MOQI Product Introduction (English)

One line

In meetings, AI that helps you answer — not just understand.

The problem

Caption tools solve "understanding what was said." MOQI solves the next step: when it's your turn to speak, answering well. Especially for people working in a non-native language — you understand, you can speak, but organizing a complex answer in seconds under pressure is hard.

Core features

• Two-track live subtitles: you (microphone) and the other side (system audio) on separate tracks — the capture itself knows who spoke, no guessing.
• Answer prompt cards: seconds after a question lands, answer points appear on your screen only — keywords bilingual (including your native language), full sentences in the meeting language, ready to speak.
• Pre-meeting material (optional): cards work without it. Load the product brief, the pricing, the JD or the CV and they get more precise — and quote your own words. If a number is not in the material, the card says so instead of inventing one.
• Three scenario models: job interview / sales call / customer support — card organization follows the scenario.
• One-click summary & archive: key points, decisions and follow-ups written up automatically; archives live only on your own device. Every meeting is saved as a record the moment it ends — you never have to remember to archive it — and you can summarize it any time afterwards.
• Meeting history and archives (iOS too since build 76; macOS / Windows before that): browse past meetings, click one to read it. "Meeting records" are the ones not summarized yet, "Archived" are the ones with a summary — two states of the same data. Archiving an already-archived meeting asks before overwriting; it never silently replaces the previous summary. Point the archive folder at your own cloud drive and both machines read the same files (one format, both platforms).
• Translation under the other side's subtitles when their language differs from yours (since build 76): on by default, collapsible any time — collapsing hides the line on screen, the translation stays in the copied text and in the archive.
• Maximize either pane (since build 76): a button in the top corner of the prompt-card and subtitle panes; press once to fill the window, press again to restore the split you dragged.
• Apple Watch standalone modes: face-to-face translation and meeting subtitles from the wrist, with no phone or computer involved (since build 69 the watch talks to our server directly). Before a meeting, "Sync materials from iPhone" pulls over what you typed on the phone, and every card after that carries it; during one, tapping a subtitle line flips its Me / Them label and the next card follows the correction (since build 72).

Works with headphones on (the structural difference vs a hardware recorder)

MOQI captures at the system-audio layer — it takes the meeting app's own output stream, not the air in the room. Wear noise-cancelling headphones, AirPods, anything: the other party's voice is still captured.

By contrast, any air-listening hardware recorder captures only your half of the conversation in that setup — and a summary built from half a conversation isn't merely incomplete, it reads complete while reaching the wrong conclusions.

Basis: multiple rounds of owner testing with headphones during early development (including real interview sessions). ⚠️ Invalidation: both platforms' capture paths have changed since. To attach a build number to this claim, re-confirm once on the current build.

Where your data goes (stated plainly)

Audio and transcripts are forwarded through our servers to the transcription/model providers — we do not store them; meeting archives and logs stay on your own device. Capture never starts before an explicit consent confirmation, revocable at any time — after withdrawal, even results already in flight are discarded. Until there is an account system, no personal information is collected: your invite code is your entire identity.

Platforms

Infrastructure (one line)

Two-region backend with automatic client-side failover: if the primary dies mid-meeting, subtitles and prompt cards recover within about half a minute — measured, not aspirational.

Beta caveats (the honest part)

• Transcription accuracy depends on accent / noise / domain vocabulary; cards can occasionally miss. Screenshots of cards saying something untrue are especially welcome — "a card promising something the product doesn't do" is a top-priority defect here.
• The Windows installer is not code-signed yet; SmartScreen will warn (explained on the download page).
• First launch needs an invite code (one per person, with its own quota). Get one on the request page: https://moqi.taolab.dev/invite (if the page is down, email dtzp555@gmail.com instead).

Join the beta

1. Open https://moqi.taolab.dev/invite, leave your email and pick a platform (Windows / macOS / iOS). The invite code and install steps arrive by email within minutes (up to 4 live codes per person sharing one quota; if you lose it, request again with the same email — you get an extra code, your installed devices keep working, up to 4 live codes).
2. Windows: install from https://dl.taolab.dev and paste the code at first launch; macOS / iOS: reply to the invite email with your Apple ID email, we send a TestFlight invite; paste the code at first launch, and the iPhone syncs it to your Apple Watch.
3. Feedback by replying to the email; screenshots of cards saying something untrue are the most useful thing you can send.

Screenshot guide