What MemoFlow does with this language
| Live captions while recording | Yes, when connected. Meeting minutes bill at 1 Credit / minute on Basic and Plus; Pro includes a monthly allowance. |
|---|---|
| Transcript after recording | Plus and Pro. The whole session is recognised again automatically when it ends; if that cannot finish safely, the live transcript is kept. |
| Speaker labels | Plus and Pro. Numbered speakers where attribution is confident; you can rename them. |
| Live Interpretation from Japanese | Yes — into any of the 33 target languages except Japanese itself. |
| Live Interpretation into Japanese | Yes — from any of the other 29 transcription languages. |
| Meeting summary | Plus and Pro. Drafted from a saved note as a separate result; the transcript is not overwritten. |
| Search | Save the transcript as a note and its words appear in note search. |
| Platforms | iPhone (iOS 16.0+) and Android (8.0 (API 26)). No desktop or web app. |
Before you record
Choose 日本語 before starting. MemoFlow does not switch language during a session, so a meeting held in Japanese with English product names in it should still be recorded as Japanese.
Three things are worth reading against the audio before a Japanese transcript goes into minutes or an e-mail:
- People's and company names, where the same reading can be written with different kanji
- English terms and product names spoken inside Japanese sentences
- Numbers, dates and amounts
文字起こし is not 議事録
A transcript records what was said; minutes record what was decided and what is still open. MemoFlow keeps the two apart. The transcript stays as recorded. On Plus and Pro, the AI meeting write-up drafts decisions, next steps and open questions from a saved note as a separate result, which you check before it becomes the record everyone else reads.
Japanese and English in one conversation
A Live Interpretation session runs in one direction. Choose Japanese as the source and English as the target to follow a Japanese speaker, or the reverse to follow an English one; any of the 33 targets works in place of English. Each line shows the original next to the machine translation, so a phrase that looks wrong can be checked against what was actually said.
Because the session listens for one language, the other side of a two-way conversation is not interpreted in the same session. The person you need to follow decides the direction.
How Japanese results are scored
Japanese is written without spaces between words, so word error rate says little about it. MemoFlow's ASR benchmark method scores Japanese by character error rate. No Japanese result has been published yet, and this page makes no accuracy claim.
Limits to know
- The language is chosen before recording; there is no automatic language detection.
- Live captions need a connection. Offline recording continues without them.
- No accuracy figure is published for any language.
- Machine translation, not a human interpreter.
- Formal speech styles (敬語) and dialects have not been evaluated separately.