Video Diffusion Models

Step-by-step painting tutorials are used for learning artistic techniques, but existing video resources lack interactivity and personalization. While recent generative models have advanced artistic image synthesis, they still struggle to generate temporally and structurally consistent videos. Video models excel in consistent frame generation, yet fine-tuning an image-to-video model on forward painting sequences creates a long-range dependency on the first-frame reference, disrupting its pretrained attention pattern and hindering reconstruction. We address this issue by training on reversed sequences that better preserve the reference-attention behavior. Building on this, we present Loomis Painter, a framework for reconstructing coherent painting processes. We further introduce a media-transfer training strategy that decouples input appearance from the target process, enabling generation across different artistic media. Trained on a curated dataset of real-world painting processes, our method outperforms existing baselines across LPIPS, DINO, and CLIP metrics and is consistently preferred in expert and non-expert user studies. Code available at https://github.com/Markus-Pobitzer/wlp.

Video inverse problems are fundamental to streaming, telepresence, and AR/VR, where high perceptual quality must coexist with tight latency constraints. Diffusion-based priors currently deliver state-of-the-art reconstructions, but existing approaches either adapt image diffusion models with ad hoc temporal regularizers - leading to temporal artifacts - or rely on native video diffusion models whose iterative posterior sampling is far too slow for real-time use. We introduce InstantViR, an amortized inference framework for ultra-fast video reconstruction powered by a pre-trained video diffusion prior. We distill a powerful bidirectional video diffusion model (teacher) into a causal autoregressive student that maps a degraded video directly to its restored version in a single forward pass, inheriting the teacher's strong temporal modeling while completely removing iterative test-time optimization. The distillation is prior-driven: it only requires the teacher diffusion model and known degradation operators, and does not rely on externally paired clean/noisy video data. To further boost throughput, we replace the video-diffusion backbone VAE with a high-efficiency LeanVAE via an innovative teacher-space regularized distillation scheme, enabling low-latency latent-space processing. Across streaming random inpainting, Gaussian deblurring and super-resolution, InstantViR matches or surpasses the reconstruction quality of diffusion-based baselines while running at over 35 FPS on NVIDIA A100 GPUs, achieving up to 100 times speedups over iterative video diffusion solvers. These results show that diffusion-based video reconstruction is compatible with real-time, interactive, editable, streaming scenarios, turning high-quality video restoration into a practical component of modern vision systems.

Diffusion-based text-to-video models are increasingly capable, but mask-based editing over hundreds of frames remains challenging: naïve long-video generation suffers from memory blow-up, window seams, and temporal drift, while existing editors often require specialized modules or heavy fine-tuning. We present Overlapping High-Order Co-Denoising, a lightweight framework that turns a single pre-trained text-to-video model into a unified inpainting-outpainting editor. We train only LoRA adapters using mixed interior and border masks together with a dual-region loss that improves synthesis inside the mask while explicitly preserving known content. At inference, we denoise long latent sequences using overlapping windows, apply second-order Heun sampling within each window, and fuse overlaps with Hamming-weighted blending to reduce boundary artifacts and improve temporal coherence. On InpaintBench (30 real-world videos, 81--300 frames), our method outperforms Wan 2.1 variants and VACE in background faithfulness (SSIM/LPIPS), temporal consistency (tLPIPS), and text alignment (CLIP), and scales to long horizons, demonstrated up to 800 frames, with memory bounded by the chosen window size.

Although recent video generative models are getting more capable of following external camera controls, imposed by either text descriptions or camera trajectories, they still struggle to generalize to unconventional camera motions, which is crucial in creating truly original and artistic videos. The challenge lies in finding sufficient training videos with the intended uncommon camera motions. To this end, we propose VividCam, a training paradigm that enables diffusion models to learn complex camera motions from synthetic videos, releasing the reliance on collecting realistic training videos. VividCam incorporates multiple disentanglement strategies that isolate camera motion learning from synthetic appearance artifacts, ensuring more robust motion representation and mitigating domain shift. We show that our design synthesizes a wide range of precisely controlled camera motions using surprisingly simple synthetic data. Notably, this synthetic data often consists of basic geometries within a low-poly 3D scene and can be efficiently rendered by engines like Unity. Our video results can be found in https://wuqiuche.github.io/VividCamDemoPage/ .

