Does it support mixed Chinese and English streaming mode?

[yourengod]

trafficstars

[C-Loftus]

Don't think it does. I believe you could train the model if you had a mixed dataset but for this it would probably be easier to use something like lingua-go / lingua-rs to detect the language and then forward it to a different model. I have a tracker for this in a separate project https://github.com/C-Loftus/QuickPiperAudiobook/issues/18