How to train any Coqui TTS model on custom dataset, containning the text, speech and its corresponding phoneme?

[subhajyotisaha01]

Hello,

I have a dataset, which contains the text, speech and its corresponding phonemes( or phonemes embedding in numerical format ). I have find out it is straight forward to train a coqui TTS model, if I have only text and speech data, where coqui library itself finds out the phonemes, store it in phoneme_cache path and train a TTS model. But I am unable to find how I will over-write the coqui library, if I have a custom dataset, which already has the phonemes within the dataset file itself.

It will be extremely helpful, if I can find some reference for that.

Apart from that I will also be grateful, if I can get some help to find out how the padding is done in these TTS library.