View Project
● Built a lightweight 1D diff usion model in PyTorch to generate short audio waveforms effi ciently, using a minimal dataset of sine waves and real-world audio clips.
● Designed a tiny U-Net architecture optimized for resource-constrained environments (CPU/low-end GPU).
● Implemented an adaptive noise schedule and effi cient denoising process for high-quality waveform synthesis.
Built with
