Video Diffusion Models

Causal video generators must predict from the past, but they need not learn only from it. In streaming autoregressive video diffusion, each emitted segment becomes a commitment that future segments must preserve. Standard training, however, only asks each causal state to explain the present. This creates what we call a representation-level planning gap: states that fit the current segment may discard identity, layout, and motion information needed for a consistent future. We introduce Video-Mirai, a training-only method that closes this gap without changing causal inference: the generator rolls out causally, a frozen foresight encoder reads the completed rollout non-causally, and a lightweight predictor distills the resulting stopped-gradient targets into causal states. Future frames supervise representations, never generator inputs. At inference, the encoder and predictor are discarded, leaving the original architecture, per-step FLOPs, and KV-cache behavior unchanged. Video-Mirai improves a strong Causal-Forcing baseline on 5-second VBench from 83.8 to 84.6 in terms of Total Score. On 30-second rollouts beyond the training horizon, subject consistency improves from 84.9 to 88.5 and background consistency from 90.2 to 91.9. Ablations identify future-conditioned targets as the key ingredient, and probes show that future frames become more decodable from current features. Causality should constrain inference, not representation supervision. Our study highlights that visual autoregressive models need foresight. Project page: https://y0uroy.github.io/Video-Mirai.

Recent world action models leverage video foundation models by aligning broad visual-dynamics priors with executable robot actions. We revisit this alignment from a distributional perspective. Existing formulations typically narrow the aligned prior into an observation-conditioned policy distribution over future actions. In contrast, we keep the distribution broader by modeling the joint space of interaction videos and executable hand trajectories under multiple conditioning regimes. We propose Donk, a unified video-action denoising model for dexterous hands. With language, an initial image, and the initial hand state, Donk samples future videos and bimanual MANO trajectories as an action policy. Without the image condition, the same denoising architecture samples paired video-action rollouts from a text-conditioned distribution, turning the aligned video prior into a data engine. Across action, video, and text-only generation evaluations, Donk improves dexterous trajectory accuracy, preserves strong video fidelity, and produces smooth text-conditioned action rollouts under the same unified training recipe.

Audio-driven human motion video generation aims to synthesize realistic and temporally coherent human animations from a single static image, with applications in talking-head synthesis, co-speech gesture generation, and dynamic presentations. Moving beyond conventional keypoint-based methods that often struggle to capture subtle motion dynamics, We propose a novel implicit-motion framework for generating realistic and temporally coherent human motion videos from a single static image and audio. Our approach uses a two-stage pipeline that decouples motion prediction from rendering. The first stage integrates appearance priors and hierarchical depth cues into a region-aware attention mechanism to model latent motion features. The second stage employs a Mamba-enhanced diffusion model to directly predict these features from audio and the source image, enabling unsupervised learning of fine-grained motion patterns. This decoupled architecture enhances flexibility and efficiency. Trained on a new 380-hour high-quality dataset, our method outperforms prior work across multiple public benchmarks and our collected data in accuracy, naturalness, and temporal coherence, setting a new state-of-the-art.

As autonomous vehicle capabilities advance, the safe evaluation of driving policies in long-tail scenarios remains a critical bottleneck. In closed-loop simulation, the driving policy model actively interacts with the environment, where its actions dynamically update the simulator state and directly influence the next set of generated sensor observations. While recent reconstruction-based neural simulators offer photorealism, they are fundamentally constrained by their initial captured data and struggle to generalize to highly dynamic or novel scenes. To overcome these limitations, we introduce OmniDreams, a foundation generative world model mid- and post-trained from the Cosmos diffusion model to autoregressively generate action-conditioned videos in real time. By leveraging the rich visual priors of Cosmos and mid- and post-training on 21k hours of driving scenarios, OmniDreams synthesizes complex, unobserved phenomena that are hard for traditional simulators to capture, such as extreme weather and unpredictable dynamic agent behaviors. Crucially, it autoregressively conditions its photorealistic sensor generation on past frames, the current simulator state, and immediate driving actions. Deployed in a closed-loop system with the Alpamayo 1 policy model and AlpaSim orchestrator, OmniDreams acts as a highly responsive, reactive environment, providing a scalable and comprehensive solution for training and evaluating next-generation autonomous driving policies. We additionally show preliminary results indicating that a world-action model (WAM) post-trained from OmniDreams achieves strong performance on the Physical AI Autonomous Vehicles NuRec dataset, surpassing the VLA-based Alpamayo 1.5 research policy model while using only 1/5 the total parameters. These results highlight the potential for a real-time world model like OmniDreams to also serve as a backbone for policy architectures.