This study focuses on a challenging yet promising task, Text-to-Sounding-Video (T2SV) generation, which aims to generate a video with synchronized audio from text conditions, meanwhile ensuring both modalities are aligned with text. Despite progress in joint audio-video training, two critical challenges still remain unaddressed: (1) a single, shared text caption where the text for video is equal to the text for audio often creates modal interference, confusing the pretrained backbones, and (2) the optimal mechanism for cross-modal feature interaction remains unclear. To address these challenges, we first propose the Hierarchical Visual-Grounded Captioning (HVGC) framework that generates pairs of disentangled captions, a video caption, and an audio caption, eliminating interference at the conditioning stage. Based on HVGC, we further introduce BridgeDiT, a novel dual-tower diffusion transformer, which employs a Dual CrossAttention (DCA) mechanism that acts as a robust ``bridge" to enable a symmetric, bidirectional exchange of information, achieving both semantic and temporal synchronization. Extensive experiments on three benchmark datasets, supported by human evaluations, demonstrate that our method achieves state-of-the-art results on most metrics. Comprehensive ablation studies further validate the effectiveness of our contributions, offering key insights for the future T2SV task. All the codes and checkpoints will be publicly released.

Recent video generation models have achieved remarkable progress and are now deployed in film, social media production, and advertising. Beyond their creative potential, such models also hold promise as world simulators for robotics and embodied decision making. Despite strong advances, current approaches still struggle to generate physically plausible object interactions and lack object-level control mechanisms. To address these limitations, we introduce KineMask, an approach for video generation that enables realistic rigid body control, interactions, and effects. Given a single image and a specified object velocity, our method generates videos with inferred motions and future object interactions. We propose a two-stage training strategy that gradually removes future motion supervision via object masks. Using this strategy we train video diffusion models (VDMs) on synthetic scenes of simple interactions and demonstrate significant improvements and generalization to rigid body and hand-object interactions in real scenes. Furthermore, KineMask integrates low-level motion control with high-level textual conditioning via predicted scene descriptions, leading to support for synthesis of complex dynamical phenomena. Our experiments show that KineMask generalizes to different VDMs and achieves strong improvements over recent models of comparable size. Ablation studies further highlight the complementary roles of low- and high-level conditioning in VDMs.

In light of recent progress in video editing, deep learning models focusing on both spatial and temporal dependencies have emerged as the primary method. However, these models suffer from the quadratic computational complexity of traditional attention mechanisms, making them difficult to adapt to long-duration and high-resolution videos. This limitation restricts their applicability in practical contexts such as real-time video processing. To tackle this challenge, we introduce a method to reduce both time and space complexity of these systems by proposing VRWKV-Editor, a novel video editing model that integrates a linear spatio-temporal aggregation module into video-based diffusion models. VRWKV-Editor leverages bidirectional weighted key-value recurrence mechanism of the RWKV transformer to capture global dependencies while preserving temporal coherence, achieving linear complexity without sacrificing quality. Extensive experiments demonstrate that the proposed method achieves up to 3.7x speedup and 60% lower memory usage compared to state-of-the-art diffusion-based video editing methods, while maintaining competitive performance in frame consistency and text alignment. Furthermore, a comparative analysis we conducted on videos with different sequence lengths confirms that the gap in editing speed between our approach and architectures with self-attention becomes more significant with long videos.

