chunked/nltk_chunking/model_doc_jukebox/chunk_4.txt

We show that the combined model at scale can generate high-fidelity and diverse songs with coherence up to multiple minutes.