2 ms·
Not Stable Diffusion. MusicLM is a neural codec language model. Basically if GPT was predicting the next audio token while conditioned on text.
by famouswaffles 4y ago
Not Stable Diffusion. MusicLM is a neural codec language model. Basically if GPT was predicting the next audio token while conditioned on text.
- ShamelessC 4y agoSo like DALLE-1? FWIW, the various diffusion models (tend to) use cross attention to an attention-based approach.