AudioLDM
AudioLDM copied to clipboard
Higher bit-depth model
I know upscaling is going to be released on Friday (which is OMG), but is it possible to retrain the model with 44.1? Is it feasible? Would a typical GPU run that? I think having a richer depth even in generation might give much clearer and crispier generations.
Retraining with 44.1 is totally feasible. A typical GPU can do this job. Our next stage of research will address the higher sampling rate.