Our model handles Cantonese-to-written-Chinese well, but Cantonese grammar is flexible, and in some passages the AI mixes written and spoken register in the same track.
You do not need to fix this line by line. Use the editor's own subtitle translation to convert the whole track into a clean written or spoken version.
Converting a track to pure written or spoken Chinese
Open the subtitle editor.
Click Add translation (shown as Translate on mobile).
The window that opens uses the current subtitle track as the source — usually Cantonese, or whichever language matches the video. If it is not the one you want to convert, switch to that track first.
Under Target language, choose the output you want, for example Traditional Cantonese (Written) or Traditional Cantonese (Spoken). The list does not show the same version as the source track or versions you have already translated, so if the one you want is missing, this method cannot convert it.
Click Start translation.
A new subtitle track is added when processing finishes. That is your corrected version.
Worth knowing
The language you pick at upload matters most. When you upload, set the spoken language to match the audio. If the video is mainly Cantonese, choose Cantonese — a wrong source language lowers transcription accuracy before any of this starts.
Timing. Translation usually finishes within a few minutes. If no new track has appeared five minutes after you click Start translation, refresh the page.
This works for other languages too — the same route translates a track into English, Japanese, Korean and 80+ others.