数据集:

teticio/audio-diffusion-256

中文

De-noising Diffusion Probabilistic Model trained on teticio/audio-diffusion-256 to generate mel spectrograms of 256x256 corresponding to 5 seconds of audio. The code to convert from audio to spectrogram and vice versa can be found in https://github.com/teticio/audio-diffusion along with scripts to train and run inference.