Time of music

[Sodayeyeye]

There is a question about the time to generate music, can only generate 10s at most?

[dorienh]

Until Mustango 2.0 comes out that is the limit indeed.

[kkaiwwana]

hi @Sodayeyeye, to generate longer music under current setup is actually pretty doable. Tranfering a super simple "copy & paste" trick that I introduced in this paper (wasn't for audio gen tho) can generate arbitrary length audio in this case. And more importantly, it's training-free! I implemented it for mustango to generate arbitrary long music, which has a varying style while remains favorable temporal consistency. The implementation can be found in this forked repo with >1min demo audio provided (I personaly think that sounds not bad).

hi @dorienh, would you like to take a look? And if you found it interseting, I'm happy to send a PR as a plug-and-play enhancement to the current verison of mustango.

[dorienh]

@kkaiwwana This is interesting, thanks for sharing. Do you happen to have any output examples. Will look into your paper.

[kkaiwwana]

@dorienh yes, please check this forked repo, I provided many generated examples.