Overview

Video diffusion models synthesize or transform frame sequences by denoising video latents, often in diffusion transformers, conditioned on text, images, reference clips, or structured controls. Researchers use temporal attention, autoregressive chunking, and pose, camera, depth, or physics cues to improve motion, identity, geometry, and interaction; related systems support editing, restoration, avatars, and world-model prediction. Evaluation spans prompt alignment and visual quality, using benchmarks such as VBench and FVD, as well as controllability, physical plausibility, and downstream action performance. A central challenge is balancing fidelity and precise control over long videos with the speed and memory needed for streaming, while limiting drift and physically implausible events.

Latest papers

812 papers72 papers in the last four weeks, down 36% on the four weeks before. 0.5% of all new papers.

Oct 19, 2026

VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Token

Diffusion-based video matting has recently emerged as a leading approach, leveraging video diffusion priors from internet-scale pretraining to achieve strong zero-shot generalization from limited synthetic data. However, the heavy compute of these diffusion backbones limits inference to a short chunk of just 6 to 24 frames, so even ordinary videos span multiple chunks. Because each chunk is matted independently from coarse mask guidance, existing methods fail to maintain consistency across chunk boundaries, producing flickering edges and discontinuous transparency. We address this with a mask-guided diffusion-matting framework built on two ideas. The RGB and mask conditioning are represented as separate token streams that interact through full spatio-temporal self-attention, letting the model adaptively combine fine RGB texture with coarse mask layout rather than committing to a fixed channel-wise fusion. On top of this, we propose a novel pair of learnable embeddings, the preserve token and the refine token, that act as a per-frame conditioning interface and enable a sliding-window inference scheme in which already-generated mattes are propagated across chunk boundaries for local temporal consistency, complemented by a globally shared reference anchor at every chunk to prevent error accumulation in long sequences without any additional training. Across standard video matting benchmarks and in-the-wild long footage, our framework outperforms prior diffusion-based methods in both fine-detail accuracy and temporal consistency, and uniquely maintains consistency across chunk boundaries on long videos.
Oct 19, 2026

Dynamics-Aware Sparse Attention for Efficient Autoregressive Video Diffusion

Autoregressive Diffusion Transformers have emerged as a powerful paradigm for long video generation. However, they are often constrained by prohibitive computational overhead and substantial memory footprints. This inefficiency arises from dense computation mechanisms that allocate uniform processing resources across all spatiotemporal regions, failing to leverage the inherent temporal redundancy of video where motion is typically confined to sparse areas. In this paper, we propose Dynamics-Aware Sparse Attention (DASA), a unified framework that shifts from dense computation to sparse, motion-driven generation to eliminate both storage and computational redundancies. Inspired by video codec principles, our approach explicitly decouples video content into dynamic and static regions. To mitigate memory bottlenecks, we selectively compress the historical KV cache by retaining only dynamic features within the generated chunk. Furthermore, to reduce computational overhead, we employ a region-level computation allocation strategy based on localized dynamics. Additionally, a dynamics-guided distillation method is introduced to enhance performance via a lightweight distillation training process. Experiments on VBench demonstrate that our approach accelerates autoregressive video generation by 2.16x, while preserving high visual fidelity.
Oct 19, 2026

RECAP: Reciprocal Sparse Attention and Hierarchical Caching for Efficient Video Diffusion

Diffusion transformers generate high-quality videos but remain computationally expensive. Sparse attention and temporal caching reduce computation within and across denoising steps, respectively, offering a natural composition. However, predictions produced with sparse attention become sources for temporal reuse, coupling their errors along the sampling trajectory. We observe that improving the accuracy of these sources recovers fidelity without changing the reuse schedule, and that sparsification errors remain directionally aligned across nearby steps, suggesting that correction information can be shared rather than recomputed. These observations motivate RECAP, a training-free method that couples sparse attention with hierarchical caching through reciprocal correction, adaptively coordinated by online error feedback. Cached dense refinement corrects omitted interactions in current sparse estimates, while fresh sparse statistics track input changes and regulate the aging correction. The resulting predictions then renew the sources for whole-step prediction reuse, allowing prediction history to advance between dense refreshes. We evaluate RECAP on multiple video diffusion models against sparse-attention methods, caching methods, and their direct compositions. On HunyuanVideo at 720p 5-second generation, RECAP achieves a 3.05x speedup and outperforms all tested direct compositions in both fidelity and inference speed.
Oct 19, 2026

Wan-Alpha++: Animating a Single RGBA Asset into Controllable Transparent Video Effects

