Discriminator-Guided Cooperative Diffusion for Joint Audio and Video Generation

Akio Hayakawa, Masato Ishii, Takashi Shibuya, Yuki Mitsufuji

Keywords: Diffusion models, Multi-modal data, Audio-visual generative models

TL;DR: We construct an audio-video joint generative model with a low computational cost using frozen pre-trained diffusion models for each modality, computing a joint guidance through the gradient of a discriminator for real video-audio pairs and fake ones.

Abstract:

In this study, we aim to construct an audio-video generative model with minimal computational cost by leveraging pre-trained single-modal generative models for audio and video.
To achieve this, we propose a novel method that guides single-modal models to cooperatively generate well-aligned samples across modalities.
Specifically, given two pre-trained base diffusion models, we train a lightweight joint guidance module to adjust scores separately estimated by the base models to match the score of joint distribution over audio and video.
We show that this guidance can be computed through the gradient of the optimal discriminator distinguishing real audio-video pairs from the fake ones independently generated by the base models.
On the basis of this analysis, we construct a joint guidance module by training this discriminator.
Additionally, we adopt a loss function to make the gradient of the discriminator work as a noise estimator, as in standard diffusion models, stabilizing the gradient of the discriminator.
Empirical evaluations on several benchmark datasets demonstrate that our method improves both single-modal fidelity and multi-modal alignment with a relatively small number of parameters.

Primary Area: generative models

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 5876