Pre-Trained Video Diffusion Models

Generating complete digital twins from videos requires precise camera control, global scene coverage, and strict spatial-temporal consistency constraints that remain challenging for perspective video generators due to their limited field of view (FoV). Their narrow FoV forces long or multi-view trajectories, amplifying cross-view inconsistency and temporal drift. We argue that 360° video generation offers a natural solution: panoramic coverage simplifies trajectory design and provides a strong global context for maintaining coherence. We introduce Pantheon360: Taming Digital Twin Generation via 3D-Aware 360° Video Diffusion, a controllable 360° video generation framework that synthesizes high-fidelity videos from sparse 360° inputs. The key idea is an explicit 3D Cache, reconstructed from the input, which serves as a geometric scaffold for any user-defined camera path. This allows the diffusion model to focus on photorealistic texture refinement while the 3D Cache enforces global geometric consistency. Experiments show that Pantheon360 achieves superior visual quality and unmatched geometric coherence, enabling reliable and flexible 360° scene generation for downstream simulation and digital-twin applications.

Video world models should maintain evolving states when evidence is unobserved, yet current generators often freeze hidden states upon interruption. This is not simply a capacity problem: pretrained video diffusion transformers already possess KV-cache mechanisms capable of non-local retrieval, but they are rarely trained to use them as dynamic memory. We introduce ReMind, a framework eliciting dynamic memory behavior via memory-oriented data, event-aware training, and cache adaptation. Organized around a taxonomy of 100+ dynamic events, we build a camera-annotated training mixture combining VLM-filtered real videos, generated hard dynamics, synthetic camera loops, and memory-interruption augmentations. Each clip is converted into a frame graph with protected anchors, degraded intervals, and explicit temporal gaps. A node-structured curriculum, including node-drop, noisy memory, frontier continuation, and reference-cache training, forces the model to retrieve relevant past states across interruptions rather than relying solely on local continuity. PM-RoPE, an elegant camera-phase RoPE extension, unlocks spatiotemporal retrieval at a single-attention cost while preserving pretrained pathways. ReMind achieves the best overall scores on STEVO-Bench and recovery tasks. Furthermore, general image-to-video evaluations confirm this curriculum avoids catastrophic forgetting. We will open-source our code, data, and models.

Consistent 3D geometry estimation from streaming RGB input is crucial for real-world applications such as autonomous driving, embodied AI, and large-scale reconstruction. While modern monocular geometry foundation models achieve strong single-image accuracy, they exhibit severe temporal inconsistency on continuous input, notably dominated by scale--shift drifting. Through targeted empirical analysis, we trace this instability to its root cause: fluctuations in latent feature statistics, whose mean and variance directly determine the predicted depth's scale and shift. Building on this insight, we introduce Dynamic Feature Normalization (DyFN), a lightweight, causal recurrent module that dynamically and robustly modulates feature statistics to maintain stable geometry over time. We adapt powerful pretrained monocular geometry models for streaming by finetuning only DyFN, a mere 2% additional parameters, while keeping the backbone frozen, thereby achieving temporal consistency without compromising single-image accuracy. Extensive experiments across four benchmarks show that DyFN effectively eliminates temporal artifacts such as disjointed layering and positional jitter, and achieves state-of-the-art temporal stability, improving over prior streaming methods by up to 14% and even outperforming heavier non-causal video baselines. Project Page: https://shawlyu.github.io/DyFN

Incorporating camera intrinsics into video generation models offers a principled way to control not only scene dynamics but also the imaging process that governs visual appearance. Prior work has primarily focused on extrinsic control, such as camera pose and motion, while treating intrinsic camera parameters as implicit or fixed. A key bottleneck is the lack of large-scale video datasets with accurate and diverse temporally varying camera metadata, which makes learning absolute camera parameterizations difficult. As a result, current models struggle to incorporate photographic camera behavior, including depth-of-field transitions, exposure variations, lens distortions, and color processing, in a controllable and temporally consistent manner. We introduce DeltaCam, a video diffusion framework that models camera behavior through $Δ$-parameterized neural camera adaptors, operating on relative changes in camera motion and intrinsics instead of absolute states. By learning this differential formulation from synthetic video data, we mitigate reliance on precise real-world camera labels and enable smooth, consistent control over imaging factors such as focal length, aperture, ISO, color temperature, and lens distortion. We extend this framework to real-world footage through two mechanisms: finetuning the controls on real image-metadata pairs for precise shot matching, and extracting disentangled embeddings for implicit video-to-video style transfer without requiring explicit camera parameters. By effectively separating scene content from intrinsic imaging behavior, DeltaCam enables camera-consistent video generation and editing operations that are difficult to achieve with existing models. Ultimately, our results establish a practical and scalable approach for bridging synthetic control and real-world photographic emulation.

Recent video-based world models have made pixel-space environments interactive at the camera level: users can navigate viewpoints while the model generates coherent visual continuations. Yet their action spaces remain incomplete: users can move the camera, but cannot act on individual objects. Since real-world interaction is inherently object-centric, such models remain closer to passive scene observers than truly manipulable environments. We present WorldCraft, a framework that expands interactive video world models from camera navigation to object-level trajectory actions. Given a user click and a sketched path, WorldCraft generates future frames in which the selected object follows the prescribed trajectory while the camera continues to navigate the scene. WorldCraft achieves this through a trajectory-centric control pipeline: First, Normalized World Trajectory (NWT) represents user-drawn motion in a camera-invariant world coordinate system and dynamically re-projects it under the current camera pose, separating object motion from camera-induced screen-space displacement; Spatial-Pathway LoRA (SP-LoRA) then injects this world-space signal through the model's spatial-control pathway, adding object manipulation capability while preserving the pretrained camera controller; finally, Trajectory-Anchored State Persistence (TASP) treats the world trajectory as a persistent spatial state and refreshes autoregressive memory after trajectory-conditioned generation, allowing moved objects to reappear at their updated positions after leaving the camera view. Experiments show that WorldCraft enables accurate object control, preserves the video-based world model's camera fidelity under camera-only evaluation, and maintains object state across long autoregressive rollouts with off-camera excursions.

Instruction-based video editing requires transforming a source video according to a natural-language instruction while preserving irrelevant content and remaining temporally coherent. We argue that existing Diffusion Transformer (DiT) editors struggle with this task for two structural reasons. First, conditioning signals are fed undifferentiated into all transformer blocks, forcing a single token stream to encode both global editing intent and fine-grained visual evidence. Second, the cross-attention patterns that govern the edit are supervised only indirectly through pixel-level reconstruction, leaving the model's internal reasoning process under-constrained. To address both limitations, we propose RVEDiT, an implicit Reasoning Video Editing DiT framework built around two complementary components. The first, Granularity-Routed Token Conditioning, introduces learnable editing tokens distilled from a multimodal LLM and routes them to shallow blocks, while reserving native visual and textual tokens for deeper blocks, thereby inducing a coarse-to-fine editing process inside the backbone. The second, Reference-Anchored Attention Alignment, employs a parameter-sharing reference branch during training and maximizes the mutual information between the attention features of the editing and reference branches, regularizing the model's internal reasoning without incurring any additional inference cost. Experiments on standard instruction-based video editing benchmarks show that RVEDiT consistently outperforms state-of-the-art baselines, with particularly strong gains on localized and compositional edits.

Recent progress of video diffusion models have enabled extensive simulation of the physical world. While simulation with hand object interaction has been less explored. We propose DexSIM, a dexterous simulation framework for simulating dexterous manipulation in real-time. While previous works utilizing video diffusion and 3D reconstruction focus on navigation, dexterous manipulation has been limited while it has extensive applications for creating interactive experiences with the simulated world and for generating synthetic data for robotics. Existing methods lack real-time interactivity and long-term spatial consistency and memory. We propose a 2-stage training framework for DexSIM. First we train a bi-directional video diffusion model by jointly embedding the hand action trajectory and video in a unified feature space. We utilize gaussian heatmap hand encoding for more accurate hand representation. Then we conduct a roll-out based autoregressive training with updated spatial cache as attention sink for spatial memory, which improves long-term consistency and 3D aware dexterous manipulation simulation. DexSIM outperforms the baseline on pixel and semantic similarity, motion fidelity, and hand projection accuracy. It also allows new applications such as hand motion transfer and runs at 15.24 FPS real-time interactivity.

