supabase under your account, encrypted at rest.
Floating subtitles window
A 560 × 220 always-on-top caption bar you can park over Zoom, Teams, a browser, or nothing.
The second window opens at one click and shows live transcription + translation in whatever configuration you're in. It's native macOS, respects Spaces, and doesn't interfere with screen sharing.
- Always-on-top, opt-in per-session
- Drag anywhere, snap to edges
- Configurable font size and opacity
- Picture-in-picture on macOS 14+ (optional)
TXT, SRT, Markdown, and ZIP
Move your transcript wherever your next tool is. No upgrades required.
All formats are generated locally on your Mac. SRT includes the original and translated captions. Markdown includes speaker labels and timestamps. ZIP bundles the audio (.webm), the JSON transcript, and your notes.
- SRT: for video editors — clean speaker + timestamp format
- Markdown: for writers — speaker headers + quotes
- TXT: for quick paste into email or a ticket
- ZIP: for archiving — everything in one file
Custom vocabulary for your jargon
Names, product codes, medical terms, legal citations — tell Deepgram once, never re-edit again.
Entries are stored locally and passed to Deepgram at session start via the `keywords` URL parameter. Up to 500 entries per account; bring more in via CSV import.
- Per-account dictionary (not per-session)
- CSV import / export
- Phonetic hints for unusual names
- Also applies during translation as a terminology bias
Dictionary entries leave your Mac only as URL parameters to Deepgram during active recording — they're not stored in our database.
Try all six in the same free 30 minutes.
No feature is gated. Starter, Main, and Bulk unlock exactly the same capabilities — only the number of minutes differs.