Video world models are a foundational generative technology for embodied AI and the Metaverse, yet existing approaches are inherently limited to a single agent observing from a single perspective. Extending these models to multi-agent settings introduces two critical challenges: data scarcity (coordinated multi-view recordings are prohibitively expensive to collect for general open-domain scenarios) and world state alignment (independently generated video streams cannot ensure that shared physical environments and events evolve consistently across views). To address these challenges, we propose MetaWorld, a novel framework that scales multi-agent video world models to open-domain environments directly from single-view videos. First, we introduce Monocular World-State Unrolling (MWSU) to explicitly decompose monocular footage into the camera operator's ego-motion and the visible subject's spatial trajectory. This camera-trajectory decomposition naturally extracts synchronized multi-agent motion data within a shared 3D space, completely bypassing the need for multi-camera setups. Second, for precise visual control, we develop the Subject-Aware World Generator to enable appearance-driven simulation conditioned on per-agent identity images. Finally, to ensure both views are grounded in the identical physical reality, we propose World-State Alignment, a per-frame inter-branch cross-attention mechanism inserted at every transformer layer of the video DiT. By jointly synchronizing the denoising process, WSA enforces both static geometric consistency and dynamic motion consistency, encouraging that the shared 3D environment and physical events remain well-aligned across both egocentric views. Extensive experiments demonstrate that MetaWorld achieves superior cross-view consistency and identity fidelity, establishing a highly scalable, physics-driven paradigm for multi-agent video world modeling.

The recent "Reasoning with Video" paradigm utilizes Video Generation Models (VGMs) to generate temporally coherent visual trajectories to complete reasoning tasks. Although state-of-the-art VGMs excel at visual quality, they often struggle to understand and follow task-specific rules, leading to logical failures across diverse reasoning scenarios. Existing efforts try to utilize Vision-Language Models (VLMs) as problem pre-solvers to produce or refine textual guidance for the VGM. However, textual descriptions fail to capture intricate spatiotemporal details, and VGMs often struggle to faithfully execute fine-grained or long-tail instructions even with a valid plan. While VLMs struggle as solvers, they possess strong perception capabilities to evaluate process-constraint satisfaction and final-goal achievement. Leveraging this strength, we introduce a paradigm shift that transitions the role of VLMs to "teachers". Specifically, a VLM teacher extracts task-specific rules to formulate differentiable rewards, guiding a VGM Reasoner via test-time online optimization of a lightweight LoRA module. This strategy enables adaptive test-time optimization and extends the reasoning capabilities beyond the VGM's intrinsic boundaries. Evaluations on symbolic (VBVR-Bench) and general-purpose (RULER-Bench) video reasoning benchmarks show that the proposed method yields a 16.7-point average performance gain, outperforming the VLM-as-Solver paradigm (+0.4 points) and Best-of-N scaling (+2.2 points) by a large margin at comparable test-time cost. These findings reveal that integrating VLMs as test-time teachers offers a promising paradigm for achieving generalizable video reasoning. Project Page: https://VLM-as-Teacher.github.io/

Identity-preserving video generation (IPVG) aims to synthesize high-fidelity videos that follow text prompts while faithfully preserving a reference identity. Despite recent progress, existing IPVG methods still struggle to balance high-level semantic control and low-level identity fidelity. To bridge this gap, we propose ST-DRC, an effective Spatial-Temporal Decoupled Reference Conditioning framework for identity-preserving text-to-video generation. At the framework level, ST-DRC performs latent in-context feature injection by encoding the reference image with the video VAE and concatenating it with noisy video latents, enabling rich low-level identity details to be accessed without additional adapters. To separate identity-aware reference retrieval from appearance copying, we introduce TASS-RoPE, a Temporal-Adjacent Spatial-Shifted RoPE scheme that places reference tokens near the video sequence in time but shifts them in space, allowing reference information to flow through spatio-temporal attention while suppressing pixel-level copy-paste shortcuts. To further prevent shortcut learning and strengthen the otherwise diluted identity supervision in the diffusion objective, we combine appearance-invariant reference augmentation with face-guided identity objectives, encouraging the model to preserve identity under variations in color, pose, and layout. At inference time, we introduce a three-stream reference classifier-free guidance strategy that independently controls text adherence and reference fidelity. Experiments demonstrate that ST-DRC achieves strong identity preservation, prompt alignment, temporal consistency, and video quality with a lightweight design built on LTX-2.3. Our method ranks among the top submissions in the facial identity-preserving video generation track, validating the effectiveness of spatial-temporal decoupled reference conditioning.