Camera-controlled video generation has achieved remarkable progress in recent years. However, existing video-to-video re-rendering methods primarily rely on Supervised Fine-Tuning using synthetic datasets. At present, there is an extreme scarcity of synchronized, multi-view real-world video data. Consequently, the prevailing paradigm often exhibits limited generalization when processing out-of-distribution real-world videos, with models struggling to accurately adhere to physical scales and camera trajectories. To bridge this gap, we propose Geo-Align, the first Reinforcement Learning framework specifically designed for camera-controlled video re-rendering. Built upon a pretrained model, we optimize the model through a scale-aware perceptual reward mechanism. Specifically, we introduce a metric 3D estimator to extract precise camera trajectories from generated videos, explicitly penalizing deviations in rotation and translation. Furthermore, we meticulously designed a data pipeline strategy based on real-world conditioning videos and target camera trajectories derived from synthetic data, eliminating the reliance on paired data. Extensive experiments demonstrate that Geo-Align consistently outperforms existing supervised learning baselines in both precise camera controllability and visual fidelity, indicating the effectiveness of our method.

Recent advances have substantially improved real-time interactive video generation in the autoregressive regime. However, most existing few-step autoregressive video generation methods, often distilled from a corresponding many-step teacher, default to a 4-step sampling configuration, which still incurs considerable latency during deployment and suffers from severe quality degradation when the number of sampling steps is further reduced, particularly in the one-step setting. Trajectory-style consistency distillation methods often produce videos with weak dynamics, while DMD-based approaches, such as Self-Forcing, tend to yield blurry frames. To address this challenge, we propose One-Forcing, a simple yet effective approach which augments the DMD objective with an auxiliary GAN loss for high-quality and efficient one-step video generation. Experiments on VBench show that One-Forcing achieves a total score of 83.76, establishing state-of-the-art performance among one-step causal video generation methods and remaining competitive with strong many-step approaches. We further demonstrate that one-step framewise autoregressive generation can be achieved stably with merely one-third of the training cost of the chunkwise model, a setting that prior methods have failed to achieve successfully.

Diffusion transformers have achieved remarkable success in high-quality video generation, yet their reliance on spatiotemporal 3D full attention incurs prohibitive computational cost due to the quadratic complexity of attention. Block sparse attention is a common approach to mitigate this by focusing computation on important regions. However, attention maps in DiTs exhibit inherently dynamic and fine-grained sparsity, which causes existing block sparse attention methods to degrade significantly in quality, especially at high sparsity ratios. In this paper, we revisit block sparse attention and derive a theoretical lower bound on attention recall to characterize the key factors governing its effectiveness. Guided by these insights, we propose DFSAttn, a training-free sparse attention framework that enables dynamic, fine-grained sparsification efficiently. DFSAttn incorporates three core designs: Hilbert curve-based token reordering to achieve fine-grained sparsity while preserving efficient GPU execution, hierarchical block scoring for accurate block importance estimation, and sparse mask caching with adaptive ratios to balance accuracy and efficiency. Experimental results demonstrate that DFSAttn consistently outperforms prior methods under high sparsity, achieving up to 2.1$\times$ end-to-end speedup while maintaining high generation quality. Our code is open-sourced and available at https://github.com/jessica-hujie/DFSAttn.

The rapid evolution of generative video foundation models has propelled the field toward professional-grade cinematic synthesis. To achieve such demanding quality, the community transitions towards Reinforcement Learning (RL) and agentic workflows. However, reliable evaluation has emerged as a critical bottleneck. Existing benchmarks predominantly evaluate ''whether it is right'' (basic prompt-following) while fundamentally neglecting ''whether it is good'' (cinematic quality, acting, and aesthetics). Furthermore, current automated metrics lack the domain-specific rigor required to provide trustworthy signals, creating a severe credibility gap between human aesthetic perception and machine scoring. To bridge this gap, we introduce EvalVerse, a comprehensive, pipeline-aware, and expert-calibrated evaluation framework. We treat video generation assessment not merely as an engineering task, but as a core scientific problem: the systematic digitization of subjective cinematic expertise. First, we organize domain knowledge into an evaluation taxonomy aligned with the professional filmmaking workflow (pre-production, production, and post-production). Second, we distill human expert judgments into a curated dataset with large-scale human annotations. Third, we inject this knowledge into Vision-Language Models (VLMs) through an expert-calibrated fine-tuning strategy, enabling the VLM to perform explicit Chain-of-Thought reasoning. Compared to previous works, EvalVerse not only retains compatibility with foundational ''rightness'' metrics, but also significantly expands the criteria to ''goodness'' and broaden the task coverage to complex multi-shot sequencing and audio-visual integration. Consequently, by providing granular diagnostic signals, EvalVerse transcends a static leaderboard and establishes a fundamental infrastructure for future work, such as reward models and evaluator agent.

Generative text-to-video systems can produce visually fluent educational clips, but they rarely encode the pedagogical content knowledge (PCK) needed for effective instruction, including prerequisite-aware sequencing, learner-adaptive depth, and sustained cognitive engagement. We present CourseBlueprint, a course-grounded pipeline for adaptive pedagogical video generation. Given a topic and learner persona, the system generates a structured teaching blueprint in a single forward pass over an undergraduate biomedical-imaging corpus (BMED 2300; twenty-three lectures, 1,116 slides). Instead of ad-hoc prompt chaining, the pipeline uses typed intermediate representations with validation: a scaffolding module builds a stage-labeled prerequisite concept graph with deterministic cycle removal, an adaptive controller assigns per-concept style specifications, and an engagement generator produces narration following a fixed hook->retrieval->core->analogy->forward contract. A deterministic slide-image override further grounds the rendered video by reusing instructor slides whenever retrieval confidence is high. We also release a reusable benchmark corpus and an evaluation harness combining repeated LLM-judge scoring with regex-grounded objective metrics. In a five-topic ablation, removing the engagement contract reduces the engagement score from 5.00 to 1.20, the adaptive score from 4.80 to 3.40, Flesch readability from 38.0 to 19.8, and analogy and retrieval-prompt counts to near zero. The slide-image override converts a 0/9 corpus-grounding failure into 9/10 successful slide matches on the same topic. These results show that pedagogical video quality depends less on surface fluency than on explicit, typed instructional contracts that make scaffolding, adaptation, engagement, and grounding auditable.

We present CoMoGen, a controllable video generation framework that generates realistic interactive dynamics from a single binary mask sequence conditioned on an input image. CoMoGen introduces a lightweight MaskAdapter that encodes binary mask sequences into a latent residual signal, injected into the Multi Modal Diffusion Transformer (MMDiT) model through a cosine-weighted schedule. Unlike the hierarchical coarse-to-fine design of UNet architectures, MMDiT operates as a sequence of uniform transformer blocks, making it difficult to identify which layers are responsible for the motion generation. Therefore, we propose a novel way to determine "Motion Layers" operating in the attention space of MMDiT. We fine-tune the model by using Low-Rank Adaptation (LoRA) to the Motion Layers, without requiring any architecture change in the MMDiT. This selective adaptation enables our method to focus on motion-critical components, yielding reduced computational cost. Despite its simplicity, CoMoGen enables precise subject motion and plausible interactions with surrounding humans, objects, and scenes. Comprehensive experiments on different datasets show that CoMoGen consistently outperforms prior controllable video generation methods and achieves state-of-the-art performance in motion fidelity and perceptual realism. Project page: mericadil.github.io/CoMoGen.

Autoregressive video diffusion models have enabled real-time, action-conditioned world generation. However, sustaining a persistent world, where revisiting a previously seen viewpoint yields consistent content, remains an open problem. Full KV-cache attention preserves this consistency but breaks real-time constraints: memory footprint and attention cost grow linearly with rollout length. Sliding window inference restores throughput but discards long-term consistency. We propose WorldKV, a training-free framework with two components: World Retrieval and World Compression. World Retrieval stores evicted KV-cache chunks in GPU/CPU memory and selectively retrieves scene-relevant chunks via camera/ action correspondence, inserting them back into the native attention window without re-encoding. World Compression prunes redundant tokens within each chunk via key-key similarity to an anchor frame, halving per-chunk storage to fit 2x more history under a fixed budget. On Matrix-Game-2.0 and LingBot- World-Fast, WorldKV matches or exceeds full-KV memory fidelity at roughly 2x the throughput, and is competitive with memory-trained baselines without any fine-tuning. Project Page: https://cvlab-kaist.github.io/WorldKV/

Training Deep Neural Networks for tracking individual cells in biomedical videos requires a large amount of annotated data. The annotation of videos for cell tracking is very time consuming and often requires domain expertise; this explains the limited availability of public annotated data to address important medical problems like tissue repair or cancer treatment. Generating synthetic videos along with their Ground Truth annotations is a promising solution that relies, as a foundational first step, on the synthesis of single cell annotations (or phantoms). Phantoms need to be time consistent, as they have to replicate biological processes that are specific to the cell types. In this work, we propose a novel framework for generating videos of cell phantoms in the Elliptical Fourier Descriptors (EFDs) domain, a compact and geometrically interpretable representation for 2D closed contours. We represent the cell phantom evolution as a multivariate time series of EFD coefficients, introducing a strong prior for cell morphology and enabling the efficient generation of sequences that evolve coherently in time. Our experimental validation proves that modelling the temporal evolution in EFD space enables the generation of biologically plausible phantom videos. Our method can be used in generative pipelines for synthesizing annotated data for cell tracking, thus strongly mitigating the annotation effort for creating new datasets. Our code is available for download here: https://github.com/FrancescoBenedetto99/efd-cell-video-gen.

