amabilee
commented
2 weeks ago The mms-tts model may not support direct Chinese characters. Instead, it often requires phonetic transcriptions, such as Pinyin or other romanization systems, particularly for Hakka and Min Nan languages.
For Hakka, you can use the Hakka Romanization System (e.g., Pha̍k-fa-sṳ). For Min Nan, you might consider using Pe̍h-ōe-jī (POJ) or Tâi-lô.
I am using mms-tts for speech synthesis in Hakka or Min Nan languages. I tried it out on the demo page at https://huggingface.co/spaces/mms-meta/MMS. However, when I input text in Chinese characters, the output audio is empty. How should I input the text? Are there any tutorials or tools to convert text into the format required by the model? Thank you.