Question about multilingual support #2

According to the paper, it was trained with only 6 languages.
When I tried to generate with another language (like Vietnamese), the mouth movement is different with the original input but it can't match the actual sentence.
I noticed the project is using Whisper so it should support multiple languages, maybe the limit comes from wav2vec2? is it possible to support other languages?