RGBA video generation is essential for creating compositable visual assets in filmmaking, game development, and digital content production. However, generating high-quality RGB appearance alongside temporally coherent transparency remains a significant challenge. Existing text-to-RGBA video methods often entangle RGB and alpha distributions and offer limited control over the precise appearance of a target asset. In addition, incorporating image conditioning typically suffers from the emergence-disappearance effects, which fail existing first-frame or first-and-last-frame conditions due to their fully transparent boundary frames. To address these issues, we present Wan-Alpha++, a unified framework that brings controllable RGBA video generation to both text- and image-to-video diffusion backbones through a shared RGBA latent space, extending our conference version Wan-Alpha from the text-conditioned setting to image-conditioned generation with an interior reference frame. Our method first learns shiftable RGB-alpha distributions while preserving the pretrained RGB prior. Specifically, we introduce a transparency-aware bidirectional diffusion objective to produce distinguishable RGBA latent representations, and a transparency-guided mean sampler that shifts the noise distribution to improve transparency generation and enable spatial controllability. Building upon this RGBA generation prior, we introduce temporal phase reparameterization for image-conditioned effect generation. By cyclically transforming a zero-positive-zero alpha pattern into a positive-zero-positive pattern, it relocates visible content to both temporal boundaries, facilitating effective appearance conditioning with limited RGBA data.
Oct 19, 2026

FluidGen: Physics-Aware Fluid Video Generation via Motion-Guided Post-Training

Video generation models produce visually compelling motion but systematically fail on fluid phenomena, including smoke, fire, and explosion debris, where outputs lack coherent vortical structure and collapse to bland dynamics that violate the physics of real fluids. Classical solvers achieve high physical fidelity but require detailed 3D scenes and carefully tuned parameters, making them impractical for in-the-wild imagery and video generation. We address this gap by injecting a data-driven fluid prior into a pre-trained video model, trained on a mixture of high-fidelity Houdini simulation data with paired velocity-field ground truth and curated in-the-wild fluid video data for appearance diversity. The core of our approach is a dual stream architecture that predicts per-frame velocity fields alongside RGB, supervised by simulation ground truth. This auxiliary signal regularizes generation toward physically consistent motion and exposes an interpretable physical quantity that we further leverage as a reward for velocity-aware direct preference optimization. Our model produces substantially more realistic fluid dynamics sequences than strong video baselines and optimization-based methods. Project page: https://y-zheng18.github.io/FluidGen/
Oct 19, 2026

Clean Forcing: Drift-Resistant Autoregressive Video Diffusion with a Frozen Base

Autoregressive (AR) video diffusion models generate streaming video of arbitrary length by conditioning each new chunk of frames on those already generated. However, they suffer from exposure bias, where a model trained on clean context must continue from its own imperfect outputs, and errors compound until quality collapses within seconds (drift). Existing methods either require cluster-scale base retraining, apply training-free inference corrections with inconsistent results, or train frozen-base correctors for sampling and cache errors. We instead find that drift is largely deterministic and therefore learnable, as holding the noisy state fixed while varying only the history reveals a counterfactual velocity gap that is ~95% systematic across noise seeds. Based on this observation, we introduce Clean Forcing, which trains a 5.9M-parameter LoRA corrector on the frozen base model to regress this gap, with a closed-loop stage that adds DAgger-style aggregation and a drift-contraction objective to expose the corrector to its own rollouts. Because the clean histories can be the base model's own single-shot generations, training needs no real videos, and the merged corrector adds no inference cost. In 50 s text-to-video generation, Clean Forcing reduces Δ-drift from +11.3 to +1.65 without real video and to -0.16 with 40 real clips, improving on the best published result of +3.57 with over 150x less data. Extensive experiments demonstrate that Clean Forcing outperforms prior methods in aesthetic quality and data efficiency with reduced drift for long-horizon video generation.
Oct 18, 2026

ViMoLa: Discrete Motion Tokens for Language-Guided Video Generation

Language-guided video generation often captures scene content while missing the requested motion. We present ViMoLa (Video Motion Language), which predicts a motion prior from language and injects it into a frozen image-to-video backbone. A spatial vector-quantized autoencoder compresses dense trajectories into 64 tokens from a 512-entry codebook; a 28M-parameter transformer predicts the tokens from text; and zero-gated cross-attention adapters condition every frozen Wan2.2 DiT block through forward hooks. Trained on 462K videos from Something-Somethingv2 and Kinetics-400, ViMoLa raises directional accuracy from 0.287 to 0.442 (chance 0.250), with similar measured generative quality. Across-prompt shuffled codes achieve 0.283, supporting the importance of prompt-matched token content. A low-rank adapter reaches 0.428 at 1.45% additional forward-pass FLOPs, and reusing the tokenizer and language model with newly trained CogVideoX adapters raises accuracy from 0.301 to 0.398. Control remains imperfect: oracle codes reach 0.531, compound-motion gains are smaller, and the adapter does not explicitly preserve token-to-patch assignments.
Oct 18, 2026

GeoART: Geometry-Aware Refinement via Test-Time Scaling for Camera-Conditioned Video Diffusion