Diffusion Transformer (DiT) has emerged as a powerful model architecture for generating high-quality images and videos. In the case of video DiT, 3D Spatio-Temporal Attention increases token length in proportion to the number of frames, sharply increasing computational cost. Token reduction methods mitigate this cost by exploiting spatial redundancy, but existing approaches rely on inaccurate similarity estimates and lightweight matching algorithms, resulting in poor matching quality and only marginal acceleration. To overcome these limitations, we propose ORBIS, an SW-HW co-designed accelerator for video DiT. ORBIS leverages the output activation from the previous timestep to obtain more accurate inter-token similarity, substantially improving matching quality and enabling a higher token reduction ratio. We further introduce a Distribution-Aware Token Matching (DATM) algorithm that captures global token distribution and explicitly minimizes token-pair loss for additional gains. To fully hide DATM latency, we design specialized, deeply pipelined hardware and minimize its hardware cost through quantization, occupying only 2.4% of total area with negligible accuracy loss. Extensive experiments show that ORBIS achieves about 2x higher token reduction ratio than the state-of-the-art approach, AsymRnR, while delivering up to 4.5x speedup and 79.3% energy reduction compared to an NVIDIA A100 GPU.

Video Virtual Try-On (VVT) aims to seamlessly replace a garment on a person in a video with a new one. While existing methods have made significant strides in maintaining temporal consistency, they are predominantly confined to non-interactive scenarios where models merely showcase garments. This limitation overlooks a crucial aspect of real-world apparel presentation: active human-garment interaction. To bridge this gap, we introduce and formalize a new challenging task: Interactive Video Virtual Try-On (Interactive VVT), where subjects in the video actively engage with their clothing. This task introduces unique challenges beyond simple texture preservation, including: (1) resolving the semantic ambiguity of interactions from standard pose information, and (2) learning complex garment deformations from video where interactive moments are sparse and brief. To address these challenges, we propose iTryOn, a novel framework built upon a large-scale video diffusion Transformer. iTryOn pioneers a multi-level interaction injection mechanism to guide the generation of complex dynamics. At the spatial level, we introduce a garment-agnostic 3D hand prior to provide fine-grained guidance for precise hand-garment contact, effectively resolving spatial ambiguity. At the semantic level, iTryOn leverages global captions for overall context and time-stamped action captions for localized interactions, synchronized via our novel Action-aware Rotational Position Embedding (A-RoPE). Extensive experiments demonstrate that iTryOn not only achieves state-of-the-art performance on traditional VVT benchmarks but also establishes a commanding lead in the new interactive setting, marking a significant step towards more dynamic and controllable virtual try-on experiences.

Existing video-language models can generate factual descriptions of road events but lack control over how these events are expressed: their tone, urgency, or style. This limits deployment in communication-critical settings where the effectiveness of a message depends on both content and presentation, not just factual accuracy. To mitigate this, we introduce a comprehensive dataset-model-evaluation suite for tone-controllable road video captioning. Our human-validated data generation pipeline expands road-video corpora with diverse tonal annotations and multi-tone captions, yielding the RoadTones-51K dataset. We propose RoadTones-VL-CoT, a controllable video-to-text model that also generates tone-conditioned Chain-of-Thought intermediate drafts for interpretability. We also introduce RoadTones-Eval, a new evaluation suite that jointly measures factual consistency and tone adherence. In addition, we conducted a user study whose results validate caption quality, tone control, and factual consistency. Together, these contributions lay the foundation for context-sensitive tone-controllable video captioning.

Autoregressive video diffusion models (ARVDs) have emerged as a promising architecture for streaming video generation, paving the way for real-time interactive video generation and world modeling. Despite their potential, the substantial inference cost of ARVDs remains a major obstacle to practical deployment, making model quantization a natural direction for improving efficiency. However, quantization for ARVDs remains largely unexplored. Our empirical analysis shows that directly applying existing quantization schemes developed for standard diffusion transformers to ARVDs leads to suboptimal performance, revealing quantization behaviors that differ from those observed in bidirectional diffusion models. In this paper, we identify two critical challenges in quantizing ARVDs: (C1) Highly unbalanced frame-wise quantization sensitivity. Error accumulation during autoregressive generation can induce severely skewed quantization sensitivity across frames, following an exponential-like decay pattern. (C2) Prominent and heterogeneous outlier patterns in weights. Weight distributions exhibit pronounced outlier channels, whose patterns vary substantially across layer types and block depths. To address these issues, we propose Q-ARVD, a novel framework for accurate ARVD quantization. (S1) To tackle the highly unbalanced frame-wise sensitivity, Q-ARVD incorporates a final-quality aware frame-weighting mechanism into the quantization objective. (S2) To prevent heterogeneous outliers from degrading performance, Q-ARVD introduces an outlier-aware adaptive dual-scale quantization, which automatically detects the presence and quantity of outlier channels for an arbitrary layer, and isolates them to protect normal channels. Extensive experiments demonstrate the superiority of Q-ARVD.

Diffusion models have achieved impressive performance in video generation, but their iterative denoising process remains computationally expensive due to the large number of tokens processed at each timestep. Recently, progressive resolution sampling has emerged as a promising acceleration approach by reducing latent resolution in early stages. However, scaling this idea to video generation remains challenging, as the additional temporal dimension introduces diverse spatio-temporal demands across different videos, and compressing only a single dimension often leads to limited acceleration or degraded quality. Therefore, we propose DVG, a Dynamic Video Generation framework that jointly allocates computation across time and space, automatically selecting content-aware acceleration strategies without manual tuning or retraining. DVG achieves near-lossless acceleration across models and tasks, reaching up to 7 times speedup on HunyuanVideo and HunyuanVideo-1.5, and 18 times when combined with distillation, demonstrating its potential as a key component in today's large-scale efficient video generation systems. Our code is in supplementary material and will be released on Github.

Autoregressive long video generation often adopts bounded-memory streaming for efficiency, typically combining local windows for short-term continuity with static early-frame sinks as long-range anchors. However, this fixed allocation keeps early frames cached even when the current visual state has substantially diverged from them, while discarding potentially more relevant intermediate history. As a result, the retained long-range context may become less adaptive and bias generation toward outdated cues; in severe cases, RoPE-induced phase re-alignment can homogenize inter-head attention and cause sink collapse, where content regresses toward sink frames. We propose DySink, a retrieval-based framework that maintains a compact memory bank and selects visually relevant historical frames as dynamic frame sinks. DySink couples adaptive retrieval with a sink anomaly gate, which detects excessive inter-head consensus over retrieved context and suppresses collapse-prone context. Experiments on minute-long videos show that DySink consistently improves temporal quality over strong baselines while also achieving higher dynamic degree, enabling coherent and more natural long-horizon visual evolution. The code and model weights are released at https://github.com/yebo0216best/DySink.

