OdiMeet — Every voice recorder deserves a transcript. On your iPhone.

A recorder you already own. A transcript you actually keep.

Recorder Mode

That iPhone SE in the drawer becomes the meeting recorder: 48 kHz WAV, keeps recording with the screen locked, survives calls and interruptions. Buy nothing.

Stealth mode

Screen goes black at minimum brightness with a slow white breathing light. Looks asleep; is recording. Tap to see the timer and controls, and they fade away again.

Import from any recorder

Sony, Olympus, Tascam, Zoom, wireless-mic transmitters, Voice Memos, files exported from any app. Plug the recorder into USB-C and pick the file in Files.

Two iPhones — who said what

One in front of each speaker, paired over Wi‑Fi and Bluetooth, recording in sync. Whichever phone is louder is the person talking — 12–17 dB of separation in our tests — and the transcript is labelled automatically, renamable after merging. Nothing to buy.

Wireless lavalier mic sets

DJI Mic Mini and RØDE Wireless Micro tested and certified. Each transmitter becomes one speaker's track; plug the receiver into the iPhone and the only setting is Stereo / Split in the maker's app.

Speaker separation on one iPhone (estimated)

With a single phone, optional on-device voice-print separation labels the transcript. It is an estimate — similar voices and cross-talk confuse it — and the interface says so. For accuracy, use either setup above.

Two on-device engines

Apple's iOS 26 engine for live text and English; SenseVoice (Alibaba FunASR, downloaded once) for the most accurate Mandarin, Cantonese, Japanese and Korean. Auto-detect or choose the language.

Live text

Watch the transcript appear as people speak. Swipe once for full-screen text, swipe again for the summary.

Live summary, your key

Every 30 seconds, running notes; every 5 minutes the summary is rewritten from the raw transcript. Powered by your own Gemini key — transcript text goes straight to Google, audio never does. On by default once you add a key; turn off in Settings.

Final summary, optionally fact-checked

When the full-quality transcript is done, one more call produces the summary of record. Turn on Google Search grounding and questionable claims get checked and marked as corrected.

Yours to keep

Copy any tab in one tap. Export TXT, Markdown, SRT, VTT or JSON. Recordings sit in the Files app as plain WAV — no lock-in, no proprietary format.

1.1 · live

Who said what — from physics, not guesswork.

Every AI note-taker guesses speakers from voice prints, and every one of them admits it is unreliable. OdiMeet is taking a different route: every old iPhone is a microphone. Put one in front of each participant, let each record, and they merge automatically when you stop.

Each phone hears its own speaker 10–20 dB louder than anyone else (median 12–14 dB in our real tabletop recordings). For every fraction of a second, the loudest track is the person talking. Attribution becomes a measurement, not a prediction — and when two people talk over each other, both are kept.

No hardware to buy. The same trick works with multi-transmitter wireless mics such as the Insta360 Mic Pro or DJI Mic.

Alice

−18 dB

Bob

speaking

Chen

−16 dB

Loudest phone per 200 ms window → the speaker. Near-100% attribution, crosstalk preserved.

1.1 · live

An iPhone array, recording in sync.

Interviewing someone? Put one iPhone in front of each person. OdiMeet pairs them over Wi-Fi and Bluetooth — no network, no account — starts every phone recording at the same instant, and keeps their clocks aligned.

Stop, and the tracks come back to one phone and merge automatically into a single conversation, already labelled by who sat where. You know exactly who said what, because each voice had its own microphone.

Two phones for a one-on-one, more for a round table. The recorder you already own, turned into a microphone array.

You3REC0:00

pairing…

Guest3REC0:00

Paired peer-to-peer · one countdown · both recording in sync.

FAQ

How is this different from Plaud, Otter or the other AI note-takers?

