Open Jimmy1456 opened 11 months ago
HI Jimmy! We do intend to make instructions on how to train our expressive models... But as for now, I can only tell that it is the same scheme of the original Coqui TTS repository (at the time we forked it) :by defining a config.json (you can check some available on the recipes) and then running the TTS/bin/train_tts. The only add-on here for running a model is that you have to also define inside the config.json a style encoder config (otherwise it will run the default configuration in TTS/style_encoder/configs/style_encoder_config.json)... In the following weeks, I'll try to release a recipe for training our models on the EMOVDB dataset and i'll update you this issue! :)
Thank you for your great work! Can you provide a more detailed guideline to train a model?