Spectrally-Guided Diffusion Noise Schedules

Carlos Esteves, Ameesh Makadia
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:28285-28304, 2026.

Abstract

Denoising diffusion models are widely used for high-quality image and video generation. Their performance depend on noise schedules, which define the distribution of noise levels applied during training and the sequence of noise levels traversed during sampling. Noise schedules are typically handcrafted and require manual tuning across different resolutions. In this work, we propose a principled way to design per-image noise schedules for pixel diffusion, based on the images spectral properties. By deriving theoretical bounds on how efficacy of minimum and maximum noise levels, we design "tight" noise schedules that eliminate redundant steps. During inference, we propose to conditionally sampled such noise schedules. Experiments show that our noise schedules improve generative quality, particularly at the low-step regime.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-esteves26a, title = {Spectrally-Guided Diffusion Noise Schedules}, author = {Esteves, Carlos and Makadia, Ameesh}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {28285--28304}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/esteves26a/esteves26a.pdf}, url = {https://proceedings.mlr.press/v306/esteves26a.html}, abstract = {Denoising diffusion models are widely used for high-quality image and video generation. Their performance depend on noise schedules, which define the distribution of noise levels applied during training and the sequence of noise levels traversed during sampling. Noise schedules are typically handcrafted and require manual tuning across different resolutions. In this work, we propose a principled way to design per-image noise schedules for pixel diffusion, based on the images spectral properties. By deriving theoretical bounds on how efficacy of minimum and maximum noise levels, we design "tight" noise schedules that eliminate redundant steps. During inference, we propose to conditionally sampled such noise schedules. Experiments show that our noise schedules improve generative quality, particularly at the low-step regime.} }
Endnote
%0 Conference Paper %T Spectrally-Guided Diffusion Noise Schedules %A Carlos Esteves %A Ameesh Makadia %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-esteves26a %I PMLR %P 28285--28304 %U https://proceedings.mlr.press/v306/esteves26a.html %V 306 %X Denoising diffusion models are widely used for high-quality image and video generation. Their performance depend on noise schedules, which define the distribution of noise levels applied during training and the sequence of noise levels traversed during sampling. Noise schedules are typically handcrafted and require manual tuning across different resolutions. In this work, we propose a principled way to design per-image noise schedules for pixel diffusion, based on the images spectral properties. By deriving theoretical bounds on how efficacy of minimum and maximum noise levels, we design "tight" noise schedules that eliminate redundant steps. During inference, we propose to conditionally sampled such noise schedules. Experiments show that our noise schedules improve generative quality, particularly at the low-step regime.
APA
Esteves, C. & Makadia, A.. (2026). Spectrally-Guided Diffusion Noise Schedules. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:28285-28304 Available from https://proceedings.mlr.press/v306/esteves26a.html.

Related Material