Cantonese translation for real meetings
Meetings in Hong Kong and Guangdong rarely stay in one language. Live Translate transcribes mixed Cantonese, English and Mandarin as it is spoken and puts bilingual captions on screen while the meeting is still running.
Ordinary translation
Cantonese is usually just converted to simplified characters, or recognised as Mandarin with a fallback — spoken particles get dropped.
Live Translate
Native Cantonese recognition; spoken phrasing is normalised into clean written form before translation, not translated word-by-word.
No switching mid-sentence
Auto-detect handles Cantonese, English and Mandarin inside one sentence. Nobody has to pause the meeting to change a dropdown.
Colloquial Cantonese, not character-by-character
Spoken Cantonese is normalised into natural written form first, so everyday words come out as meaning rather than literal characters.
Your names, spelled your way
A glossary binds product, client and system names to the meeting so they survive both transcription and translation.
Searchable afterwards
Timestamped bilingual transcript plus an auto summary — the meeting becomes a document, not a memory.
How it works
- 1Start a meeting and leave the source language on auto-detect.
- 2Optionally bind a glossary so product and client names come out right.
- 3Speak. Bilingual captions appear as you go; the summary is ready when you finish.
Questions about Cantonese
Do I have to switch languages when people mix Cantonese and English?
No. Set the source to auto-detect and mixed Cantonese / English / Mandarin in a single sentence is transcribed continuously — which is how meetings in Hong Kong and Guangdong actually sound.
Can it handle colloquial Cantonese?
Spoken Cantonese differs a lot from written Chinese, and word-for-word translation breaks on everyday terms. Cantonese goes through a normalisation step that turns speech into natural written form before translating, instead of mapping character by character.
Our product names keep coming out wrong.
Use the glossary. Add product, client and system names up front and bind it to the meeting — both transcription and translation then follow your spelling. One-off names can go into the temporary-terms box when you start a meeting.
Does a noisy room matter?
Yes. Accuracy is noticeably better in a quiet room; use an external or headset mic and keep the device close to whoever is speaking. Silence is gated on the client so quiet passages are not turned into phantom text.
Can I review the meeting afterwards?
Yes. Timestamped bilingual captions stay in your account with an auto-generated summary, so you can search back to who said what and when.