Existing 4D-driven video diffusion models primarily target plausible generation, but faithful 4D editing requires preserving source-observed regions while synthesizing disoccluded or out-of-view content. We identify Evidence-Role Mismatch: reliable source-backed evidence, unreliable rendered cues, and unsupported regions are entangled in a single conditioning signal, causing preservation drift, ghosting, and unstable extrapolation. We propose PREX (Preserve, Reveal, Expand), a region-aware framework that decomposes the target spatiotemporal volume into Preserve, Reveal, and Expand roles according to observation support and scene extent. PREX builds observation-backed appearance cues with calibrated confidence and injects them into a frozen video diffusion backbone through a region-aware adapter, trained with proxy tasks without requiring paired edited videos. We further introduce PREBench, a diagnostic benchmark with curated edits, region-role masks, and human-aligned metrics that complement global video-quality and 4D-control evaluations. Experiments show that PREX reduces region-structured failures while maintaining strong visual quality and 4D edit control capability. Project Page: https://ricepastem.github.io/PREX-Open

Extending the generation horizon of video diffusion models to long sequences remains a long-standing and important challenge. Existing training-free approaches fall into two categories: extensions of bidirectional models, which are tightly coupled to specific architectures and suffer from quality degradation over long horizons, and autoregressive models, which accumulate drift errors due to exposure bias and tend to produce repetitive motion patterns. To address these issues, we propose a novel but simple inference-time approach for long video generation that is architecture-agnostic and requires no additional training. Our method generates long videos via overlapping sliding windows, where predicted clean samples from adjacent windows are blended via \emph{Tweedie matching} to enforce both \textbf{manifold constraint and temporal consistency} across overlap regions. \emph{Stochastic early-phase sampling} then synchronizes per-window trajectories by injecting fresh noise after each Tweedie matching correction in the high-noise phase, before transitioning to deterministic ODE sampling to preserve fine-grained visual fidelity. Applied to various video generation models, our method generates videos several times longer than the native window length while outperforming both training-free and autoregressive baselines in temporal consistency and visual quality, and further extends to audio-video joint generation and text-to-3DGS without any fine-tuning.

Diffusion models provide powerful priors for zero-shot video inverse problems, but their real-time deployment is hindered by two inefficiencies: high initial latency caused by holistic video restoration, and low throughput resulting from multiple VAE passes to enforce measurement consistency in pixel space. To overcome these limitations, we propose Autoregressive Video Inverse problem Solver (AVIS). The AVIS framework leverages autoregressive video diffusion models to restore videos in a streaming manner, naturally eliminating latency bottlenecks. Specifically, AVIS initializes reverse diffusion with a measurement-consistent estimate, reducing the required sampling steps. Compared to leading non-autoregressive solvers, AVIS drastically reduces initial latency from 114s to 4s and increases throughput from 0.71 to 1.18 FPS while achieving superior restoration quality. We further introduce a highly accelerated variant, dubbed AVIS Flash, that enforces measurement consistency solely on the first chunk. AVIS Flash substantially boosts throughput to 5.91 FPS on a single RTX 4090 GPU while maintaining competitive performance and achieving a favorable efficiency-performance trade-off, paving the way toward real-time deployment.

Long-horizon video generation suffers from two intertwined issues. First, there is drift, where video quality degrades over time. Second, there are continuity issues which manifest as object permanence issues, or improperly rendering transient content (e.g., an object that appears in non-consecutive frames changing color/style). Recent work has focused on autoregressive distillation techniques that attack both problems simultaneously. We instead choose to focus on drift directly and introduce \textbf{Anchored Tree Sampling (ATS)}: a training-free inference-time scheduler that replaces left-to-right rollout with sparse-to-dense, anchor-bounded imputation organized as a tree. A root call produces sparse anchors over the full horizon, recursive refinement generates intermediate anchors, and final leaf spans are synthesized between neighboring anchors. This reduces the critical path from $K$ sequential rollout steps to $L+1$ tree-hierarchical steps and converts horizon-compounding drift into anchor-bounded drift. We focus on V2V generation in the \emph{static-camera} regime, where sparse anchors over the horizon are well approximated by the dense conditioning signal, and the base model can produce them without retraining. We evaluate ATS against two contemporary autoregressive baselines on Wan $2.1$ $+$ VACE, across five conditioning modalities (inpainting, outpainting, edge, pose, depth). We show that ATS outperforms both competitors in overall quality, as well as in drift prevention. We additionally demonstrate stable $\geq 40$-minute generation on LTX-$2.3$ across the same five modalities. We conclude by proposing a path forward to extend ATS to arbitrarily long T2V generation, as well as the dynamic-camera and multi-shot regimes.

Recent diffusion models achieve strong photorealism and fluency in video generation, yet remain fragile under abstract, sparse or complex conditions, leading to poor performance in professional production workflows such as storyboard sketches and clay render conditions. Existing video generation models, either inject conditions through adapters or couple a generic vision-language model (VLM) within a diffusion backbone, leaving a capability gap and failing to produce the videos that align with the user's creative intent. We present CogOmniControl, a reasoning-driven framework that factorizes controllable video generation into creative intent cognition and generation. Specifically, we train a specialized CogVLM using authentic anime production data. Compared to generic VLMs, it generates more professional and clear outputs, accurately cognizing user creative intent from sparse and abstract conditions and tuning these cues into dense reasoning output. Besides, CogOmniDiT unifies the controls from various conditions through in-context generation and is aligned to the CogVLM reasoning outputs via reinforcement learning. Furthermore, leveraging CogVLM's robust capability in guiding video generation, we release its potential in planning specific evaluators and enable a Best-of-N selection for the generated videos. This integration transforms the entire framework into a closed-loop "harness-like" architecture. We further introduce CogReasonBench and CogControlBench, built from professional workflows data that carry genuine creative intent rather than simulated ones. Experiments on two benchmarks show that CogOmniControl surpassed the existing open-source models. The project website: https://um-lab.github.io/CogOmniControl/

Recent generative video models achieve impressive visual quality but remain constrained by limited physical consistency and controllability. Existing video generation methods provide minimal physical control, and single-image-to-3D conversion approaches often suffer from object interpenetration. Furthermore, physics-based scene-level 3D generation methods exhibit spatial misalignment, stylized artifacts, and inconsistencies with the input data, restricting their use in realistic interactive video synthesis. We propose TelePhysics, a training-free framework that converts a single image into a physically consistent and controllable video through holistic scene-level 3D reconstruction. By representing the full scene geometry in a unified spatial coordinate system, TelePhysics resolves object penetration and alignment ambiguity. Unlike prior methods, this formulation enables accurate scenelevel multi-object interactions and introduces richer, complex control types for advanced mechanicsbased manipulation. By decoupling simulation from rendering, TelePhysics bypasses latency-heavy priors, achieving real-time physical interaction previews paired while preserving photorealistic visual fidelity. Experimental results demonstrate that TelePhysics substantially outperforms prior methods in physical fidelity, spatial coherence, and controllability. The open-source code is available at https://github.com/xinzhang007/TelePhysics.

Image-to-video models often generate videos that remain overly static, compared to text-to-video models. While prior approaches mitigate this issue by weakening or modifying the image-conditioning signal, they often require additional training or sacrifice fidelity to the reference image. In this work, we identify reference-frame dominance as a key mechanism behind motion suppression. We observe that non-reference frames in I2V models allocate excessive self-attention to reference-frame key tokens, causing reference information to be over-propagated across time and suppressing inter-frame dynamics. Based on this finding, we propose DyMoS (Dynamic Motion Slider), a training-free and model-agnostic method that rebalances the attention pathway from generated frames to the reference frame during initial denoising steps. DyMoS leaves both the input image and model weights unchanged and introduces a single scalar parameter for continuous control over motion strength. Experiments across multiple state-of-the-art I2V backbones demonstrate that DyMoS consistently improves motion dynamics while maintaining visual quality and fidelity to the reference image.

World simulators can provide safe and scalable environments for training Physical AI systems before real-world deployment. Large video generation models are emerging as a promising basis for such simulators because they can generate diverse and realistic visual futures. However, using them as world simulators requires physically faithful video continuations, namely, generated videos that preserve the physical state implied by the conditioning input, and evolve in ways consistent with basic physical principles. We propose PhyWorld, a video generation world model designed to produce temporally coherent and physically faithful scene continuations through two-stage post-training. In the first stage, we improve video-to-video continuation with flow matching fine-tuning, encouraging stable visual attributes and coherent motion dynamics across frames. In the second stage, we align generated dynamics with physical principles using Direct Preference Optimization (DPO) over physics preference pairs, guiding the model toward outputs with higher physical plausibility. To evaluate PhyWorld, we use both standard video-quality benchmarks and a dedicated physical-faithfulness benchmark with per-law scoring. Experiments show that PhyWorld improves video consistency, achieving an average score of 0.769 on VBench compared with 0.756 or below for state-of-the-art baselines. PhyWorld also improves physical plausibility, reaching an average score of 3.09 on our physical-faithfulness benchmark compared with 2.99 for the strongest baseline. These results suggest that post-training large video generation models with continuation and physics-preference signals can make them more effective world simulators for Physical AI.