Hair simulation and rendering are challenging due to complex strand dynamics, diverse material properties, and intricate light-hair interactions. Recent video diffusion models can generate high-quality videos, but they lack fine-grained control over hair dynamics. We present ControlHair, a hybrid framework that integrates a physics simulator with conditional video diffusion to enable precise and controllable dynamic hair rendering. ControlHair adopts a three-stage pipeline: it first encodes physics conditions into per-frame geometry using a simulator, then extracts per-frame control signals, and finally feeds control signals into a video diffusion model to generate videos with desired hair dynamics. This cascaded design decouples physics reasoning from video generation, supports diverse physics, and makes training the video diffusion model easy. Trained on a curated 10K video dataset, ControlHair outperforms text- and pose-conditioned baselines, delivering precisely controlled hair dynamics. We also demonstrate three use cases of ControlHair, including dynamic hairstyle try-on, bullet-time effects, and cinemagraphic. Project page: https://linwk20.github.io/controlhair-web.

Video diffusion models have achieved remarkable progress in recent years, yet generating high-resolution videos remain a fundamental challenge. Most video generators are trained at limited spatial resolutions due to the scarcity of high-resolution 4K video data and the prohibitive computational cost of large-scale training on such data. Most video diffusion models are trained on 720p videos and are therefore effectively limited to generating videos at similar resolutions during inference. To address this gap, we propose CineScale. CineScale, to the best of our knowledge, is the first tuning-free inference framework enabling pretrained video diffusion models to generate high-quality videos at resolutions far beyond those seen during training. Our key observation is that generation quality degrades at higher resolutions because positional encodings shift beyond their training distribution, producing blurred details and structurally incoherent videos. To address this gap, we introduce Adaptively Rectified RoPE. Our extensive experiments show that CineScale enables pretrained diffusion models, despite never being trained on high-resolution data, to generate high-fidelity 4K video without any fine-tuning, improving local detail and sharpness while preserving temporal coherence. This demonstrates that high-resolution generation capabilities can be unlocked purely at inference time.

We present Vivid-VR, a DiT-based generative video restoration method built upon an advanced T2V foundation model, where ControlNet is leveraged to control the generation process, ensuring content consistency. However, conventional fine-tuning of such controllable pipelines frequently suffers from distribution drift due to limitations in imperfect multimodal alignment, resulting in compromised texture realism and temporal coherence. To tackle this challenge, we propose a concept distillation training strategy that utilizes the pretrained T2V model to synthesize training samples with embedded textual concepts, thereby distilling its conceptual understanding to preserve texture and temporal quality. To enhance generation controllability, we redesign the control architecture with two key components: 1) a control feature projector that filters degradation artifacts from input video latents to minimize their propagation through the generation pipeline, and 2) a new ControlNet connector employing a dual-branch design. This connector synergistically combines MLP-based feature mapping with cross-attention mechanism for dynamic control feature retrieval, enabling both content preservation and adaptive control signal modulation. Extensive experiments show that Vivid-VR performs favorably against existing approaches on both synthetic and real-world benchmarks, as well as AIGC videos, achieving impressive texture realism, visual vividness, and temporal consistency. The codes and checkpoints are publicly available at https://github.com/csbhr/Vivid-VR.

Recent progress in transformer-based architectures has demonstrated remarkable success in video generation tasks. However, the quadratic complexity of full attention mechanisms remains a critical bottleneck, particularly for high-resolution and long-duration video sequences. In this paper, we propose NABLA, a novel Neighborhood Adaptive Block-Level Attention mechanism that dynamically adapts to sparsity patterns in video diffusion transformers (DiTs). By leveraging block-wise attention with adaptive sparsity-driven threshold, NABLA reduces computational overhead while preserving generative quality. Our method does not require custom low-level operator design and can be seamlessly integrated with PyTorch's Flex Attention operator. Experiments demonstrate that NABLA achieves up to 2.7x faster training and inference compared to baseline almost without compromising quantitative metrics (CLIP score, VBench score, human evaluation score) and visual quality drop. The code and model weights are available here: https://github.com/gen-ai-team/Wan2.1-NABLA