Diffusion models have shown remarkable success in video generation. However, whether such models are truly aware of the 3D structure underlying visual observations, rather than simply reproducing plausible 2D projections, remains an open question. In this work, we investigate this question through human motion control, a task that requires precise modelling of 3D human geometry, motion, camera viewpoint, and scene context. Unlike prior methods that rely on rendered 2D motion guidance videos, we propose a render-free framework that conditions video generation directly on compressed 3D human mesh tokens. This representation preserves full 3D geometric information while enabling a unified token-based generation pipeline that processes video tokens jointly with motion tokens in a DiT-based architecture. This design requires the model to reason jointly about appearance, 3D structure, and camera viewpoint during video generation. Experimental results demonstrate strong performance on human motion control benchmarks, while reducing artifacts induced by view-dependent 2D guidance and trajectory-pose mismatches during editing. These findings suggest that video diffusion models, when equipped with mesh tokenization, can better capture complex 3D human structures and their interactions with the surrounding environment.

Modern vehicle platforms are equipped with a rich sensor suite, including LiDAR, calibrated multi-camera rigs, and accurate ego-motion, that in principle offers strong signal for re-rendering a driving scene from novel viewpoints. A growing line of recent work leverages video diffusion models for this task, using their generative priors to synthesize plausible novel views from sparse vehicle observations. In practice, however, existing methods exploit only a fragment of this signal, and their quality tends to degrade as the target trajectory departs from the recorded driving path. We argue that this is fundamentally a multi-sensor fusion problem: sparse LiDAR reprojections supply accurate but incomplete metric geometry, surround-view reference imagery supplies dense appearance but no metric depth, and camera poses tie the two together across views. We introduce StreetNVS, a video diffusion framework that jointly conditions on all three signals through a Reference-Enhanced Camera Attention module based on a relative ray-level positional encoding. We develop a two-stage curriculum training strategy that gradually exposes the model to increasingly sparse LiDAR. On the Waymo Open Dataset, StreetNVS substantially outperforms state-of-the-art baselines under sparse LiDAR conditioning, matches methods that rely on 10-100 times denser point clouds. We further show capabilities of synthesizing coherent videos along extreme out-of-trajectory paths such as elevation, lane-shift, pullback, and rotation. Our website: https://streetnvs.github.io

To study the ability to infer physical dynamics from videos and extrapolate them forward in time, we assemble a dataset of 2D Material Point Method (MPM) physical simulations covering rich physical phenomena such as deformable objects, fluids, kinetic objects, and emitters. We study code generation and video diffusion approaches on this dataset, identifying their strengths and weaknesses by varying the amount of physically relevant side information. The code generation model, beyond giving a working demonstration of automatic synthesis of MPM simulations, reveals that such an approach struggles with inferring physical parameters from visual input, but relative to video diffusion, produces physically and temporally stable extrapolations forward in time, while the video diffusion model more strongly identifies geometric properties from visual input but produces physically implausible extrapolations.

We present PAI-Studio, a new reference-conditioned video synthesis task that addresses a long-standing challenge in cinematic background replacement: generating dynamic backgrounds aligned with foreground motion while preserving foreground identity, matching reference scene appearance, and achieving globally consistent illumination with realistic foreground relighting. Existing open-source systems and commercial APIs cannot simultaneously ensure motion-consistent background generation, high-fidelity foreground relighting and foreground identity preservation, often resulting in static backgrounds, inconsistent boundaries, and noticeable compositing artifacts. To bridge this gap, we build upon a Diffusion Transformer video backbone and reformulate the problem as an in-context conditional generation task. Through bidirectional attention, our model jointly captures foreground dynamics and background reference information within a unified architecture. We further construct a 30K-scale dataset sourced from high-quality films and online videos to support this task. Extensive evaluations demonstrate that our method significantly outperforms existing open-source and commercial API solutions.

