Word-for-word Cantonese/English meeting transcription — fillers, stutters and all.
Tick the consent box above to enable uploading
Terms guide transcription and are saved with the recording for reuse on the next upload.
Telling us the headcount keeps the speaker split accurate. Leave blank to let the system estimate - it tends to over-split without it.
Record 30-60 seconds of your own voice and new transcripts will label your diarized speaker as 「本人」 automatically - shown as an AI match you confirm, never applied silently. You can only enrol your own voice. The recording itself is NOT kept: the system stores only a voiceprint (it cannot be turned back into audio), and you can delete it at any time.
Read this aloud in your normal speaking voice:
I am reading this passage for about one minute so the system can capture the sound of my voice. In meetings I usually speak the way I am speaking now, sometimes mixing in terms like without prejudice, discovery and filing. This recording itself will not be stored; only a voiceprint is kept, and I can delete it whenever I choose. I will keep reading a little longer to make sure the sample is at least thirty seconds. Thank you.