Camera-conditioned video diffusion models often suffer from inaccurate camera control and geometric inconsisten- cies because camera constraints and estimated geometry are only weakly enforced during denoising. Existing reme- dies often require costly geometry-aware fine-tuning or ad- ditional 3D supervision. We introduce GeoART, a training- free self-refinement framework that improves camera con- trollability through geometry-aware test-time scaling. At selected denoising steps, GeoART generates multiple can- didates within the Predict-and-Perturb cycle and eval- uates them using a geometry prior, without directly modify- ing the denoising trajectory. A confidence gate suppresses unreliable geometric evidence, while rank-calibrated prob- abilistic selection combines score magnitude and relative rank to choose the next refinement state. Experiments demonstrate consistent improvements in camera controlla- bility, multi-view consistency, and visual quality over exist- ing sampling and self-refinement methods, without training or modifying the pretrained video diffusion model
Oct 18, 2026

GeoV2V: Geometry-Grounded Video Diffusion Model for Driving Scene Generation

Novel view synthesis under trajectory changes is essential for autonomous driving simulation, yet existing methods struggle to generate consistent videos when extrapolating to unseen viewpoints such as lane shifts. Reconstruction-based approaches maintain geometric consistency but degrade in visual quality under large viewpoint changes, while reconstruct-then-restore methods often produce visually plausible frames yet suffer from geometric and temporal inconsistencies. We present GeoV2V, a geometry-grounded video-to-video diffusion framework that synthesizes temporally coherent driving videos along novel trajectories by integrating LiDAR and depth-derived geometric priors with full-video conditioning in the Wan 2.1 backbone. A key challenge in learning cross-lane transformations is the lack of synchronized multi-trajectory supervision in real-world datasets. To address this, we also introduce Para4D, a synthetic dataset capturing synchronized multi-lane driving videos within dynamic scenes, which provides explicit supervision for lane-shift generation and serves as a strong training prior for geometry-consistent view synthesis. Experiments on Waymo, nuScenes, and Para4D demonstrate state-of-the-art performance in visual quality, geometric consistency, and robustness to complex lighting dynamics under large trajectory shifts. Project page: https://xiyuche.github.io/GeoV2V/
Oct 17, 2026

BodyBridge: Grounded Body Layers for Sim-to-Real Translation of Vulnerable Road Users

Simulation-to-real driving video diffusion models can translate photorealistic scenes, but they typically represent pedestrians and cyclists, vulnerable road users (VRUs), as generic object classes. This is insufficient for vulnerable road users whose safety-relevant appearance depends on articulated pose and identity. We present BodyBridge and Grounded Body Layers (GBL), a simulator-derived, pixel-aligned representation that combines scene geometry, canonical per-pixel body correspondence, and exact VRU masks. GBL provides explicit articulation to the video diffusion model, emphasizes VRU regions during control-branch post-training, and supports additive composition with scene-geometry control at inference. It also enables per-joint evaluation of generated VRUs and a failure taxonomy for identity errors. On CARLA-to-real translation, BodyBridge improves near-range [email protected] by 22% over the released depth-conditioned baseline. Replacing normal-map body control with canonical correspondence reduces body flips by 18%, while composition reduces near-range facing flips by 27%. These results show that explicit body grounding complements scene-level control and improves pose-faithful VRU generation. All code, data, and model checkpoints will be made publicly available to support reproducible, pedestrian-aware autonomous driving research.
Oct 16, 2026

Persistent Planar Memory for Video World Models

Video world models that generate and explore environments at scale need a persistent 3D memory of what they have produced. Plane primitives offer a natural candidate: a compact set of oriented planes captures the dominant structural elements (e.g., walls, floors, facades) of indoor and outdoor scenes, while requiring only 7 parameters per primitive. We present PPMem, a video world model that adopts plane primitives as its persistent 3D memory. PPMem augments a video diffusion transformer with a planar prediction head and a Depth Token Fusion (DTF) module: each autoregressive chunk jointly produces RGB video and planar geometry, which merges into a persistent plane memory that conditions subsequent chunks via rendered depth. The resulting memory is three orders of magnitude more compact than per-pixel point clouds (~100K primitives, 5MB for a full scene), geometrically explicit (metric depth, normals, exportable surfaces), cross-view consistent (jointly predicted by the shared backbone), and adaptive (refined by the merging procedure as generation proceeds). On DL3DV and RealEstate10K, PPMem outperforms prior 3D-grounded world models in video quality while jointly producing metric-scale depth, surface normals, and exportable planar geometry, sustaining coherence across hundreds of frames under large viewpoint changes. Project page: https://hyzcluster.github.io/ppmem.
Oct 16, 2026

Video Diffusion Avatars: Taming Video Diffusion Models into Controllable 3D Avatars

