Custom Augmentations?

I’m working with music representations for vision models, and in this area there are possibilities for augmentation techniques that aren’t implemented in the fastai library (for good reason!). I’m just wondering how I would go about adding custom augmentations?