Recent video generative models have greatly improved the realism of AI-generated videos, yet their outputs still exhibit artifacts such as temporal inconsistencies, structural distortions, and semantic incoherence. While Multimodal Large Language Models (MLLMs) show strong visual understanding capabilities, their ability to perceive and reason about such artifacts remains unclear. Existing benchmarks often lack systematic evaluation of artifact-aware perception and fine-grained diagnostic reasoning, especially across diverse AI-generated video domains beyond photorealistic content. To address this gap, we introduce Artifact-Bench, a comprehensive benchmark for evaluating MLLMs on AI-generated video artifact detection and analysis. We first establish a three-level hierarchical taxonomy of realism artifacts, covering photorealistic, animated, and CG-style videos. Based on this taxonomy, Artifact-Bench defines three complementary tasks: real vs. AI-generated video classification, pairwise realism comparison, and fine-grained artifact identification. Experiments on 19 leading MLLMs reveal substantial limitations in artifact perception and reasoning, with many models approaching random or even below-random performance in challenging settings. We further observe significant misalignment between MLLM judgments and human perceptual preferences, highlighting their limited reliability as general evaluators for AI-generated video realism.

Recent video editing models have converged on a unified conditioning design: a single diffusion transformer jointly consumes text, source video, and reference images, and one set of weights covers replacement, removal, style transfer, and reference-driven insertion. The design is flexible, but it assumes that the user already provides model-ready text, reference images, and spatial grounding for local edits, which real requests often omit. We present Aurora, an agentic video editing framework that pairs a tool-augmented vision-language model (VLM) agent with a unified video diffusion transformer. The VLM agent maps a raw user request to a structured edit plan aligned with the transformer's conditioning channels, thereby resolving textual and visual underspecification before generation. We train the VLM agent with supervised data for complete edit planning and reference-image selection, together with preference pairs for robust tool use and instruction refinement. We introduce AgentEdit-Bench to evaluate agent-enhanced video editing under textual and visual underspecification. Experiments on AgentEdit-Bench and two existing video editing benchmarks show that Aurora improves over instruction-only baselines and that the VLM agent transfers to compatible frozen video editing models. Project page: https://yeates.github.io/Aurora-Page

We present LongLive-2.0, an NVFP4-based parallel infrastructure throughout the full training and inference workflow of long video generation, addressing speed and memory bottlenecks. For training, we introduce sequence-parallel autoregressive (AR) training, instantiated as Balanced SP, which co-designs the efficient teacher-forcing layout with SP execution by pairing clean-history and noisy-target temporal chunks on each rank, enabling a natural teacher-forcing mask with SP-aware chunked VAE encoding. Combined with NVFP4 precision, it reduces GPU memory cost and accelerates GEMM computation during training, the proportion of which increases as video length grows. Moreover, we show that a high-quality infrastructure and dataset enable a remarkably clean training pipeline. Unlike existing Self-Forcing series methods that rely on ODE initialization and subsequent distribution matching distillation (DMD), LongLive-2.0 directly tunes a diffusion model into a long, multi-shot, interactive auto-regressive (AR) diffusion model. It can be further converted to real-time generation (4 to 2 denoising steps) with standalone LoRA weights. For inference on Blackwell GPUs, we enable W4A4 NVFP4 inference, quantize KV cache into NVFP4 for memory savings, and boost end-to-end throughput with asynchronous streaming VAE decoding. On non-Blackwell GPU architectures, we deploy SP inference to match the speed on Blackwell GPUs, while the quantized KV cache can lower inter-GPU communication of SP. Experiments show up to 2.15x speedup in training, and 1.84x in inference. LongLive-2.0-5B achieves 45.7 FPS inference while attaining strong performance on benchmarks. To our knowledge, LongLive-2.0 is the first NVFP4 training and inference system for long video generation.

Autoregressive video generation has improved rapidly in visual fidelity and interactivity, but it still suffers from long-term inconsistency and memory degradation. Most existing solutions either compress historical frames using predefined strategies or retrieve keyframes based on coarse implicit attention signals, both of which fail to handle evolving prompts with shifting entity references, leading to identity drift, character duplication, and attribute loss. To address this, we propose IAMFlow, a training-free identity-aware memory framework that explicitly models and tracks persistent entity identities, enabling consistent generation across prompt transitions. Specifically, an LLM extracts entities with visual attributes from each prompt and assigns unique global IDs for identity-aware memory, while a VLM asynchronously verifies and refines attributes from rendered frames, enabling explicit entity tracking in place of implicit similarity-based matching. To keep the proposed framework computationally practical, we design a systematic inference acceleration pipeline, including asynchronous visual verification, adaptive prompt transition, and model quantization, which achieves faster generation than existing baselines. Furthermore, we introduce NarraStream-Bench, a benchmark for narrative streaming video generation that features 324 multi-prompt scripts spanning six dimensions and a three-dimensional evaluation protocol that integrates both traditional metrics and multimodal large language model-based assessments. Extensive experiments show that IAMFlow, despite being training-free, achieves the best overall performance on NarraStream-Bench, outperforming the strongest baseline by 2.56 points, while achieving a 1.39$\times$ speedup over the most efficient baseline in the 60-second multi-prompt setting.

Video generation models produce visually compelling results but systematically violate physical commonsense -- on VideoPhy-2, the best model achieves only 32.6% joint accuracy. We identify a specification bottleneck: text prompts are lossy compression of the physical world, omitting the parameters that fully determine dynamics, and no amount of model scaling can recover what was never specified. From this diagnosis we derive three properties that physics conditioning must satisfy -- sufficiency, dynamism, and verifiability -- and show that no existing approach satisfies all three. We present NEWTON, in which video generation is demoted from the system output to one action inside an agent's toolbox: a learned planner orchestrates physics-aware tools (keyframe generation, scientific computation, prompt refinement) to construct rich conditioning, and a verifier closes the loop for iterative re-planning. The planner is the sole trainable component, optimized on-policy via Flow-GRPO inside the live multi-turn loop. On VideoPhy-2, NEWTON improves joint accuracy from 21.4% to 29.7% on LTX-Video and from 30.7% to 37.4% on Veo-3.1, without modifying either generator. Our project page: https://Newton026.github.io/newton

Recent advances in autoregressive video diffusion have enabled sequential and streaming video generation. However, long-horizon generation requires increasingly large KV caches, making efficient compression without sacrificing quality challenging. Existing methods mostly select historical frames based on attention scores, but their context decisions remain coarse. When multiple frames are generated in the same chunk, these methods often apply a shared history selection to the whole chunk, score historical frames solely by attention, and assign head-wise budgets either uniformly or by attention-pattern heuristics rather than explicit head-importance estimation. We show that frames within the same generated chunk can depend on distinct historical frames, that the same historical frame can receive different attention scores as its relative temporal distance to the current frames changes, and that masking different heads induces unequal generation degradation. Motivated by these findings, we propose \textbf{Focused Forcing}, a training-free KV selection method that focuses cached history along both generated-frame and head dimensions. For each generated frame, Focused Forcing preserves the most relevant and distinctive historical frames by combining attention scores with diversity scores of historical frames, while assigning larger budgets to heads with higher estimated importance. Across multiple autoregressive generation paradigms, Focused Forcing achieves up to $\textbf{1.48}\times$ end-to-end acceleration without training, while \textbf{improving visual quality and text alignment}. \textit{Our code will be released on GitHub.}

Stable Audio 3 is a family of fast latent diffusion models (small, medium, large) for variable-length audio generation and editing. Since our models can generate several minutes of audio, variable-length generations are key to avoid the cost of producing full-length generations for short sounds. We also support inpainting, enabling targeted audio editing and the continuation of short recordings. Our latent diffusion models operate on top of a novel semantic-acoustic autoencoder that projects audio into a compact latent space, enabling efficient diffusion-based generation while preserving audio fidelity and encouraging semantic structure in the latent. Finally, we run adversarial post-training to both accelerate inference and improve generation quality, reducing the number of inference steps while improving fidelity and prompt adherence. Stable Audio 3 models are trained on licensed and Creative Commons data to generate music and sounds in less than a 2s on an H200 GPU and less than a few seconds on a MacBook Pro M4. We release the weights of small and medium, that can run on consumer-grade hardware, together with their training and inference pipeline.

