Launching today
Vertex Meet
Meeting minutes that hear and see, not AI summaries
1 follower
Meeting minutes that hear and see, not AI summaries
1 follower
Vertex Meet records your meeting on iPhone and writes the discussion into complete minutes: every topic, decision and action item. On Cloud Plus, photograph the slide or whiteboard and the numbers land in the minutes under the topic discussed. Sensitive meeting replaces names, ID numbers and contact details in the text on your device before sending. Tasks sync with Apple Reminders. Audio never stored on our servers; free tier on your own OpenAI key. English, German, Finnish, Swedish, Estonian.
Hi Product Hunt,
I'm Andro, from Estonia. I run a small consulting company and sit through a lot of meetings. Every AI tool I tried gave me a summary, and the details that matter were exactly what got summarized away. I could never send the result to the other participants without rewriting it first. The second problem was waiting: after every meeting you sit and watch a progress bar.
So we built Vertex Meet.
It records on the iPhone and then does two passes. The first rewrites the raw transcript into clean prose without dropping anything: filler words out, everything else stays, chronology preserved. The second reorganizes that clean text into topics, decisions and action items. The rule we enforce in the prompt is "reorganize, never delete" - a 500-word output from an 8000-word meeting is treated as a failure, not a summary.
Five things I'd point out:
The minutes can see. Audio never carries a table or a chart, so on Cloud Plus you can photograph the slide, the spreadsheet on the screen or the whiteboard while recording. The photo gets a timestamp, the model reads what is on it, and the numbers end up in the minutes under the topic that was being discussed at that moment - not as a separate attachment. A table on the slide stays a table in the minutes and in the export. No extra hardware, just the phone you already have. This shipped this month.
Sensitive meeting mode. Turn it on before you record, and your own phone finds the names, ID numbers, phone numbers, emails and bank account numbers in the text and swaps them for pseudonyms before anything is sent for processing. Your device puts the real details back into the finished minutes, and a line above the protocol tells you how many were replaced. Two honest limits: a name that only ever appears at the start of a sentence can survive, because the first word of a sentence is capitalised anyway, and the audio itself is not masked, although according to OpenAI's own documentation audio sent for transcription is not retained. So this is data masking, not anonymisation, and we do not call it anything else. It is a Pro feature, and it is the thing that decides whether a lawyer, an HR lead or an auditor is allowed to use a tool like this at all.
Privacy is architectural, not a promise. Audio is never stored on our servers, photos stay on the device, minutes live in your own iCloud, and on the free tier you bring your own OpenAI key, so requests go straight from the phone and we never see anything.
Corrections are fast. When the transcription mishears a name or a term, a long press on that word replaces it everywhere in the minutes at once. That is the difference between "nice output" and "document I can send".
Speed on the top tier comes from transcribing during the meeting in 5-minute chunks, so the minutes are ready moments after you hit stop. The hard part was iOS suspending the app about 30 seconds after the phone locks, with no working pattern for a background request where you also need the response body. We got past that, and then the part I did not expect: what people actually wanted was not to be stuck on the progress screen at all. So you can move to another part of the app while the minutes are still being written, and on the paid plans you can start the next meeting straight away while the previous protocol finishes in the background. On a day with three meetings back to back, that turned out to matter more than shaving seconds off the processing.
It works in English, German, Finnish, Swedish and Estonian, interface and minutes both. Free tier is BYOK, paid plans if you'd rather not deal with keys. iOS only for now.
I'd really like feedback from people who have to produce minutes others rely on - engineering, legal, boards, research. What would make you trust an AI-written protocol enough to send it unedited?
Hi Product Hunt, Andro here.
Let me tell you about one thing instead of listing features.
I kept doing the same thing after every meeting. Take the minutes, paste them into an AI, and work from there: turn the discussion into a working document in the business planning or designing digital solution. That habit is the whole reason I wanted the full discussion written down rather than a summary. You cannot plan from highlights.
The problem was everything else sitting in that text. Real meetings are full of other people's names and other companies' business. So every time I ended up cleaning the text by hand before I dared to paste it anywhere, and that is exactly the kind of dull step that makes you quietly stop doing the useful thing.
So last week I added an export made for this. You download the minutes in a version where those details are already replaced with pseudonyms, and it goes straight into whatever model you use. No prep, no second guessing about what is in there. It is masking rather than anonymisation, and I would rather be precise about that than oversell it.
Small feature. It changed how I use the app more than I expected.
I am curious whether others have the same habit. Do you run your meeting notes through an AI afterwards for whatever follow-up reason, and what do you do about the parts that should not leave the room?