Video diffusion models (VDMs) trained on internet-scale data encode a rich prior on appearance, motion, and physics that is difficult to replicate with explicit 3D rendering and costly 3D data capture. We introduce Video Diffusion Avatars (VDAs), a method that creates photo-realistic, multi-view consistent avatars from a single image by fine-tuning a VDM for precise control over camera viewpoint, facial expression, and upper-body pose. Departing from 3D morphable models (3DMMs), VDAs use a learned expression space that lifts the expressiveness ceiling of 3DMMs, e.g., for the tongue and mouth interior. However, learned expression codes exhibit heavy viewpoint bias, as our DISTRACT benchmark reveals. We show that this bias can be removed through disentanglement, enabling their use as driving signals for multi-view consistent avatars. Camera and upper-body pose are controlled jointly via proxy-mesh renderings, giving independent control over expression, pose, and viewpoint. A camera-triplet training strategy makes the approach efficient: our models train for 34 hours on a single A100 using only public multi-view data. VDAs outperform state-of-the-art single-image 3D avatars on novel-view reenactment, reducing face LPIPS by 29% and improving perceptual quality (JOD) by 66% on BecomingLit. A static 3DGS reconstruction probe further confirms that our generations are the most multi-view consistent. Built on a video prior, VDAs produce effects beyond the reach of explicit renderers, including physically plausible hair dynamics, shadows, reflections, and refractions.
Oct 16, 2026

I2V-DETACH: Source Grounding Detachment for Unauthorized Image-to-Video Generation

Recent diffusion models have advanced image-conditioned video generation, but their accessibility raises risks of unauthorized manipulation. Image immunization has emerged as a promising defense for Image-to-Video (I2V) protection, yet existing methods often rely on degrading visual quality or temporal coherence. In this work, we revisit I2V immunization as reducing source-content preservation, shifting the focus from output degradation to weakening source-conditioned content propagation. To this end, we propose I2V-DETACH, an image immunization method that prevents unauthorized I2V generation through source grounding detachment. I2V models preserve source content by propagating source-image cues into video latents during denoising. Accordingly, I2V-DETACH disrupts this process in two complementary ways: it suppresses source-video grounding to reduce source cue propagation, and strengthens intra-video interactions to shift denoising toward video-side dynamics with reduced reliance on the source image. To obtain informative optimization signals, we initialize auxiliary video latents with source-oriented states that remain within the source-conditioned generation regime while deviating from source-preserving trajectories. Extensive experiments across diverse diffusion architectures show that I2V-DETACH substantially reduces source-content preservation, improves over prior immunization methods, and exhibits strong black-box transferability.
Oct 16, 2026

Sink-Sparse Hybrid Attention for Efficient Inference in T2I/T2V Diffusion Transformers

Diffusion Transformers (DiTs) have become a common backbone for high-quality text-to-image and text-to-video generation, but repeated dense self-attention remains a major inference bottleneck as token sequences grow and the same operator is invoked across many denoising steps. Our study suggests that dense attention in pretrained DiTs contains reusable layer/head-specific high-mass structure that can be identified from a small offline calibration set and reused across future generations. This leads to Sink-Sparse Hybrid Attention, a train-free attention replacement that reuses static sink masks and supplements them with a lightweight pooled-KV branch, while leaving model weights and sampling schedules unchanged. On Wan2.1-T2V-1.3B, the method attains 1.54×\times end-to-end speedup while maintaining generation quality. On Z-Image-Turbo, it attains 1.57×\times end-to-end speedup with competitive GenEval2 performance. These results indicate that reusable importance-aware sparse support can provide an effective train-free acceleration path for both T2V and T2I DiTs.
Oct 15, 2026

OmniXRef: A Unified Framework for Multimodal Multi-Reference Video Editing

Creative intent in video editing can be expressed through text instructions (e.g., "remove the people"), image references (e.g., subject appearance), video references (e.g., camera movement), or a combination of the above. We introduce OmniXRef, a unified framework for multimodal multi-reference video editing that allows users to combine variable numbers of text, image, and video references to control diverse attributes of a source video. We first devise a unified input design with a reference-bank interface and introduce R2R^2-RoPE to support generalization to more references. Then our model, OmniXRef-Model, uses a single-stream diffusion transformer to jointly process the source video and multimodal context from an MLLM in an interleaved sequence. To address the scarcity of training data for video-reference editing, we construct OmniXRef-Data using synthesis pipelines for different editing tasks and combinations of references. We use curriculum learning to progressively increase task complexity and resolution. To support systematic evaluation of video as a control modality, we introduce OmniXRef-Bench, a benchmark for video-reference attribute transfer across Camera, Motion, Content, Style, and Visual Effects (VFX). Comprehensive evaluations show that our OmniXRef-Model outperforms task-specific baselines and commercial models.
Oct 15, 2026