In video generation models, particularly world models, training large-scale video diffusion Transformers (such as DiT and MMDiT) poses significant computational challenges due to the extreme variance in sequence lengths within mixed-mode datasets. Existing bucket-based data loading strategies typically rely on "equal token length" constraints. This approach fails to account for the quadratic complexity of self-attention mechanisms, leading to severe load imbalance and underutilization of GPU resources. This paper proposes \textit{AdaptiveLoad}, an integrated optimization framework consisting of two core components: (1) A dual-constraint adaptive load balancing system, which eliminates long-sequence bottlenecks by simultaneously limiting memory consumption and computational load ($B \times S^p \le M_{\text{comp}}$); (2) A fused LayerNorm-Modulate CUDA kernel, which utilizes a D-tile coalesced reduction strategy to increase throughput and alleviate memory pressure. Experimental results on the Wan 2.1 world model demonstrate that our method reduces the computational imbalance rate from 39% to 18.9%, improves peak VRAM utilization efficiency by 22.7%, and achieves an overall training throughput increase of 27.2%.

Video style transfer aims to render videos in a target artistic style while preserving content, structure, and motion. While image stylization has advanced rapidly, video stylization remains challenging due to temporal inconsistency. Most existing methods stylize frames or keyframes and enforce consistency via heuristic temporal propagation, which is brittle under occlusions, disocclusions, and long-term motion, leading to drift and flickering artifacts. We argue that a fundamental bottleneck lies in the lack of large-scale triplet data and a principled training paradigm that jointly models and disentangles style, content, and motion.To address this, we introduce VISTA-1000, a synthetic dataset with 1,000 styles and motion-aligned triplets of style reference, clean video, and stylized video, and propose a diffusion-transformer-based in-context video style transfer framework with a lightweight style adapter for robust style extraction. Extensive experiments demonstrate SOTA performance in style fidelity, temporal consistency, and content preservation.

Diffusion-based \textit{image-to-video} (I2V) generation has become a central direction in generative models by turning a reference image, with optional conditions, into a temporally coherent video. Compared with broader video generation settings, this task places stricter demands on content consistency, identity preservation, and motion coherence. Although the literature grows rapidly, existing works mostly discuss I2V generation within broader topics and still lack a dedicated taxonomy together with a systematic analysis centered on this field. This work addresses that gap by treating diffusion I2V generation as a standalone subject. It first reviews the task formulation, model architectures, datasets, and evaluation metrics, and then organizes existing methods through a taxonomy based on architecture and training paradigm. It further distills four core designs, namely condition encoding, temporal modeling, noise prior design, and spatial-temporal upsampling, and discusses representative application scenarios together with major open challenges.

Latent diffusion models have emerged as the dominant paradigm for many generation tasks including audio generation such as text-to-audio, text-to-music and text-to-speech. A key component of latent diffusion is an autoencoder (VAE) that compresses high-dimensional signals into a low frame rate continuous representation that is conducive for downstream prediction. Regularizing these VAEs is challenging, as there is a trade-off between over-regularized (poor output quality) and under-regularized (difficult to predict) latent representations. We propose a framework for studying this trade-off through compression and train Audio VAEs at specific bitrates via target-KL regularization. This allows direct comparison to well-studied discrete neural audio codec models, and the construction of rate-distortion curves for audio VAEs. We evaluate the impact of target-KL regularization on text-to-sound generation and find that sweeping compression rates is helpful in identifying the optimal generation setting.

Trajectory-controlled video generation has become essential for controllable video generation. While current methods perform well under small-view camera motions, they degrade significantly with large-view motions. Existing solutions for extreme-view synthesis typically require dedicated video pairs, demanding substantial annotation effort. To address these limitations, we propose Dynamic Extreme VIew Synthesis-GRPO (DEVIS-GRPO), a GRPO-based framework for trajectory-controlled video generation, the first online policy gradient method for extreme view video generation. Central to our approach is a novel sampling strategy: Accumulative Dynamic Extreme VIew Synthesis (ADEVIS), which achieves large-view camera motions by progressively accumulating small-view increments. This method delivers two key advantages: 1) enhanced training efficiency, as it eliminates the need to warm-start the policy model by collecting expensive paired large-view videos, and 2) increased sampling diversity, achieved by flexibly varying trajectory configurations. Finally, we designed a multi-level consistency-quality reward function to select high-quality samples for model optimization. Experiments on the Kubric-4D, iPhone, and DL3DV datasets demonstrate our method's superiority. On Kubric-4D, we achieve relative improvements of 21.57% in PSNR and 7.31% in SSIM over the second-best method in non-occlusion areas. On iPhone, LPIPS is reduced by 18.56%.

We present HighSync, an end-to-end diffusion-based framework for high-fidelity lip synchronization that generates photorealistic talking-face videos aligned with arbitrary input audio. Existing approaches consistently struggle to reconcile image quality with synchronization accuracy, producing either visually degraded outputs or temporally inconsistent lip movements. HighSync addresses both challenges simultaneously and, to our knowledge, is the first lip sync model to operate natively at 512*512 resolution, positioning it as a viable solution for professional production environments such as the film and broadcast industries. Central to our approach is the identification and systematic elimination of a data leakage phenomenon that has silently undermined temporal modeling in prior work, preventing models from developing a genuine dependence on the audio signal. Comprehensive evaluations across both perceptual quality and synchronization accuracy metrics confirm that HighSync achieves state-of-the-art performance on both fronts. Source code, pre-trained models, and supplementary video results are publicly available at: https://github.com/saeed5959/high_sync

Recent advancements in generative video models demonstrate high visual fidelity, yet their integration into enterprise environments is restricted by temporal inconsistencies and severe brand misalignment. Current monolithic architectures struggle to enforce rigid brand constraints, frequently hallucinating unapproved visual assets. We introduce Genflow, a Compound AI System designed to enforce brand consistency in generative media production. Our architecture integrates a retrieval-based 'Brand DNA' extraction module to parameterize generation according to established corporate identity guidelines. Furthermore, we implement an Adversarial Multi-Agent Quality Control (QC) loop. Instead of a single-pass generation, this pipeline employs evaluator agents to iteratively critique generated frames against the extracted parameters, prompting generator models to refine outputs until a deterministic consensus is reached. By transitioning to a multi-stage, self-correcting pipeline, Genflow improved the yield of brand-compliant video generations from 42% to 89%, establishing a robust framework for scalable, enterprise-grade generative systems.

Text-to-video (T2V) generation has rapidly progressed in visual fidelity, yet its ability to faithfully represent multiple cultures within a single prompt remains underexplored. We introduce MAVEN, a multi-agent prompt refinement framework designed to improve cultural fidelity in both mono-cultural and cross-cultural T2V generation. MAVEN decomposes prompts into person, action, and location dimensions, handled by specialized agents operating in parallel or sequentially. To support systematic evaluation, we contribute a new benchmark of 243 culturally grounded prompts and 972 corresponding videos, spanning three cultures (Chinese, American, Romanian), three action categories, and both mono-cultural and cross-cultural scenarios. Evaluations combining CLIP-based metrics, VLM-as-judge assessments, and videoquality measures show that multi-agent refinement, particularly parallel specialization, significantly improves cultural relevance while preserving visual quality and temporal consistency. The dataset and code are available at https://github.com/AIM-SCU/MAVEN

Audio super-resolution (SR), also referred to as bandwidth extension (BWE), aims to reconstruct high-fidelity signals from low-resolution (LR) or band-limited (BL) observations, an inherently ill-posed task due to the ambiguity of missing high-frequency (HF) content. This survey provides a comprehensive overview of the field, with a particular focus on the paradigm shift from discriminative mapping to modern generative modeling. We first review early discriminative deep neural network (DNN) models, which formulate BWE/SR as a deterministic mapping problem and are prone to regression-to-the-mean effects and spectral over-smoothing. We then systematically review generative approaches, including autoregressive (AR) models, variational autoencoders (VAEs), generative adversarial networks (GANs), diffusion and score-based models, flow-based methods, and Schrödinger bridges. Across these approaches, we examine key design aspects, including representation domain, architecture, conditioning mechanisms, and trade-offs among reconstruction fidelity, perceptual quality, robustness, and computational efficiency. Furthermore, we discuss emerging directions involving large language models (LLMs) and multimodal foundation models, and highlight open challenges in perceptual evaluation, phase modeling, and real-world generalization. By providing a structured taxonomy and unified perspective, this survey establishes a comprehensive foundation and offers a practical roadmap for advancing BWE/SR from deterministic point estimation toward distribution-aware generative modeling.