We present a post-training quantization (PTQ) approach for Wan2.1-T2V-14B, a 14-billion-parameter text-to-video diffusion transformer, targeting the W8A8 HiFloat8 (HiF8) format on Ascend 910B NPUs. A central challenge in quantizing video DiT models is the heterogeneous activation distribution across transformer blocks: boundary blocks (the first and last few blocks) exhibit fundamentally different statistical properties from middle blocks, making uniform quantization ineffective. We conduct a systematic per-block activation analysis across all 40 WanAttentionBlocks and use the findings to motivate a boundary-protection strategy that retains the first two and last three blocks in BF16 while quantizing the remaining 35 blocks with W8A8 HiF8. The proposed PTQ method matches or marginally exceeds the BF16 baseline on all five VBench dimensions evaluated, indicating no measurable accuracy loss within the 5-prompt evaluation set. An ablation study over four protection configurations confirms that full boundary protection yields the highest average VBench score, validating the data-driven block selection. We additionally investigate quantization-aware training (QAT) as a complementary fine-tuning stage and analyze the conditions under which it fails to outperform plain PTQ on single-card hardware.

Large video diffusion models achieve strong visual quality but remain expensive to deploy because each sample requires many denoising steps and a large resident parameter footprint. This paper studies a deployment-oriented compression pipeline for Wan2.2-T2V-A14B by combining few-step distribution-matching distillation with low-bit quantization. The pipeline follows the model's dual-expert denoising route, calibrates the high-noise and low-noise branches separately, protects sensitive entrance layers, and uses HiF4-style low-bit representation to improve dynamic-range coverage. Quantization is calibrated on the distilled few-step student rather than on the original long-step trajectory, reducing activation-distribution mismatch during inference. The proposed co-design keeps the quantized model close to the same-step full-precision model and surpasses the original full-precision baseline at 8 and 20 steps on average. The 20-step setting gives the best quality-efficiency trade-off in the tested configurations.

Video Diffusion Models (VDMs) synthesize high-quality videos, but precise camera and motion control remains challenging when large viewpoint shifts, changing occlusions, or complex dynamics reveal unseen regions. Diffusion priors alone can yield structural inconsistencies, while warped guidance can contain incomplete geometry and misleading view-dependent appearance cues. Real2SAM2Real complements VDM priors with a coarse 3D foreground cache of topologically complete meshes for intuitive editing and animation. Generative 3D lifting estimates these object and human meshes, including unobserved surfaces, from a single reference image. Normal maps rendered under specified camera trajectories and foreground motions guide video synthesis in the VDM, while the reference image provides appearance conditioning. Soft spatial-aligned injection and minimally invasive fine-tuning incorporate this guidance while largely preserving pretrained synthesis priors. Decoupled conditioning enables independent appearance and geometry editing. We train on around 2,000 monocular clips using instance-masked normal sequences with instance-level spatial perturbations to reduce sensitivity to differences between video-estimated training normals and normals rendered from coarse 3D caches at inference. Training requires neither explicit 3D assets nor paired multi-trajectory recordings. Real2SAM2Real supports decoupled camera and multi-instance motion control, including articulated humans. Across these challenging settings, experiments show strong spatiotemporal consistency and better overall control accuracy and video fidelity than evaluated state-of-the-art methods. Preliminary demonstrations suggest potential robotics applications.

Video world models (WMs) have shown promise for policy evaluation and improvement by imagining realistic future observations conditioned on ego-robot actions. While WMs can model distributions over futures, policy evaluation and improvement typically rely on nominal imaginations, which can miss high-impact outcomes of robot actions unless prohibitively many samples are drawn. To enable robust policy evaluation and improvement over WM imaginations, we propose StressDream, which steers imaginations toward high-impact yet plausible outcomes specified at inference time by optimizing the initial noise of diffusion-based WMs. However, optimizing high-dimensional noise is challenging: the optimization must reason about nuanced, scene-dependent target events in generated videos while avoiding out-of-distribution (OOD) noise that yields implausible imaginations. We address this with two complementary objectives: a semantic objective with a Vision-Language Model that provides informative gradients by reasoning about the generated video, and a plausibility objective that prevents the optimized noise from drifting OOD. With state-of-the-art video world models for autonomous driving and robotic manipulation, we show that StressDream effectively steers imaginations toward high-impact yet plausible outcomes specified by text at inference time, such as task failures, enabling robust policy evaluation and improvement by identifying actions whose plausible futures include undesirable outcomes. Video results are available at https://stressdream.github.io/.