SoftCache: Adaptive Step Size Caching for Diffusion Transformers

Diffusion and flow models generate a sample over iterative steps, each a full pass through a large network, and that pass dominates generation cost. Training-free caching accelerates them by reusing activations from recent evaluations instead of recomputing them. Dynamic caching methods decide when to refresh by accumulating a cheap probe of the network input into a staleness signal and spending a fresh evaluation once it crosses a threshold. They perform hard caching: the signal is a binary switch between refresh and reuse, and every step, fresh or cached, advances by the prescribed step size. We observe that the signal carries much more than this switch. How far it sits below the threshold, its headroom, predicts the error induced by reuse. We propose SoftCache, which performs soft caching: instead of a binary switch, the headroom sets how far each cached step goes. Where the headroom is large, the sampler jumps past several scheduled nodes, and as a refresh nears, its steps shrink back toward the original step. These adaptive jumps may land between the scheduled nodes, taking the sampler off its fixed grid. SoftCache requires no training and adds no per-backbone settings: on FLUX for images and on HunyuanVideo and Wan2.1 for video, at speedups from 3.5×3.5\times to 6.9×6.9\times, it is ahead of every baseline on all four fidelity metrics.
Oct 15, 2026

Perceive, Predict, and Plan: a Unified World Action Model for Egocentric Dexterous Manipulation

We introduce P3^{3}-WAM, a unified world action model for egocentric dexterous manipulation that learns 3D perception, action-conditioned prediction, and goal-directed planning through three complementary modes: Perceive, which reconstructs articulated hand motion and camera trajectories from video; Predict, which generates visual futures under prescribed actions; and Plan, which jointly generates action trajectories and their anticipated visual consequences conditioned on an initial image, a language instruction, and an optional visual goal. Our model integrates a compact, structured 3D action representation into a pretrained video diffusion transformer through a mixture-of-transformers architecture with shared base weights, modality-specific adaptation, and bidirectional visual-action attention. Randomized modality visibility lets video and 3D actions serve as either conditioning context or prediction targets across the three modes. Experiments show that unified training improves all three capabilities over single-mode training, including gains in planning quality under a vision-language model evaluation. Through unified training, our model outperforms specialized baselines on 3D hand reconstruction and action-conditioned video prediction across four datasets, while producing goal-directed manipulation videos with stronger instruction adherence and action progression than video-generation baselines. Project Page: https://p3-wam.pages.dev/
Oct 15, 2026

Velox: Spatial Memory Makes World Models Efficient

Video world models turn actions and camera motion into long visual rollouts, making video a natural interface for interactive simulation. Such interaction requires responsive generation, yet coherent rollouts must retain visual memory of what has already been generated. Chunked video diffusion models usually keep this context in growing KV caches, so later chunks become progressively slower. To address this bottleneck, we introduce Velox, a training-free attention router for region-level context selection using frustum slab correspondences from streaming geometry. Velox represents historical blocks as camera frustum slabs with depth intervals and obtains current-block depth through recency-priority multiwarp. For well-covered regions, frustum slab overlap softly modulates each layer's model-derived block importance. Uncovered regions retain the model-derived route without geometric bias. The fixed token-block granularity matches FlashAttention-style tiled execution and reduces the attention term that grows with rollout length. Frustum slab construction runs asynchronously with denoising on the same H200. At chunk 48, Velox accelerates attention by 4.36×4.36\times and denoising by 2.43×2.43\times. On the full 158-case WBench navigation split, it remains within 0.15 API-free average points of dense LingBot-Fast-Cam. Please see our project page at https://velox-world.com/velox/ for more details.
Oct 15, 2026

Sparse Attention for Video Generation Acceleration via Growing Sparsity and Reduced Search

Diffusion Transformers (DiTs) dominate video generation but suffer from high inference cost due to the quadratic complexity of self-attention. Sparse attention mitigates this by skipping unnecessary computation in the attention matrix, but existing methods apply a fixed sparsity ratio throughout denoising and struggle to accurately determine which computations to skip. To address these limitations, we analyze attention dynamics through attention entropy and mean attention distance, and find that attention becomes increasingly concentrated across denoising steps. We further observe that blocks selected early continue to capture a large fraction of the total attention weight at later steps. These findings suggest two acceleration opportunities: the sparsity ratio should grow as denoising progresses, and earlier search results can be reused to reduce the search space. Based on these findings, we propose Growing Sparsity and Reduced Search (GSRS), a training-free framework that retains more computation early and progressively reduces it later, while restricting subsequent searches to an early candidate set to reduce search overhead. GSRS adopts Top-p selection for adaptive sparsity control, supported by a precise evaluation scheme that computes exact attention weights within the FlashAttention computation at negligible overhead. Experiments on HunyuanVideo and Wan 2.1 validate that GSRS achieves 1.55×\times--2.21×\times speedup while maintaining generation quality comparable to full attention.
Oct 14, 2026