Recent diffusion-based video generators have achieved remarkable visual fidelity and prompt controllability, yet scaling them to ultra-high-resolution (UHR) long videos remains prohibitively expensive. The difficulty is especially pronounced for long single-shot generation where a continuous scene must preserve global temporal coherence, and fine-grained spatial details without relying on clip transitions or autoregressive shot stitching. In this work, we revisit this challenge from the perspective of decoupled modeling. We argue that existing video diffusion models already encode strong local visual priors, while the main bottleneck lies in efficiently extending global spatiotemporal modeling as resolution and duration increase. Based on this insight, we propose AtlaVid, a decoupled global-local framework for efficient UHR long video generation. AtlaVid first generates a low-resolution and low-FPS global semantic proxy via temporally scaled RoPE, thereby extending the temporal horizon without increasing the training token count. Guided by this proxy, a high-resolution detail branch performs joint denoising with hierarchical locality-preserving attention. Reordered spatiotemporal windows preserve geometric locality and asymmetric global-local attention injects aligned semantic guidance and preserves the model's pretrained ability. This design enables resolution-agnostic training: the model is trained only at 720P with lightweight LoRA adaptation, yet generalizes directly to 4K and beyond for longer (>10s) video synthesis. Experiments show that AtlaVid substantially improves the efficiency of ultra-high-resolution long video generation, achieving high-quality UHR long video generation with 60.9x speed up and significantly less training cost and even better performance than native 4K video generators.

Autoregressive (AR) video diffusion is a powerful paradigm for streaming and interactive video generation. However, its reliance on softmax self-attention leads to quadratic compute complexity in sequence length and memory usage due to key-value caching, which limits its scalability to long video horizons. Existing remedies (e.g., sparse attention and KV-cache compression) reduce per-step cost but still rely on a linearly growing cache or irreversibly discard past context, and thus fail to address linear memory growth and streaming context management. To address this scalability bottleneck, we propose ARL2 (Attend Locally, Remember Linearly), a hybrid attention module that replaces quadratic cross-frame attention with a fixed-size recurrent state. We decompose self-attention into two branches: an intra-frame softmax branch for spatial detail and local dependencies, and an inter-frame gated recurrent linear branch that maintains a fixed-size state for streaming context. Our key insight is that softmax attention captures fine-grained local interactions, while a recurrent state provides controllable long-range memory. This design achieves linear-time scaling with constant memory while improving temporal consistency over the full-softmax model. To prevent noisy intermediate states from corrupting memory, we update the recurrent state only after the denoised pass. To avoid within-frame information asymmetry, all tokens share the same pre-update state rather than sequential updates. To the best of our knowledge, this is the first work to convert a pretrained AR video diffusion model into a hybrid linear attention architecture, through an efficient two-stage training scheme for AR video. With 75% of layers replaced by hybrid linear attention, the model achieves up to 2.26 wall-clock speedup and 54% memory reduction, while maintaining comparable quality with improving temporal consistency.

Autoregressive video diffusion models enable open-ended generation through local attention and KV caching. However, existing training-free long-video optimization methods mainly focus on stable extension under a single prompt, making them difficult to handle interactive scenarios involving prompt switching, old scene forgetting, and historical scene recall. We identify the core bottleneck as the functional entanglement of historical KV states: stable anchors and recent dynamics are handled by the same cache policy, leading to outdated background contamination, delayed response to new prompts, and loss of long-range memory. To address this issue, we propose Echo-Forcing, a training-free scene memory framework specifically designed for interactive long video generation with three core mechanisms: (1) Hierarchical Temporal Memory, which decouples stable anchors, compressed history, and recent windows under relative RoPE; (2) Scene Recall Frames, which compresses historical scenes into spatially structured KV representations to support long-term recall; and (3) Difference-aware Memory Decay, which adaptively forgets conflicting tokens according to the discrepancy between old and new scenes. Based on these designs, Echo-Forcing uniformly supports smooth transitions, hard cuts, and long-range scene recall under a bounded cache budget. Extensive evaluations on VBench-Long further demonstrate that Echo-Forcing achieves the best overall performance in both long-video generation and interactive video generation settings. Our code is released in https://github.com/mingqiangWu/Echo-Forcing

Group Relative Policy Optimization has emerged as essential for aligning video diffusion models with human preferences, but faces a critical computational bottleneck: training a 14B parametered model typically demands hundreds of GPU days per experiment. Existing efficiency methods reduce costs through sliding window subsampling training timesteps, but fundamentally compromise optimization, exhibiting severe instability and failing to reach full trajectory performance. We present Flash-GRPO, a single-step training framework that outperforms full trajectory training in alignment quality under low computational budgets while substantially improving training efficiency. Flash-GRPO addresses two critical challenges: iso-temporal grouping eliminates timestep-confounded variance by enforcing prompt-wise temporal consistency, decoupling policy performance from timestep difficulty; temporal gradient rectification neutralizes the time-dependent scaling factor that causes vastly inconsistent gradient magnitudes across timesteps. Experiments on 1.3B to 14B parameter models validate Flash-GRPO's effectiveness, demonstrating substantial training acceleration with consistent stability and state-of-the-art alignment quality.

Human-centric video customization, particularly at the garment level, has shown significant commercial value. However, existing approaches cannot support low-latency and interactive garment control, which is crucial for applications such as e-commerce and content creation. This paper studies how to achieve interactive multi-garment video customization while preserving motion coherence using only single-garment video data. We present FashionChameleon, a real-time and interactive framework for human-garment customization in autoregressive video generation, where users can interactively switch garment during generation. FashionChameleon consists of three key techniques: (i) Instead of training on multi-garment video data, we train a Teacher Model with In-Context Learning on a single reference-garment pair. By retaining the image-to-video training paradigm while enforcing a mismatch between the reference and garment image, the model is encouraged to implicitly preserve coherence during single-garment switching. (ii) To achieve consistency and efficiency during generation, we introduce Streaming Distillation with In-Context Learning, which fine-tunes the model with in-context teacher forcing and improves extrapolation consistency via gradient-reweighted distribution matching distillation. (iii) To extend the model for interactive multi-garment video customization, we propose Training-Free KV Cache Rescheduling, which includes garment KV refresh, historical KV withdraw, and reference KV disentangle to achieve garment switching while preserving motion coherence. Our FashionChameleon uniquely supports interactive customization and consistent long-video extrapolation, while achieving real-time generation at 23.8 FPS on a single GPU, 30-180$\times$ faster than existing baselines.

Causal autoregressive video diffusion models support real-time streaming generation by extrapolating future chunks from previously generated content. Distilling such generators from high-fidelity bidirectional teachers yields competitive few-step models, yet a persistent gap between the history distributions encountered during training and those arising at inference constrains generation quality over long horizons. We introduce the Real-time Autoregressive Video Extrapolation Network (RAVEN), a training-time test framework that repacks each self rollout into an interleaved sequence of clean historical endpoints and noisy denoising states. This formulation aligns training attention with inference-time extrapolation and allows downstream chunk losses to supervise the history representations on which future predictions depend. We further propose Consistency-model Group Relative Policy Optimization (CM-GRPO), which reformulates a consistency sampling step as a conditional Gaussian transition and applies online Reinforcement Learning (RL) directly to this kernel, avoiding the Euler-Maruyama auxiliary process adopted in prior flow-model RL formulations. Experiments demonstrate that RAVEN surpasses recent causal video distillation baselines across quality, semantic, and dynamic degree evaluations, and that CM-GRPO provides further gains when combined with RAVEN.

Generative video models are increasingly studied as implicit world models, yet evaluating whether they produce physically plausible 3D structure and motion remains challenging. Most existing video evaluation pipelines rely heavily on human judgment or learned graders, which can be subjective and weakly diagnostic for geometric failures. We introduce PDI-Bench (Perspective Distortion Index), a quantitative framework for auditing geometric coherence in generated videos. Given a generated clip, we obtain object-centric observations via segmentation and point tracking (e.g., SAM 2, MegaSaM, and CoTracker3), lift them to 3D world-space coordinates via monocular reconstruction, and compute a set of projective-geometry residuals capturing three failure dimensions: scale-depth alignment, 3D motion consistency, and 3D structural rigidity. To support systematic evaluation, we build PDI-Dataset, covering diverse scenarios designed to stress these geometric constraints. Across state-of-the-art video generators, PDI reveals consistent geometry-specific failure modes that are not captured by common perceptual metrics, and provides a diagnostic signal for progress toward physically grounded video generation and physical world model. Our code and dataset can be found at https://pdi-bench.github.io/.

