PulseMulti-language support
Handling meetings that mix Japanese and English, choosing the output language, and the list of supported languages.
You can set which language transcription and summaries come out in, for Japanese meetings, English meetings, and meetings that mix the two. This article concentrates on the decisions that actually arise in a Japanese setting, covering how to choose and where the limits are.
There are two settings. Both are changed in the desktop app under Settings > Preferences > Language.
- Transcription language: which language what is spoken in the meeting is recognised as
- Summary language: which language the note generated after the meeting is written in
Menus for choosing the transcription language and the summary language, one above the other.
The app's interface — menus, buttons, settings — is currently in English only. Transcripts and summaries can be produced in Japanese, but the interface itself has not been localised. This article gives the Japanese alongside the English items on screen.
Settings by kind of meeting
| What the meeting is like | Transcription language | Summary language |
|---|---|---|
| Everyone speaks Japanese | Multi-language | Auto |
| Japanese with English words and proper nouns mixed in | Multi-language | Auto |
| Japanese and English speakers alternating | Multi-language | Auto or English |
| An English-only meeting, but you want the minutes in Japanese | Multi-language | Auto produces an English summary, so use chat to turn it into Japanese |
| An English-only meeting, and English minutes are fine | English | English |
Choosing “English” as the transcription language gives the highest quality for English but does not expect Japanese speech. To record a Japanese meeting, choose “Multi-language”.
Transcription language
There are two options in the desktop app.
- English: gives the highest quality where most of the meeting is in English.
- Multi-language: handles all the languages listed later in this article. Where several languages are spoken within one meeting, it switches between them as it transcribes.
In a discussion that mixes Japanese and English — someone at an overseas office explaining the position in English while the Japanese side confirms in Japanese — choosing “Multi-language” puts both into the same transcript.
This setting applies only to meetings recorded after you change it. A transcript already recorded cannot be translated into another language afterwards. To read what was said in a past meeting in Japanese, ask chat to translate that passage.
Summary language
- English: the note is generated in English regardless of the language spoken in the meeting.
- Auto: the note is generated in whichever language was spoken most in the meeting.
For a Japanese note from a Japanese meeting, choose “Auto”. In a meeting that is roughly half Japanese and half English, the output language may vary from meeting to meeting. To fix the output, either choose “English” or read the generated note and correct what you need.
To change the language of a past note, change this setting, open the note and run regenerate.
Switching language mid-meeting
You can also change language after a meeting has started. Reselect the language from the meeting controls on the recording screen.
The current language and a selection menu among the controls shown on the note screen while recording.
Just after switching, transcription may pause briefly while the new language's recognition model loads. In meetings where the languages divide clearly — Japanese in the first half, English in the second — timing the switch to a break between agenda items makes any gap less noticeable.
A language changed mid-meeting is also saved as the default for later meetings. To make it a temporary change for that meeting only, set it back afterwards under Settings > Preferences > Language.
Improving accuracy in Japanese meetings
With the language set correctly, keeping the following in mind means fewer corrections when you read the transcript back.
- Register proper nouns in advance: company names, department names, product names and external contacts can be registered under Customising transcription. In Japanese, the right sound written with the wrong characters is a common error, and registration makes a noticeable difference. Note that registration makes a word more likely to be recognised; it does not standardise variant spellings such as “サーバー” and “サーバ” onto one form. Where your company has settled on a spelling, replace it after exporting.
- The transcript keeps what was actually said: back-channel responses, restarts and the exchange of pleasantries are all recorded as spoken. Where you need something readable, share the enhanced note rather than the transcript.
- Output is horizontal: even where your internal forms and minutes use vertical writing, transcripts and summaries are produced as horizontal text. If you need vertical layout, set it in your word processor after exporting.
- When people talk over each other: recognition accuracy drops when several participants speak at once. For attributing speech to each speaker, see Speaker labels.
Supported languages
The desktop app (macOS / Windows) supports 31 languages. These are the languages the transcription provider currently supports.
| Region | Languages supported |
|---|---|
| East and South-East Asia | Japanese, Chinese (Mandarin), Chinese (Cantonese), Korean, Vietnamese |
| South Asia | Hindi, Marathi, Urdu |
| Middle East and West Asia | Arabic, Hebrew, Persian, Turkish |
| Africa | Afrikaans, Xhosa, Zulu |
| Europe | English, French, German, Italian, Spanish, Portuguese, Dutch, Catalan, Galician, Danish, Swedish, Norwegian, Finnish, Estonian, Romanian, Russian |
If you need a language not on this list, contact tharada@the-unchain.com. We will let you know if the transcription provider adds support.