StreamErase: Towards Ultra-Fast Streaming Video Erasing

Current video object removal methods suffer from prohibitive latency (minutes per video) and non-causal processing that prevents real-time interaction. We present StreamErase, enabling sub-second latency with up to 26 FPS streaming generation on a single GPU. Our approach first augments a text-to-video model with mask input to generate high-quality videos that erase target objects and synthesize coherent backgrounds, but does not perform inference on the fly. As such, we progressively transfer the erasing capability from a powerful bidirectional foundation model to an efficient unidirectional streaming eraser, enabling long-horizon erasing without sacrificing visual fidelity. Several key challenges arise when generating videos over long, potentially infinite horizons: (1) bridging the domain gap from training on finite length and extrapolating to long horizons, (2) sustaining high quality by preventing error accumulation, and (3) maintaining fast inference, without increasing computational cost as the context grows. To address these challenges, we carefully designed sliding-window causal attention with attention sinks. By incorporating self-rollout with attention sinks and KV cache rolling during training, we simulate inference-time extrapolations with a fixed context window, enabling constant-speed generation of arbitrarily long videos. Finally, we establish a dedicated benchmark, LongErase, for streaming long-video object removal. Our model achieves strong performance compared to existing video object erasing models and generalizes well to real-world video scenarios, uniquely enabling infinite-length streaming. With StreamErase, users can input a mask and see the erasing results unfold in real-time, delivering a truly interactive experience.
Oct 14, 2026

StateCraft: Augmenting Video World Models with Abstract-State Reasoning

Video world models aim to simulate complex environments with realistic appearance and dynamics. Maintaining consistency over long interactions requires tracking past actions and understanding how those actions affect the environment. Video world models can only generate partial observations of the entire world at a time. Remembering past observations alone may not be sufficient in every scenario, as they may no longer reflect the world after a sequence of actions. We introduce a paradigm that decouples a persistent, updatable abstract state from the visual state. The abstract state keeps track of the entities' changes as the world evolves, and is used to condition the visual state, which represents the current view. We propose StateCraft, a novel architecture that pairs a language model with a video diffusion model to evolve the dynamic world. We evaluate our approach on generative interfaces, which pose a substantial challenge because a single action may change parts of the state that are not currently visible. To support training and evaluation, we develop a data collection pipeline that produces informative, densely annotated trajectories at scale without human or agent supervision, together with a benchmark measuring generation quality, memory, and controllability. On this benchmark, our approach improves surpasses the state-of-the-art baselines. We demonstrate that StateCraft maintains high self-consistency and strong generation quality over long trajectories. Because the abstract state is explicit, it can be edited directly, and generation follows edits to states never observed in the data. Finally, a proof-of-concept 3D experiment, in which actions change the illumination of an unobserved room, shows that the paradigm can also be applied to spatial environments. Project page: https://insait-institute.github.io/StateCraft/.
Oct 13, 2026

Kandinsky S2V 1.0: Audio-Driven Avatar Video Generation

We present Kandinsky S2V 1.0, an open family of speech-to-video models for controllable audio-driven human animation, comprising Kandinsky S2V 1.0 Lite (2B parameters) and Kandinsky S2V 1.0 Pro (21B parameters). From a single reference image, a speech waveform, and a text prompt, the models generate long-form 25 FPS video in which lip motion, expression, pose, and gesture follow the speech. The architecture draws text and audio conditioning from one Qwen2.5-Omni model and adds a reference-aware three-scale visual memory. Training starts from the Kandinsky 5.0 Video foundation models [1] and combines mixed reference-memory layouts with an area-adaptive face loss. The training corpus contains 1.17M single-speaker clips selected by active-speaker detection, with captions generated directly in English and Russian. A side-by-side human evaluation against the strongest open-source baseline and two proprietary commercial models shows that Kandinsky S2V 1.0 Pro performs comparably to or better than these models on average and is also preferred for lip synchronization. On common speech to video benchmarks, Kandinsky S2V 1.0 Lite and Kandinsky S2V 1.0 Pro rank among the top performing models according to mean rank over five reported metrics. Beyond the base models, we provide π-Flow distilled variants of both Kandinsky S2V 1.0 Lite and Kandinsky S2V 1.0 Pro. The distilled Kandinsky S2V 1.0 Pro model generates video in 16 sampling steps without classifier-free guidance, achieving a 5.4 times end-to-end speedup over the base model, while matching or exceeding it on nearly all reported benchmark metrics. To accelerate open research and deployment in speech-to-video generation, we release the code, model checkpoints, and diffusers integration under the MIT license.
Oct 13, 2026

Long-horizon Video Generation with Slot Memory Addressed By Region Tokens

Memory for long-horizon video generation must satisfy two competing requirements: it must remain bounded so generation can continue indefinitely, while still recovering what was observed at a place when the camera returns, regardless of how long ago it was seen. Reconstruction-based memories provide spatial recall by caching and reprojecting explicit geometry, but require growing storage and depend on estimated 3D structure. Recurrent memories keep a constant-size state, but retention is governed by temporal recurrence, causing old observations to fade with time. We introduce SMART, a slot memory addressed by region tokens, which combines bounded storage with spatially addressed recall. SMART stores content in a constant-size slot recurrence whose writes are addressed by the observing camera and whose reads are queried by the rendering camera, without constructing an explicit 3D representation or maintaining a growing cache. Consequently, recall is conditioned on inter-view geometry rather than on recency. Each read additionally provides a confidence estimate that conditions a chunk-causal diffusion transformer, allowing it to selectively trust retrieved content while relying on its generative prior elsewhere. On long trajectories in Memory Maze and Minecraft, SMART maintains high-fidelity recall far beyond its training horizon while using constant memory and per-step cost.
Oct 13, 2026