Dense self-attention is the compute and quality bottleneck of long-video diffusion inference: cost grows quadratically with the sequence length, and beyond the training horizon the model converges to near-static output, that is, "frozen" repetitive video. State of the art approaches are either too costly, e.g., they require retraining, or fail to satisfy both performance and quality objectives in a scalable manner. To this end, we introduce Long Video Sparse Attention (LVSA), a training-free model-agnostic block-sparse attention for video diffusion transformers that combines a structured window pattern with rotating global anchors, thus removing the fixed-grid bias which causes long-range temporal artifacts. LVSA, combined with a FlashInfer kernel, reduces compute up to 3.17x on Wan 2.1 1.3B at a 6x horizon, 2.98x on Wan 2.1 14B at a 6x horizon, and 3.33x on HunyuanVideo 1.5 at a 1.5x horizon, compared to dense attention. Beyond reducing compute, LVSA enables HunyuanVideo 1.5 generation at a 2x horizon, which is otherwise out-of-memory on a single GPU. Moreover, LVSA provides speedups up to 2.41x compared to RIFLEx and 3.27x compared to UltraViCo on Wan 2.1 1.3B. To demonstrate applicability across diverse platforms, we apply LVSA on NPUs and achieve speedups up to 2.71x on Wan 2.2 A14B and 3.24x on Wan 2.1 1.3B compared to dense attention. To evaluate quality in a fair way, we introduce VQeval, a tool properly scoring loopy video failures, which instead are rewarded in state of the art evaluators like VBench-Long. LVSA is quality-neutral for generation at training horizon length and quality-positive at extended lengths.

Streaming video generation models typically rely on temporal-centric memory, which organizes historical context as raw frames, chunk segments, or unclustered tokens. This organization frequently leads to identity drift and semantic inconsistency when entities exit the frame or during interactive prompt transitions. To address these limitations, we propose SlotMemory, an object-centric Key-Value memory mechanism for streaming video diffusion. Our approach shifts the memory abstraction from "when" an event occurred to "what" is being represented by decomposing the transformer's key-value manifold into discrete, reusable semantic slots. By utilizing these slots as routing addresses to index and store high-fidelity key-value tokens, we enable entity-level persistence and prompt-aware retrieval across long horizons. Evaluated on 60-second interactive narratives using the Wan2.1-T2V-1.3B backbone, SlotMemory achieves a state-of-the-art quality score of 81.61 and a 22.8 percent relative improvement in dynamic consistency over the strongest existing streaming baseline. Our results demonstrate that structured semantic representation, rather than raw temporal capacity, is the essential primitive for persistent long-form video synthesis. Our codes and checkpoints are available at https://tj12323.github.io/SlotMemory/.

Precise camera pose control is critical for video diffusion, yet maintaining geometric consistency remains a challenge. Existing methods that directly inject numerical camera parameters into the diffusion backbone often fail to bridge the gap between abstract coordinates and visual content, leading to structural distortions. To address this issue, we propose CameraNoise, a flow-to-noise warping method that encodes camera motion into a temporally coherent stochastic representation. Unlike conventional conditioning, CameraNoise embeds camera poses directly into the noise space. This decouples motion from scene appearance while faithfully preserving trajectory dynamics. Specifically, we introduce a novel Geometry-guided Reprojection Flow and a noise warping algorithm, which jointly preserve the Gaussian prior of diffusion and ensure consistent noise propagation under camera transformations. By integrating CameraNoise into the diffusion process, our framework delivers stable, high-fidelity videos. Extensive experiments demonstrate that our approach significantly outperforms prior methods in both visual quality and trajectory faithfulness. The project page and code are available at: https://gulucaptain.github.io/CameraNoise/.

Recent progress in video diffusion models has enabled remarkable generative fidelity, yet leveraging these priors for restoration remains limited by the strong coupling between conditional and unconditional branches in standard classifier-free guidance. We introduce a training-free framework that enhances distorted and low-resolution videos by decoupling these signals in time. Our proposed Decoupled Time Guidance (DTG) evaluates the unconditional branch at a cleaner diffusion timestep, providing a lookahead prior that preserves geometry while suppressing replication of warped content. This temporal bias is annealed throughout sampling, allowing the model to transition from structure correction to detail refinement without retraining. Combined with any off-the-shelf restoration module in a plug-and-play manner, our approach improves perceptual coherence and restores plausible structure in AIgenerated and real-world videos alike. To facilitate evaluation, we curate GenWarp480, a benchmark of 4,400 distorted 480p videos synthesized from diverse text-to-video models. GenWarp480 focuses on characteristic generative degradations such as warped faces, body misalignments, and spatial artifacts, providing a purpose-built testbed for assessing robustness to generative errors. Extensive experiments demonstrate that our method achieves significant improvements in structural fidelity and temporal stability without any model training.