Four ways. Hardware: they sell you a recorder; OdiMeet uses the iPhone you already have, the voice recorder in your drawer, or the wireless mics you bought for video. Where the audio goes: they upload it to their cloud and meter it by the minute; OdiMeet transcribes on the phone with no account and no meter. Whose AI: they pick the model and bill you monthly; OdiMeet uses your own Gemini key, so you pay Google's actual price and see every request. Who said what: they guess from voice-prints; OdiMeet measures it, by distance or by separate tracks.

Does anything leave my phone?

Audio never does. Transcript text leaves only if you turn on Gemini summaries — and then it goes directly from your phone to Google with your key; there is no OdiMeet server in between. If you also enable Google Search fact-checking, specific claims in the transcript become search queries. Live summaries are on once you add a key and can be turned off in Settings; Search fact-checking is off by default.

Which iPhones work?

Any iPhone that runs iOS 26 — from the iPhone SE (2nd generation) and iPhone 11 up. Apple's engine is comfortable on all of them, and the optional SenseVoice model is fast even on the older ones. A recent iPhone with an OLED screen also makes stealth mode truly black.

What is the lowest-end iPhone you have tested on?

An iPhone SE (2nd generation) with 64 GB, running iOS 26. It is one of our everyday test devices — recording, on-device transcription and the two-phone sync recording all run on it.

What about meetings that mix Chinese and English?

Set the transcription language to Chinese and SenseVoice keeps English words in English inside the Chinese sentence — the right setting for Taiwan-style mixed meetings. Auto-detect listens to the first 30 seconds and picks one language, which suits imported recordings whose language you do not know.

Does it identify speakers?

Yes — and not by guessing. Two iPhones: one in front of each speaker, recording in sync; whichever phone is louder is the person talking, 12–17 dB of separation in our tests, and the transcript is labelled automatically. Wireless lavalier sets (DJI Mic Mini and RØDE Wireless Micro certified): each transmitter is one speaker's track — the most accurate. A single phone can use optional on-device voice-print separation; it is an estimate, confused by similar voices and cross-talk, and the interface says so. In all three you can rename speakers after merging.

Do I still need an external microphone, like a DJI Mic?

For most meetings, no. An iPhone on the table within about half a metre of its speaker records clearly, and a phone per participant is enough for speaker attribution. An external wireless mic earns its place in three situations: noisy rooms or long tables (a lavalier at the collar is 15–20 dB closer to the voice than a phone on the table, roughly twice the separation); when attribution must be near-perfect, as in depositions; and when phones can't sit on a table — walking interviews, field work. If you already own a set, plug its receiver into the iPhone and it works — the ones we have tested and certify are the DJI Mic Mini and the RØDE Wireless Micro (see the next question).

Does it support wireless lavalier mics?

Yes. We have tested the DJI Mic Mini and the RØDE Wireless Micro; both work perfectly: plug the receiver into the iPhone, set the channel mode to Stereo / Split in the maker's app (the only setting you need to touch), and pick "One iPhone + wireless mic set" in New recording — each transmitter becomes one speaker's track, and who said what goes straight into the transcript. Details: DJI Mic Mini, RØDE Wireless Micro. Other receivers that work the same way should too, but we only call a set "certified" once we have tested it.

Are summaries Gemini-only?

Gemini comes first because it is the easiest key to get and the cheapest to run. Amazon Bedrock, OpenAI, Anthropic and custom endpoints are coming — any OpenAI-compatible API will work, including a local model on your own computer such as Ollama. With a local model, transcript text never reaches Google or anyone else.

What does it cost?

90 days free, then $2.99 a month, $29.99 a year, or $39.99 once for lifetime — see Plans. Every plan includes everything; subscribe only for the months you need. Gemini usage is paid directly to Google with your own key, typically well under a dollar for a two-hour meeting with live summaries on.

When can I get it?

Now: download OdiMeet on the App Store, with a 90-day free trial. Requires iOS 26.

Your spare iPhone is now an AI voice recorder.
Your audio never leaves it.