Text on screen in about 3.3–3.7 seconds[5]
Median time from speech to a line on screen, per meeting, in tests of the Mac app with Sangei Cloud.
- 3.3–3.7
- seconds, median from speech to text on screen
Live translation, while the meeting is still happening
In short
KaigiFlow listens to your microphone or your Mac's system audio and puts speaker-labelled text on screen as people talk. No bot joins the call. You can follow up to three text outputs at once. See the list of supported languages
Follow the meeting as it happens
Point KaigiFlow at your microphone or your Mac's system audio. Speaker-labelled text appears live as people talk — no bot joins the call.
- Choose your microphone or Mac system audio as the capture source.
- Watch live transcription appear with stable speaker labels.
- Read up to three selected output languages alongside the original.
- Pause and resume without losing the conversation.
- Keep the recording on your Mac the whole time.
A SHARED THREAD
Different languages. One conversation.
An interactive illustration, not an app recording. Names and meeting content are fictional.
Synthetic sample
Moonlark Home APAC launch check-in
太好了，我們這邊也需要繁體中文的產品規格表。
Great. We'll also need the product spec sheet in Traditional Chinese.
Conversation history
Good morning everyone, thanks for joining. Let's do a quick check-in on the Moonlark Home launch.
おはようございます。東京チームです。今週、製品デモの準備が完了しました。
Good morning, this is the Tokyo team. We finished preparing the product demo this week.
早安，台北團隊已經確認了展示會場，可以容納兩百位嘉賓。
Good morning. The Taipei team has confirmed the venue, which holds two hundred guests.
That's great news. Do we have a confirmed date for the Taipei event?
目前預計是十一月十二號，星期四。
It's currently planned for Thursday, November 12.
東京では十一月五日に発表イベントを予定しています。
In Tokyo, we're planning the launch event for November 5.
Perfect, so Tokyo first, then Taipei one week later. I'll update the press calendar.
ありがとうございます。プレスリリースの英語版はいつ頃いただけますか。
Thank you. When can we expect the English version of the press release?
I can send the draft by Friday, so you'll have time to translate it.
What we measured
Our own tests, on public recordings with human transcripts. Lower error rates are better. Each figure carries how it was tested and when.
Up to 3× fewer missed words in English and Mandarin meetings[1]
Words that were spoken but never appeared, compared with standard cloud speech recognition on the same audio. Lower is better.
English meetings
- KaigiFlow
- 11.9%
- Standard cloud speech recognition
- 32.5%
- Times fewer missed words
- 2.7×
Mandarin meetings
- KaigiFlow
- 12.0%
- Standard cloud speech recognition
- 36.3%
- Times fewer missed words
- 3.0×
21% to 36% fewer errors in mixed-language meetings[3]
Conversations that switch between Cantonese, Mandarin, English and Japanese, compared with standard cloud speech recognition. Lower is better.
Cantonese and English
- KaigiFlow
- 19.7%
- Standard cloud speech recognition
- 26.9%
Cantonese and Mandarin
- KaigiFlow
- 10.7%
- Standard cloud speech recognition
- 16.5%
Cantonese, Mandarin and English
- KaigiFlow
- 11.7%
- Standard cloud speech recognition
- 18.2%
Japanese and Mandarin
- KaigiFlow
- 12.6%
- Standard cloud speech recognition
- 15.9%
Tests and dates
- [1] English and Mandarin meeting recordings with human transcripts, compared with standard cloud speech recognition. Tested October 2026. How we measure
- [3] Constructed conversations that switch between two or three languages, built from recordings with human transcripts, compared with standard cloud speech recognition. Claimed only for the mixes named. Tested October 2026. How we measure
- [5] Speed: time from speech to the line on screen, median per meeting, in tests of the real Mac app with Sangei Cloud on public speech recordings. Tested October 2026. Live text is not instant: it arrives a few seconds after each sentence. How we measure
- [9] All tests used public recordings with human transcripts. No customer recordings were used. Results are from KaigiFlow’s own tests in October 2026 and describe those tests, not every meeting. How we measure
Wherever the conversation happens.
In the room. Across the screen. In the audience.
Illustrative scenes. Product demonstrations use real app recordings.
Verified at a glance
No bot
Captures your microphone or Mac system audio instead of joining the call
On your device
On-device: nothing about the meeting leaves the device.
Translation alone is not a shared understanding
Multilingual meetings get handled in four broad ways. Each one fails differently, and the difference is usually about what you are left holding when the call ends.
Captions inside the meeting app
- What it is good at
- Nothing to install, already there when the call starts.
- Where it runs out
- Tied to that one app, and the meeting usually ends with captions rather than a record you can go back and correct.
A meeting bot
- What it is good at
- Captures the call reliably and works the same way every time.
- Where it runs out
- Everyone can see it join. Plenty of clients, employers, and regulated conversations will not allow an extra participant.
A human interpreter
- What it is good at
- The highest-fidelity option when the conversation genuinely matters.
- Where it runs out
- Booked ahead and priced per session — hard to justify for an ordinary recurring call.
KaigiFlow
- What it is good at
- Captures your microphone or your Mac's system audio, so nothing joins the call. Keeps the recording on your device.
- Where it runs out
- On a Mac, it needs Apple silicon, and capturing system audio needs the macOS Screen & System Audio Recording permission.
Compared by how each approach is used, not by vendor.
When you will reach for it
Client meetings
When your customer speaks another language, KaigiFlow shows each sentence with its translation on your Mac or iPhone while the call is still going. Afterwards you have a record you can check.
Learn more
Japanese business meetings
English, Japanese and Chinese can all be spoken in the same meeting. KaigiFlow shows each line with its translations, so you can follow along without waiting for a summary.
Learn more
Interviews and research calls
Follow an interview or research call as it happens, then work from a transcript you can check line by line. KaigiFlow is in beta on Mac, iPhone and iPad.
Learn more
Alongside video calls
KaigiFlow captures your Mac's system audio, so it works alongside call apps such as Zoom, Microsoft Teams and Google Meet. It runs as its own app next to the call, and no bot joins the meeting.
Learn more
Students and lectures
Studying a language, or studying abroad? KaigiFlow shows each sentence of the class as text with its translation on your Mac, iPhone or iPad while the lecture goes on. Afterwards you have a transcript to study from.
Learn more
Languages, verified before we list them
Choose the right language setup for your conversations.
Language coverage depends on the app build.
Questions people ask
Meeting capture, supported languages and what happens to your recordings.
Does KaigiFlow join my Zoom call?
No. KaigiFlow captures your microphone or your Mac’s system audio, so it does not join the meeting as an extra participant. You can use that capture with a meeting running on your Mac. Tell the other participants you are recording and obtain any consent required where you meet.
How is KaigiFlow different from a meeting bot or built-in captions?
KaigiFlow captures audio on your Mac, iPhone or iPad without adding a participant to the call. It gives you speaker-labelled transcription and translation, then lets you review the meeting afterward. Capture uses your microphone or Mac system audio rather than an integration with one meeting service. System audio capture requires the relevant macOS permission.
Which languages does KaigiFlow support?
You can follow up to three text outputs at once. Available meeting languages depend on your app build. See the list of supported languages
Bring clarity to your next multilingual meeting.
See the beta details for Mac, iPhone and iPad — status, compatibility, and install notes for the private beta.
Join the private beta