Long-rollout causal video diffusion has converged on a fixed-size sliding-window KV cache, with recent progress innovating within this layout by changing which tokens occupy the window or how their positions are encoded. The per-head KV layout itself, a dominant contributor to streaming memory and latency, has been mostly left unchanged. In this paper, we present the first study of Multi-Head Latent Attention (MLA) in video diffusion. VideoMLA replaces per-head keys and values with a shared low-rank content latent and a shared decoupled 3D-RoPE positional key, reducing per-token KV memory by 92.7% at every cached layer. We further investigate why MLA succeeds in video diffusion even though the spectral assumption often used to motivate it in language models does not hold: pretrained video attention is not low-rank, with 99%-energy effective rank far above any practical latent dimension. VideoMLA retains quality at compression ratios where direct spectral approximation would predict large reconstruction error. We show that the MLA bottleneck, rather than the pretrained spectrum, determines the effective rank: both spectral and random initialization occupy nearly the full rank budget from initialization, and training preserves this budget while adapting within it. On VBench, VideoMLA matches short-horizon streaming video diffusion baselines, achieves the best overall score at long horizons among evaluated methods, and improves throughput by 1.23x on a single B200.

Autoregressive video diffusion models generate streaming video by producing frames sequentially, conditioning each chunk on previously generated content. These models are structurally anchored to the first frame: its key-value representation occupies a privileged position in the attention cache and serves as the primary scene reference throughout generation. As the cleanest and most error-free position in the cache, this anchor draws disproportionate attention, suppressing video dynamics, and locking scene composition to the initial viewpoint even as the scene naturally evolves. The result is a temporally shallow video in which motion, camera movement, and scene progression are dampened in favor of static consistency. To address this, we replace the static anchor with an adaptive state, a hidden latent that the model denoises alongside content at every chunk but never renders. Rather than referencing a frozen first frame, the model generates its own scene anchor at each step by attending to both the previous state and the current content, producing a reference that evolves with the generated content. Unlike standard video generation, which encodes an absolute notion of time, our formulation treats time as relative: every generation step sees the same positional structure regardless of how far generation has progressed, and the state transition is identical at every chunk. Together, these properties introduce a recurrence into the generation process, where denoising serves as the transition function, and the KV cache serves as the carrier, requiring no external module. Experiments demonstrate that the adaptive state substantially improves video dynamics, enabling richer motion and natural scene progression within generated videos.

As video diffusion models (VDMs) advance toward world models, a key question arises: do they truly understand causality, or merely overfit to statistical temporal patterns? Existing benchmarks mostly rely on synthetic data, limiting real-world generalization due to the sim-to-real gap. We present YoCausal, a two-level benchmark inspired by the Violation of Expectation (VoE) paradigm from cognitive science. By temporally reversing real-world videos at zero cost as natural counterfactual samples, YoCausal establishes an arbitrarily extensible evaluation protocol. Level 1 introduces the Reverse Surprise Index (RSI), quantifying arrow-of-time perception via denoising loss. Level 2 introduces the Causality Cognition Index (CCI), which leverages a VLM to stratify datasets into causal and non-causal subsets, disentangling genuine causal reasoning from temporal bias. Evaluation of 13 state-of-the-art VDMs reveals that perceiving the arrow of time does not imply understanding causality, and a significant gap persists relative to human-level causal cognition.

Scaling Diffusion Transformers to generate high-resolution, long videos is constrained by the quadratic cost of self-attention, and existing sparse attention methods degrade under high sparsity. We show empirically that generation quality is determined not by the sparsity ratio itself, but by how well the sparse mask aligns with the tile-wise geometry of full attention. Based on this insight, we propose Veda, a distilled sparse attention framework that formulates tile selection as an explicit reconstruction problem from full attention. Veda integrates statistics-aware tile scoring with head-aware tiling to reduce estimation error and structural mismatch, enabling aggressive sparsity. A hardware-efficient tile-skipping kernel converts theoretical sparsity into practical wall-clock speedups. Experiments on large video diffusion models, including Waver and Wan2.1, demonstrate substantial acceleration with no noticeable degradation in generation quality. To generate 720P 10-second videos on Waver-T2V-12B, Veda achieves a 5.1$\times$ end-to-end speedup and a 10.5$\times$ self-attention speedup, reducing attention overhead from 92% to 50%. Notably, the gains increase with sequence length, indicating that Veda scales favorably with spatiotemporal resolution across models.

