Image-Based Audio Diffusion

Use image diffusion models that generate spectrogram images from audio embeddings and then convert them back to audio via a decoder adapter.