I've found ChatGPT 4o (the voice one) to be useful, I just need to prompt it first by saying `Whatever you hear in Chinese, say in English, whatever you hear in English, say in Chinese`. It refuses to cooperate if you tell it to be a live translator.

Anyways, it seems to understand context really well, I tried it out with a Chinese person and they said it worked really well.

It even works for Croatian which is awesome as it is a really small language that anything voice-related usually does not work with.

Learn how to point at something and ask "how do you say?" This will get you way further.