Recent video diffusion foundation models have achieved remarkable progress in high-quality video generation, yet turning them into real-time interactive video world models remains challenging. Interactive world models require controllable, causal, and low-latency rollout, which in practice demands a full pipeline spanning data construction, controllable fine-tuning, autoregressive training, few-step distillation, and streaming inference. In this work, we present minWM, a full-stack open-source framework for building real-time interactive video world models. minWM provides an end-to-end pipeline that converts existing bidirectional T2V/TI2V video foundation models into camera-controllable few-step autoregressive world models. Specifically, minWM first fine-tunes a bidirectional video diffusion model with camera control, and then applies the Causal Forcing / Causal Forcing++ pipeline, including AR diffusion training, causal ODE or causal consistency distillation, and asymmetric DMD, to distill it into a few-step autoregressive generator for low-latency rollout. The framework is modular and architecture-extensible: we instantiate it on representative open backbones, including Wan2.1-T2V-1.3B and HY1.5-TI2V-8B, covering both cross-attention-based condition injection and MMDiT-style architectures. minWM also supports adapting existing video world models, such as HY-WorldPlay, to new data distributions, training recipes, and latency targets. Beyond releasing runnable scripts, checkpoints, documentation, and inference code, we provide practical ablations on camera trajectory quality, controllability training steps, and minimal batch-size requirements. We hope minWM serves as a reproducible and extensible recipe for building and adapting real-time interactive video world models. Project Page: https://github.com/shengshu-ai/minWM

Distribution Matching Distillation (DMD) is a widely used paradigm for accelerating inference in few-step video diffusion models. However, DMD-style video distillation faces two coupled challenges: the fake score must track a continuously evolving generator, making training costly when frequent updates are required, while reverse-KL-style matching can be mode-seeking and conservative for preserving strong motion dynamics. To address these issues, we propose Score Gradient Matching Distillation (SGMD). SGMD adopts a fake-score perspective by directly optimizing the fake score toward the teacher, while using teacher stop-gradient Fisher as a stable distribution-matching objective. We provide a gradient analysis that motivates this objective choice under ideal tracking. Building on this, SGMD introduces a pair of dual potentials: negative-residual (NR) for outer-loop correction and residual-contraction (RC) for inner-loop tracking. Empirically, compared to DMD2, SGMD achieves an approximately $\sim 3\times$ training speedup and substantially improves motion dynamics for 4-step distilled models while preserving temporal consistency. A human study confirms that SGMD is preferred in motion quality and overall preference, while visual quality and text alignment remain comparable. Code is available at https://github.com/ModelTC/LightX2V.

Video object removal frequently struggles to simultaneously eliminate target objects and their associated physical effects (e.g., smoke, reflections, light, and ripples) in out-of-domain scenarios due to complex spatiotemporal ambiguities. While existing methods primarily rely on spatial masks, they often fail to capture weakly correlated effects, and the potential of explicit textual guidance remains underexplored. Furthermore, a fundamental optimization conflict exists in removal models between high-level semantic generalization and precise pixel-level background preservation. To address these challenges, we propose GenEraser, a novel framework for generalized and high-fidelity video object and effect removal. First, we introduce a Multi-Conditional Mixture-of-Experts (MC-MoE) paired with Bipartite Text guidance to fully exploit the multimodal priors of Diffusion Transformers, significantly enhancing the identification of complex effects. Second, a Learnable Deep ``CFG'' Fusion mechanism (LD-CFG) is developed to adaptively balance the relative dominance of mask and textual conditions across diverse scenarios. Finally, we propose a Decoupled Expert Architecture, comprising a Locator and a Preserver, to mitigate the inherent trade-off between semantic generalization and pixel alignment. Extensive experiments demonstrate that our GenEraser surpasses recent state-of-the-art approaches, achieving significant quantitative improvements (e.g., $2.16$ dB and $1.44$ dB on the ROSE Benchmark and VOR-Eval, respectively) while maintaining exceptionally robust generalization in open-world scenarios. https://cyqii.github.io/GenEraser.github.io/