Camera-controlled video generation has made substantial progress, enabling generated videos to follow prescribed viewpoint trajectories. However, existing methods usually learn camera-specific conditioning through camera encoders, control branches, or attention and positional-encoding modifications, which often require post-training on large-scale camera-annotated videos. Training-free alternatives avoid such post-training, but often shift the cost to test-time optimization or extra denoising-time guidance. We propose Warp-as-History, a simple interface that turns camera-induced warps into camera-warped pseudo-history with target-frame positional alignment and visible-token selection. Given a target camera trajectory, we construct camera-warped pseudo-history from past observations and feed it through the model's visual-history pathway. Crucially, we align its positional encoding with the target frames being denoised and remove warped-history tokens without valid source observations. Without any training, architectural modification, or test-time optimization, this interface reveals a non-trivial zero-shot capability of a frozen video generation model to follow camera trajectories. Moreover, lightweight offline LoRA finetuning on only one camera-annotated video further improves this capability and generalizes to unseen videos, improving camera adherence, visual quality, and motion dynamics without test-time optimization or target-video adaptation. Extensive experiments on diverse datasets confirm the effectiveness of our method.

Large-scale labelled driving video data is essential for training autonomous driving systems. Although simulation offers scalable and fully annotated data, the domain gap between synthetic and real-world driving videos significantly limits its utility for downstream deployment. Existing video generation methods are not well-suited for this task, as they fail to simultaneously preserve scene structure, object dynamics, temporal consistency, and visual realism, all of which are critical for maintaining annotation validity in generated data. In this paper, we present DriveCtrl, a depth-conditioned controllable sim-to-real video generation framework for realistic driving video synthesis. Built upon a pretrained video foundation model, DriveCtrl introduces a structure-aware adapter that enables depth-guided generation while preserving the scene layout and motion patterns of the source simulation, producing temporally coherent driving videos that remain aligned with the original simulated sequences. We further introduce a scalable data generation pipeline that transforms simulator videos into realistic driving footage matching the visual style of a target real-world dataset. The pipeline supports three conditioning signals: structural depth, reference-dataset style, and text prompts, while preserving frame-level annotations for downstream perception tasks. To better assess this task, we propose a driving-domain-specific knowledge-informed evaluation metric called Driving Video Realism Score (DVRS) that assesses the realism of generated videos. Experiments demonstrate that DriveCtrl consistently outperforms the base model and competing alternatives in realism, temporal quality, and perception task performance, substantially narrowing the sim-to-real gap for driving video generation.

Recent breakthroughs in video diffusion models have significantly accelerated the development of video editing techniques. However, existing methods often rely on inpainting video frames based on masked input, which requires extracting the target video mask in advance, and the precision of the segmentation directly affects the quality of the completion. In this paper, we present SEDiT, a novel one-stage video Subtitle Erasure approach via One-step Diffusion Transformer. We introduce a mask-free inference approach that enables direct erasure of the targeted subtitle. The proposed one-stage framework mitigates the sub-optimality inherent in the two-stage processing of prior models. Since subtitle removal is a localized editing task in which most pixels remain unchanged, the underlying distribution shift is minimal, making it well-suited to one-step generation under rectified flow. We empirically validate the reliability of one-step denoising and further provide a formal theoretical justification. Under the localized-editing structure of subtitle removal, the conditional optimal transport (OT) map and its induced rectified flow velocity field are Lipschitz continuous with respect to the latent variable, which underpins the theoretical feasibility of one-step sampling. To address the challenge of long-term temporal consistency, we adopt a hybrid training strategy by occasionally conditioning the model with a clean first-frame latent. This facilitates temporal continuity, allowing each segment during inference to leverage the output of its predecessor. To avoid visible seams caused by cropping and reinserting processed targets, particularly in scenarios involving substantial motion, we feed the original video directly into SEDiT. Thanks to one-step and chunk-wise streaming inference, our method can efficiently handle native 1440p video with infinite length.

Text- and image-conditioned video generation models have achieved strong visual fidelity and temporal coherence, but they often fail to generate motion governed by kinematic and geometric constraints. In these settings, object parts must remain rigid, maintain contact or coupling with neighboring components, and transfer motion consistently across connected parts. These requirements are especially explicit in articulated mechanical assemblies, where motion is constrained by rigid-link geometry, contact/coupling relations, and transmission through kinematic chains. A generated video may therefore appear plausible while violating the intended mechanism, such as rotating a part that should translate, deforming a rigid component, breaking coupling between parts, or failing to move downstream components. To evaluate this gap, We introduce MechVerse, a benchmark for mechanically consistent image-to-video generation. MechVerse contains 21,156 synthetic clips from 1,357 mechanical assemblies across 141 categories, organized into three tiers of increasing kinematic complexity: independent articulation, pairwise coupling, and densely coupled multi-part mechanisms. Each clip is paired with a structured prompt describing part identities, stationary supports, moving components, motion primitives, direction, speed/extent, and inter-part dependencies. We evaluate proprietary, open-source, and fine-tuned image-to-video models using standard video metrics, instruction-following scores, and human judgments of motion correctness and kinematic coupling. Results show that current models can preserve appearance and smoothness while failing to generate mechanically admissible motion, with errors increasing as coupling complexity grows. MechVerse provides a benchmark for measuring and improving mechanism-aware video generation from image and language inputs.

The high dynamic range (HDR) video ecosystem is approaching maturity, but the problem of upconverting legacy standard dynamic range (SDR) videos persists without a convincing solution. We propose a framework for HDR video synthesis from casual SDR footage by leveraging large-scale generative video models. We introduce a Multi-Exposure Video Model (MEVM) that can predict exposure-bracketed linear SDR video sequences from a single nonlinear SDR video input. We further propose a learnable Video Merging Model (VMM) that merges the predicted exposure-bracketed video into a high-quality HDR sequence while preserving detail in both shadows and highlights. Extensive experiments, quantitative and qualitative evaluation, and a user study demonstrate that our approach enables robust HDR conversion for in-the-wild examples from casual consumer videos and even iconic films. Finally, our model can support HDR synthesis pipelines built upon existing SDR generative video models. Output HDR videos can be viewed on our supplementary webpage: sdr2hdrvideo.github.io

Diffusion-based video generation has advanced substantially in visual fidelity and temporal coherence, but practical deployment remains limited by the quadratic complexity of full attention. Training-free sparse attention is attractive because it accelerates pretrained models without retraining, yet existing online top-$p$ sparse attention still spends non-negligible cost on mask prediction and applies shared thresholds despite strong head-level heterogeneity. We show that these two overlooked factors limit the practical speed-quality trade-off of training-free sparse attention in Video DiTs. To address them, we introduce a head-wise adaptive framework with two plug-in components: Temporal Mask Reuse, which skips unnecessary mask prediction based on query-key drift, and Error-guided Budgeted Calibration, which assigns per-head top-$p$ thresholds by minimizing measured model-output error under a global sparsity budget. On Wan2.1-1.3B and Wan2.1-14B, our method consistently improves XAttention and SVG2, achieving up to 1.93 times speedup at 720P while maintaining competitive video quality and similarity metrics.

Autoregressive video diffusion models support real-time synthesis but suffer from error accumulation and context loss over long horizons. We discover that attention heads in AR video diffusion transformers serve functionally distinct roles as local heads for detail refinement, anchor heads for structural stabilization, and memory heads for long-range context aggregation, yet existing methods treat them uniformly, leading to suboptimal KV cache allocation. We propose Head Forcing, a training-free framework that assigns each head type a tailored KV cache strategy: local and anchor heads retain only essential tokens, while memory heads employ a hierarchical memory system with dynamic episodic updates for long-range consistency. A head-wise RoPE re-encoding scheme further ensures positional encodings remain within the pretrained range. Without additional training, Head Forcing extends generation from 5 seconds to minute-level duration, supports multi-prompt interactive synthesis, and consistently outperforms existing baselines. Project Page: https://jiahaotian-sjtu.github.io/headforcing.github.io/.

Interactive real-time autoregressive video generation is essential for applications such as content creation and world modeling, where visual content must adapt to dynamically evolving event conditions. A fundamental challenge lies in balancing reactivity and stability: models must respond promptly to new events while maintaining temporal coherence over long horizons. Existing approaches distill bidirectional models into autoregressive generators and further adapt them via streaming long tuning, yet often exhibit persistent drift after condition changes. We identify the cause as conditional bias, where the teacher may provide condition-aligned but trajectory-agnostic guidance, biasing generation toward locally valid yet globally inconsistent modes. Inspired by Trust Region Policy Optimization, we propose Delta Forcing, a simple yet effective framework that constrains unreliable teacher supervision within an adaptive trust region. Specifically, Delta Forcing estimates transition consistency from the latent delta between teacher and generator trajectories, and uses it to balance teacher supervision with a monotonic continuity objective. This suppress unreliable teacher-induced shifts while preserving responsiveness to new events. Extensive experiments demonstrate that Delta Forcing significantly improves consistency while maintaining event reactivity.