Not using latest multilingual weights? #56
Description
First of all, thanks for a great repo!
I've been playing around with the API/Frontend a bit and I have noticed some differences between the resulting audio of this project and https://huggingface.co/spaces/ResembleAI/Chatterbox-Multilingual-TTS. Specifically on Swedish the results on numbers and language specific characters å, ä, ö seem to be slightly better on the HF hosted version.
Looking in the uploaded files at https://huggingface.co/ResembleAI/chatterbox/tree/main we can see that there's a fairly recently uploaded v2 version of the model uploaded.
But from what I can tell If I look at the model files within the chatterbox-tts-api-(blackwell) container, when its running, it seems that its in fact the old version that is loaded.
Am I missing something here, or could this be the case? If so, how could I update it to load the newer recent version?
Br