In recent years, training-free video generation has progressed remarkably. However, when handling complex textual instructions, existing methods still suffer from semantic ambiguity, incorrect concept binding, and cross-frame inconsistency. To address these issues, we propose KGEdit, a structured semantic control framework for text-to-video (T2V) diffusion models. Specifically, we first construct an ambiguity-aware knowledge graph (AAKG) to disentangle and disambiguate the input prompt, converting it into four types of structured semantics: identity, relation, attribute, and negative constraints. We then design a structured semantic injection module (SSIM) to inject these semantic signals into key layers of the diffusion Transformer, enabling fine-grained semantic control. In addition, we introduce a temporal-aware semantic control (TASC) module that dynamically schedules semantic objectives according to the stage-wise characteristics of the denoising process, further improving semantic alignment and temporal consistency. Experiments show that KGEdit outperforms existing methods in editing precision and temporal stability, while offering higher efficiency and controllability in text-driven interaction scenarios.

We present a method for harmonizing the lighting of a foreground video to match a target background scene, adjusting shadows, color tone, and illumination intensity (relightful harmonization). Unlike images, acquiring labeled data for videos, where identical motions are recorded under different lighting conditions, is practically infeasible and non-scalable. While one way to create such paired data is to apply existing image-based harmonization models frame by frame to a video, the resulting outputs often suffer from significant temporal jitters. We overcome this problem by introducing a novel lighting deflickering model that can stabilize the global and local lighting flickering artifacts. Our video diffusion model learns from these upgraded deflickered data with a volume of real and synthetic videos to generate high-quality video harmonization results. We further propose an asymmetric alpha mask conditioning technique to learn the clean boundaries from real videos. Experiments demonstrate that our model achieves strong temporal coherence, naturalness, cleaner boundaries, and physically meaningful lighting behavior, while maintaining strong relighting expressiveness compared to prior image-based and video-based harmonization methods.

Diffusion Transformers achieve strong video generation quality, but the quadratic cost of full attention limits efficiency. We introduce OSP-Next, an efficient text-to-video generation model that integrates sparse attention, parallelism, quantization, and reinforcement learning. OSP-Next uses a hybrid full-sparse attention architecture, where the sparse component is implemented with Skiparse-2D Attention. This fixed-pattern mechanism applies token-wise and group-wise sparse attention along spatial dimensions, leveraging locality while maintaining native compatibility with FlashAttention kernels. Based on the local equivalence of rearrangement in Skiparse-2D Attention, we further propose Sparse Sequence Parallelism (SSP), which partitions subsequences across ranks and switches sparse patterns through a single All-to-All communication. Compared with Ulysses Sequence Parallelism (SP), SSP provides a native parallel strategy for sparse attention and reduces communication volume by 75%. OSP-Next also incorporates HiF8 quantization to enable stable joint training with 8-bit quantization and sparse fine-tuning, and applies Mix-GRPO post-training to improve the performance of the sparse model. Experiments show that OSP-Next achieves a VBench total score of 83.73%, surpassing the Wan2.1 baseline. Under the 5-second 720P and 5-second 768P settings, OSP-Next achieves up to 1.64$\times$ single-GPU speedup and over 1.52$\times$ eight-GPU speedup on NVIDIA H200 GPUs. In addition, with only a 0.4% drop in VBench total score, OSP-Next-HiF8 achieves 1.69$\times$ and 2.27$\times$ speedups under the two settings on a single Ascend 950PR, demonstrating the efficiency and performance of OSP-Next across hardware platforms.

Pretrained foundation models have become an important basis for end-to-end autonomous driving. In contrast to vision-language models pretrained primarily on static image-text pairs, video generative models capture temporal dynamics and motion priors that are naturally suited for driving. We present DriveWAM, a driving world-action model that adapts a pretrained video diffusion transformer into an autoregressive video-action policy. DriveWAM organizes video and action streams into a unified temporal token sequence and trains them under a joint flow-matching objective, preserving the pretrained video-generation architecture while adapting its large-scale video priors to action generation. To incorporate high-level scene understanding, we introduce scene-evolving driving guidance, where a frozen VLM produces chunk-specific semantic intent to guide video-action generation. To keep long-horizon rollout bounded, we further introduce selective KV memory, which maintains bounded modality-aware video and action memory pools through relevance-redundancy cache selection at inference time. Experiments on NAVSIM and the PhysicalAI-Autonomous-Vehicles benchmark show that DriveWAM achieves strong planning performance, and a data-scaling study from 4k to 100k driving clips further confirms the scaling potential of world-action modeling for end-to-end autonomous driving.