Tokenizer Replacement

[humair-m]

Hey @ysharma3501 ,

Is it practical to replace the original tokenizer with new one like if we replace tokenizer with phenomes

[Alukarho]

I had been having problems training it in Spanish because, when it encountered unknown words, it would pronounce them with strange noises. It searched its knowledge, but it wasn’t able to construct the word. I tried several methods, but I couldn’t properly train it in Spanish to cover any word. Then I came up with the idea of training using phonemes; the tokens for that were already in the tokenizer, and it worked very well. There are still things to polish, but it already speaks excellently, much better than before.

[Satish-101]

@Alukarho I’m training an Indic language, mainly Punjabi, and my MiraTTS model hasn’t learned to produce intelligible speech, it’s generating gibberish/noise. Could you share your approach or pointers on how you replaced the original tokenizer with a phoneme-based one?