Spectrogram Diffusion Pipeline

Use a two-step audio generation pipeline by generating spectrogram images via diffusion models and then applying a vocoder to synthesize the final audio waveform.