4 ms·
Could this be used to train a text -> audio model? I'm thinking of an architecture that uses RVQ. Would RVQ still be necessary?
by cellis 1y ago
Could this be used to train a text -> audio model? I'm thinking of an architecture that uses RVQ. Would RVQ still be necessary?
- diyer22 1y agoI believe DDN is capable of handling TTS (text-to-speech) tasks, because with the text condition, the generation space is significantly reduced. And it's recommended to combine it with an autoregressive model (GPT) for more powerful modeling capabilities.