Abstract
Joint audio--video (AV) generators are trained on data in which \emph{what an event looks like} and \emph{what it sounds like} are spuriously correlated. We present a \emph{controlled causal study} of the resulting failure mode. In an AV structural causal model where the audio is, by construction, independent of the video's nuisance appearance, models that let audio read video directly---through cross-attention or a shared latent---learn a \emph{visual shortcut}: they predict sound from appearance rather than the causal event and, when the appearance--event correlation is broken at test time, synthesize the wrong event's sound. Crucially, the popular remedy of routing both modalities through a \emph{shared common-cause latent} does \emph{not} fix this---a bottleneck, an unsupervised shared/private factorization, and a faithful shared-prior model all grab the appearance proxy and fail like the direct model. Blocking the shortcut instead requires an \emph{intervention on the nuisance}: under the stated assumptions we prove that counterfactual invariance is necessary and sufficient to identify the causal predictor, and we verify the mechanism from feature-vector SCMs to procedural pixel video, real images with spectrogram audio, moving real digits, and a conditional generator. On a \emph{real, pretrained} V2A generator (MMAudio), an input-intervention test shows the model is far from invariant to sound-irrelevant edits, though a generic-noise control reveals it is broadly input-brittle rather than specifically colour-shortcutting---clean isolation of the shortcut needs the controlled confounds our synthetic studies provide. We characterize \emph{when} the shortcut occurs, compare the objective against supervised counterfactual augmentation, and isolate the \emph{unknown-nuisance} regime---where the intervention cannot be applied---as the central open problem.
Explore similar work
May 28, 2026cs.CV
Generative video-to-audio (V2A) models produce highly plausible soundtracks, but it remains unclear whether they capture the underlying physical processes. Existing evaluations emphasize perceptual realism and overlook physical correctness under controlled interventions. In this paper, we introduce FlatSounds, a benchmark that audits the physical reasoning of V2A models through: 1) controlled counterfactual pairs in which a single physical factor is varied, and 2) single-video pattern tests that probe internal consistency and directional trends. These settings test whether the generated audio correctly reflects specific physical properties and timings. Our evaluation of state-of-the-art models reveals a consistent trade-off: models rely more on text captions than the visual stream to infer physics and semantics. Captions generally improve physical and semantic accuracy, but paradoxically degrade temporal alignment. Our results highlight the need to move beyond audio quality toward learning physical processes directly from pixels. Finally, we find that our physics-based metrics correlate strongly with human preference tests on our own data. Project webpage: https://research.nvidia.com/labs/cosmos-lab/flatsounds/
Tingle Li, Siddharth Gururani, Kevin J. Shih +6
Jun 26, 2025cs.CV
We propose a step-by-step video-to-audio (V2A) generation method that provides finer control over the generation process and more realistic audio synthesis. Inspired by traditional Foley workflows, our approach enables incremental generation of complementary sounds, allowing users to author multiple sound events induced by a video. To avoid the need for costly multi-reference video-audio datasets, each generation step is formulated as a negatively guided V2A process that discourages duplication of sounds already present in previously generated tracks. The guidance model is trained by finetuning a pre-trained V2A model on audio pairs from non-overlapping segments of the same video, encouraging it to leverage acoustic context while remaining visually grounded, and enabling training with standard single-reference audiovisual datasets. Objective and subjective evaluations demonstrate that our method enhances the separability of generated sounds at each step and improves the overall quality of the final composite audio, outperforming existing baselines. Our project page is available at: https://ahykw.github.io/sbsv2a/.
Akio Hayakawa, Masato Ishii, Takashi Shibuya +1
May 8, 2026cs.SD
Joint audio-video generation models are rapidly approaching professional production quality, raising a central question: do they understand audio-visual physics, or merely generate plausible sounds and frames that violate real-world consistency? We introduce AV-Phys Bench, a benchmark for evaluating physical commonsense in joint audio-video generation. AV-Phys Bench tests models across three scene categories: Steady State, Event Transition, and Environment Transition. It covers physics-grounded subcategories drawn from real-world scenes, plus Anti-AV-Physics prompts that deliberately request physically inconsistent audio-video behavior. Each generation is evaluated along five dimensions: visual semantic adherence, audio semantic adherence, visual physical commonsense, audio physical commonsense, and cross-modal physical commonsense. Across three proprietary and four open-source models, we find that Seedance 2.0 performs best overall, but all models remain far from robust physical understanding. Performance drops sharply on event-driven and environment-driven transitions, and even strong proprietary systems collapse on Anti-AV-Physics prompts. We further introduce AV-Phys Agent, a ReAct-style evaluator that combines a multimodal language model with deterministic acoustic measurement tools, producing rankings that closely align with human ratings. Our results identify cross-modal physical consistency and transition-driven scene dynamics as key open challenges for joint audio-video generation.
Zijun Cui, Xiulong Liu, Hao Fang +8