Published in Applied Sciences, Volume 14, Issue 20, Article 9595, 2024
We propose CycleDiffusion, a voice conversion model based on cycle-consistent diffusion to improve quality and consistency.
Recommended citation: Dongsuk Yook, Geonhee Han, Hyung-Pil Chang, In-Chul Yoo. (2024). "CycleDiffusion: Voice Conversion Using Cycle-Consistent Diffusion Models." Applied Sciences, 14(20), 9595.
Download Paper
Published in Interspeech 2025, 2025
We propose DGMO, a training-free diffusion-guided mask optimization method for audio source separation.
Recommended citation: Geonyoung Lee*, Geonhee Han*, Paul Hongsuck Seo. (2025). "DGMO: Training-Free Audio Source Separation through Diffusion-Guided Mask Optimization." Interspeech 2025. (* equal contribution)
Download Paper
Published in CVPR 2025, 2025
A distillation-based framework for compressing diffusion models using randomly conditioned generation.
Recommended citation: Dohyun Kim*, Sehwan Park*, Geonhee Han, Seung Wook Kim, Paul Hongsuck Seo. (2025). "Random Conditioning with Distillation for Data-Efficient Diffusion Model Compression." CVPR 2025.
Download Paper