Skip to content

A different approach #6

@RustoMCSpit

Description

@RustoMCSpit

You'd honestly be far better getting MIDI drum data from real songs (easy to tag genres too) and then using a generator to use a variety of very good MIDI drum VST to randomly go over them all then randomly generating drum samples.

With all the below done as well, you'd be in a far better place.
Separated stems #3
Long progressions with large context #4 (This would automatically be achieved by the above)
No one shots #5

I've listened to a few of the audios in this dataset and honestly: garbage in, garbage out. This is not a dataset any AI should be trained on and clearly using algorithms to randomly generate four bar loops is not the way to go about this.

I do not believe most songs have licensing rights to their drum parts as drum parts are so repetitive by nature, it's like how no one owns a chord progression. Then again, that mightn't be true. Something to think about.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions