From what I understand, the easiest way to fix this would be to reset the modal at each new use.

However, this could be seen as a regression for users who want to add the same picture multiple times (and can't copy/paste for some reason).
Another way of solving this would be to 'hide' the upload action by putting it in together with the file browsing or the step submission. However I think this change is much more important and would need to be discussed because it can influence a lot more use cases.

I'll go with the first solution since it seems like the best for the end users.