🚩 Report : Legal issue(s)
Reasons both ethical and legal. The model card for this project does not mention at all what dataset was used to do the fine-tuning or where the data came from. The model itself seems to be able to produce music spectrograms from a number of popular genres, and responds well to mentioning specific high profile artists (e.g. Beyonce, Mariah Carey, Daft Punk). It seems very likely that this model was created using copyrighted music and/or music generally not authorized for use in this way. Requesting clarity and transparency from the authors.
it can't even sing well, and you have no proof that they even did train it on copyrighted music because you don't have the dataset.
Additional fine tuning and data information has been added to the model-card. This was trained using approaches similar to hugging face examples, but fine-tuning can be achieved with very small datasets using a dreambooth approach.