What will SunoAI Foundation Model be trained on?

[hu-po]

It seems like the biggest hurdle for SunoAI is the fact that the Encoder/Decoder is a non-commerical license from Facebook. My guess is the team wants to train their own foundational AI model from scratch. What dataset will you use? For context, this is the dataset according to the facebook paper:

[hu-po]

Perhaps https://www.robots.ox.ac.uk/~vgg/data/voxceleb/

It is CC BY 4.0

[gkucsko]

a continuously growing mix of proprietary and open datasets :)

[vivasvan1]

@gkucsko Could you provide the list of open datasets you use?