FlowLong is presented, a training-free, architecture-agnostic, inference-time framework that harmonizes overlapping windows sampled in parallel via Tweedie matching, a manifold-constrained blend of predicted clean samples on overlaps, and stochastic early-phase sampling that yields arbitrarily long, phase-locked audio-video.
Autoregressive video generation requires denoising the current frames while writing their key-value representations as context for future predictions. However, these two roles typically share parameters, and we find that their gradients exhibit distinct patterns and systematic negative alignment, hindering the joint op...
Zi-Han Su, Junhao Zhuang, Yao-Wei Li et al.· 1 citation
Autoregressive video models can generate minute-long videos in real time, but they produce generic subjects from text rather than specific subjects from user-provided images. Existing customization methods either require costly per-subject optimization or use pretrained conditioning networks that jointly process all vi...
Yunseung Ok, Hyunsoo Kim, Minseo Kim et al.· 0 citations
Natively training joint video-audio generation models at higher resolutions empowers them to learn richer visual details and sharper motion dynamics. However, full attention incurs quadratic cost and, as resolution increases, spreads attention over increasingly redundant tokens, diluting learning signals for informativ...
Shu-Yuan Tu, Qi Tian, Yin-Ming Huang et al.· 1 citation
This framework minimizes maximum mean discrepancy in frozen self-supervised video representation spaces in frozen self-supervised video representation spaces, using a hybrid Nystr\"om--Monte Carlo estimator to balance approximation bias and sampling variance.
Chi Zhang, Yue-Yi Liu, Hao-Yan Shi et al.· 3 citations· ⚡2
The decoupled self-forcing distillation resolves both models under one frozen teacher: conditioned on motion, it distills the renderer into a block-causal student; unconditionally, it scores rendered rollouts against real videos, supervising motion by the video it produces.
Yan-Ru An, Rui-Yan Wang, Wen-Wu Wei et al.· 0 citations
This work reformulates the video diffusion sampling as a frame-indexed stochastic process over noise levels, and constructs a continuous training trajectory along which the sampling schedule progressively evolves from independent sampling to inference-consistent sampling.
Yue-Ting Zhu, Yue-Hao Song, Kai-Chen Zhang et al.· 2 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