Focus, Align, and Diffuse: Time-Series-Aware Keyframe Diffusion for Cardiac Dynamic Synthesis from Sparse Observations

Dense cardiac dynamics are essential for assessing cardiac function, yet in practice clinicians often observe only extremely sparse measurements. However, recovering full dynamics from such sparsity is a highly ill-posed problem due to the lack of temporal cues. While auxiliary time-series signals like ECG provide physiological guidance, effectively exploiting them is challenging due to the cross-modal representation gap and phase misalignment. In this work, we present CardioFAD, a Focus--Align--Diffuse paradigm that reformulates sparse cardiac dynamic synthesis as a recoverability--alignability anchor discovery problem. Our core insight is that resolving extreme temporal ambiguity requires a compact set of anchor states that are both sufficient to recover the visual trajectory and reliable for ECG--visual alignment. We therefore learn motion-salient keyframes under two complementary principles: trajectory sufficiency, which preserves trajectory-critical intra-modal motion content, and anchor alignment, which promotes consistent inter-modal correspondence at these anchors. We implement the Focus and Align modules optimized by the two aspects. Diffuse then performs two-stage generation by first synthesizing keyframes and subsequently interpolating the remaining frames. Extensive experiments on cardiac MRI and echocardiography videos demonstrate superior visual quality, temporal coherence, and improved functional consistency under the extreme one-shot regime.
Oct 13, 2026

GAD: Geometry Attention Distillation for Consistent Video Generation

Video diffusion models generate visually compelling short-term sequences but often struggle to preserve scene geometry and object structure over time. In this paper, we first demonstrate that current video attention cannot identify reliable correspondences across frames, with responses dispersed across multiple regions even after recognizable scene structure has emerged. We argue that reliable cross-frame associations are essential for information exchange as scene structure forms in video generation. Thus, we introduce Geometry Attention Distillation (GAD), which transfers correspondences from a frozen geometry model's attention to video attention. Extracted from clean videos, these correspondences link the same static surface or moving object across frames. Since structure formation begins in deeper layers at higher noise and shallower layers at lower noise, we apply attention distillation at a student layer selected according to the noise timestep. After training, video attention focuses more consistently on corresponding locations. Experiments across all three video backbones demonstrate improved long-range scene consistency and dynamic temporal consistency, with geometric gains across generation lengths. On Wan2.2 5B, GAD improves cross-frame matching accuracy by at least 30% over the Wan baseline on both long-range benchmarks. Notably, GAD achieves these gains without additional inference modules, offering a promising approach to efficient and consistent video-based world simulation. Project page: https://fantasy-amap.github.io/GAD/.
Oct 13, 2026

LongForcing: Scaling Temporal Manifold Matching for Long-Horizon World Rollout

Autoregressive video diffusion models have emerged as a foundation for long-horizon world rollout. However, when extended far beyond their training windows, these models suffer from progressive drift: visual quality deteriorates, motion becomes inconsistent, and object identity gradually collapses as generated frames are recursively reused as context. We study a supervision-horizon gap: short-horizon distillation leaves later rollout states without direct supervision. We show that matching short-window marginals need not determine long-horizon behavior, while longer on-policy trajectories expose delayed failures to training. This motivates temporal manifold matching at scale for long-horizon world models. Building on this insight, we introduce LongForcing, a distillation approach that scales teacher supervision to long student self-rollouts. LongForcing preserves the short-horizon teacher-forcing and causal ODE-distillation stages, while replacing the realscore teacher in the final distribution-matching stage with an extended-horizon bidirectional teacher. The long-horizon teacher supervises extended student self-rollouts using the standard distribution-matching formulation, without modifying the student architecture or increasing inference cost. We evaluate LongForcing on text-to-video generation and interactive action-conditioned world models, achieving a VBench Quality score of 85.52 and improving WorldRoamBench 60s action score by 23.9 points in first-person ablations. Qualitative stress tests further sample interactive rollouts up to 24 hours, far beyond the training horizon. These results suggest that scaling supervision along the temporal axis is a practical step toward long-horizon world models. The project website and code are at https://github.com/Fantasy-AMAP/LongForcing.
Oct 13, 2026

VARI: Robust Text-to-Video Generation via Group-Structured Intent Alignment

Modern text-to-video (T2V) models generate visually compelling videos, yet meaning-preserving changes in prompt wording can substantially alter semantic realization and overall quality. Existing prompt rewriting modifies the inference input, whereas conventional post-training constructs supervision from one prompt formulation at a time; neither explicitly optimizes reliability across equivalent forms. We introduce VARI (Verified Alignment over Representations of Intent), a group-structured framework that treats automatically verified formulations of the same visual intent as a shared supervision unit. VARI builds matched form-seed rollout pools, cross-scores each video against all forms, and applies a wording-independent structured intent audit to identify valid realizations and hard semantic failures. Group-Consistent Distillation (GCD) aligns every verified form with interchangeable valid realizations, while Form-Contrastive Flow-DPO (FC-DPO) learns group-level preferences from audited positive-negative pairs, prioritizing same-seed, different-form contrasts. On HunyuanVideo-1.5, VARI attains the highest T2V-CompBench, VBench, and StoryEval scores among the evaluated methods. On prompt-form extensions of T2V-CompBench and VBench, it improves weakest-form alignment over the strongest prompt-level baselines by 2.45 and 3.55 points and reduces cross-form drop by 30.5% and 32.2%, respectively, while preserving visual quality. Human evaluation, natural-caption controls, prompt-enhancement experiments, diversity analyses, and transfer to LTX-2.3 further support the robustness and transferability of these gains, while VARI requires no additional inference-time module.
Oct 13, 2026

LinMorph: Learning to Morph Weights for Efficient Video DiT Linearization

Post-hoc linearization is an attractive way to accelerate pretrained Video Diffusion Transformers (Video DiTs), yet existing methods largely treat the quality loss induced by attention replacement as a recovery problem. We show that a more fundamental bottleneck lies upstream: initialization. Random initialization discards pretrained knowledge, whereas manual weight reuse transfers it through rigid and incomplete source--target correspondences. We introduce LinMorph, which recasts Video DiT linearization as a learnable softmax-to-linear weight transformation problem. Instead of copying predefined counterparts, LinMorph learns a cross-parameter morphism that lets each target Gated DeltaNet (GDN) parameter draw jointly from pretrained query, key, value, and output projections, including for parameters with no direct softmax counterpart. A single morphism is shared across layers and learned through local teacher alignment, while producing layer-specific initializations from their respective pretrained weights. The resulting weights are then materialized as independent GDN modules for progressive distillation and joint recovery.Since the morphism operates in weight space rather than token space, it can be learned on short, low-resolution clips and directly reused at larger video scales. Across pretrained Video DiTs, LinMorph enables markedly faster recovery and better final generation quality than random initialization and heuristic weight reuse. Results show that efficient linearization depends not only on how a converted model is recovered, but critically on how pretrained knowledge is transferred before recovery. Project page: https://lonelyling.github.io/LinMorph.
Oct 13, 2026

Hierarchical Curriculum Learning for Surgical Video Generation: Focusing on Bilateral Axillo-Breast Approach Robotic Thyroidectomy Datasets

High-fidelity surgical video generation is increasingly important for surgical education, simulation, and downstream AI training, as conventional simulation pipelines are limited by cost, realism, and scenario diversity. Adapting generic text-to-video diffusion models to surgical procedures remains difficult, however, due to complex anatomical structures, intricate tool-tissue interactions, the inherently hierarchical nature of surgical workflows, and the scarcity of curated training data outside a few well-studied procedures. We propose a surgical video generation framework that combines hierarchical procedural annotation with curriculum transfer learning. Surgical procedures are decomposed into three levels (phase, task, and action) and a pretrained text-to-video diffusion model is progressively adapted from coarse to fine-grained surgical context, supported by a dynamic sampling strategy that handles variable-resolution and variable-length surgical clips. We instantiate the framework on Bilateral Axillo-Breast Approach robotic thyroidectomy, a procedure for which no curated benchmark currently exists. Relative to the preceding task-level model, the final action-level model reduces FVD by 20.1% and FID by 15.7%, and a matched-budget ablation indicates that these gains arise from the curriculum design rather than from additional training. A user study with six clinicians rates visual realism highly while highlighting remaining gaps in anatomical accuracy and clinical reliability. The proposed framework offers a step toward AI-driven surgical simulation in procedures where curated data are scarce.
Oct 13, 2026

FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching

Recent video super-resolution (VSR) models achieve efficient one-step generation through diffusion distillation. However, high-resolution inference remains costly, and distilled models are difficult to align with human preferences. These challenges call for improvements in both model architecture and training strategy. We present FlashVSR+ to address both limitations. On the architecture side, we introduce hybrid attention based on the strong locality observed in pretrained VSR models, while retaining sparse global interaction for long-range consistency. It achieves up to 2×2\times DiT speedup and 45% lower peak memory, enabling spatially untiled 4K VSR on a single 24GB RTX 4090. On the training side, we propose Reward-Yield Distribution Matching Distillation (Ry-DMD) for preference-aligned one-step generation. Ry-DMD uses reward-weighted teacher rollouts to incorporate black-box rewards into distribution matching, without backpropagating through the reward model. We further show that Ry-DMD generalizes beyond VSR to T2I generation. Together, FlashVSR+ delivers stronger restoration quality and substantially improves high-resolution efficiency, with consistent gains in user preference. We release our code, models, and training data at \href{https://www.flashvsr.tech}{\textcolor{FlashBlue}{**www.flashvsr.tech**}}.