Training-free motion customization imposes motion patterns from reference videos onto video generators through test-time computation. Most existing methods target full diffusion models, requiring many denoising steps and high computational cost. With the rise of efficient distilled models, a natural question arises: can test-time motion customization be applied directly to distilled generators with their accelerated sampling and efficiency gains? However, our analysis reveals that existing training-free techniques fail on distilled models. Distillation fundamentally alters the denoising dynamics that prior test-time guidance relies on, and the large denoising steps of distilled generators discard the dense intermediate states that score guidance requires, rendering existing motion control strategies incompatible with fast generation. To address this limitation, we propose MotionEcho, a novel training-free test-time distillation framework that enables motion customization for distilled video generators. The key idea is to correct the student model's sampling trajectory with restricted usage of a high-quality diffusion teacher at inference time. Teacher supervises the student's denoising by re-noising the student's endpoint onto its dense trajectory to form a motion-aligned clean endpoint, then interpolating it with the student's, while an adaptive scheduling mechanism determines when and how much teacher guidance is needed. As a result, MotionEcho restores generative trajectories for distilled video generators via lightweight, adaptive test-time teacher guidance, enabling accurate motion control without compromising generation efficiency. Extensive experiments on multiple distilled video generation models demonstrate that our method significantly improves motion fidelity and visual quality while retaining the efficiency advantages of distilled generation.

Recent advances in 3D content generation have amplified demand for dynamic models that are both visually realistic and physically consistent. However, state-of-the-art video diffusion models frequently produce implausible results such as momentum violations and object interpenetrations. Existing physics-aware approaches often rely on task-specific fine-tuning or supervised data, which limits their scalability and applicability. To address the challenge, we present PhyMAGIC, a training-free framework that generates physically consistent motion from a single image. PhyMAGIC integrates a pre-trained image-to-video diffusion model, confidence-guided reasoning via LLMs, and a differentiable physics simulator to produce 3D assets ready for downstream physical simulation without fine-tuning or manual supervision. By iteratively refining motion prompts using LLM-derived confidence scores and leveraging simulation feedback, PhyMAGIC steers generation toward physically consistent dynamics. Comprehensive experiments demonstrate that PhyMAGIC outperforms state-of-the-art video generators and physics-aware baselines, enhancing physical property inference and motion-text alignment while maintaining visual fidelity.

Subject-driven video generation (SDV-Gen) aims to produce videos of a specific subject by adapting a pretrained video model, enabling personalized and application-driven content creation. To achieve this goal, per-subject tuning methods require approximately 200 A100 GPU hours to generate a customized video, whereas zero-shot methods avoid per-subject tuning but typically rely on millions of subject-video pairs for the supervision, incurring massive network fine-tuning costs (10K-200K A100 GPU hours). We propose a data- and compute-efficient zero-shot SDV-Gen framework that avoids test-time per-subject tuning and the use of large-scale subject-video pairs. Our key idea decomposes SDV-Gen into (i) identity injection learned from subject-image pairs and (ii) motion-awareness preservation maintained by a small set of arbitrary videos. We optimize the two tasks with stochastic switching, using random reference-frame sampling and image-token dropout to prevent trivial first-frame copying. Our gradient analysis shows that the two objectives rapidly evolve toward nearly orthogonal update subspaces, explaining the stable optimization. Using CogVideoX-5B, we adapt a single model with 200K subject-image pairs and 4,000 arbitrary videos in 288 A100 GPU hours. This yields about 1% of compute compared to prior zero-shot baselines (i.e., 0.4% of VACE and 2.8% of Phantom) while using no subject-video pairs, yet remaining competitive in subject fidelity and motion quality. We show that the same recipe transfers to Wan 2.1-1.3B and Wan 2.2-5B.

Rapid advancements in video diffusion models have enabled the creation of realistic videos, raising concerns about unauthorized use and driving the demand for techniques to protect model ownership. Existing watermarking methods suffer from two key limitations: they overlook temporal consistency due to conventional watermark decoders and degrade the visual quality of the generated videos. To address these issues, we introduce a robust watermarking method for latent video diffusion models named Latent Video Diffusion Watermarking (LVMark). We propose a novel watermark decoder tailored for generated videos by learning the consistency between adjacent frames. It ensures accurate message decoding, even under malicious attacks, by combining the low-frequency components of the three-dimensional wavelet domain with the color features of the video. Additionally, we train a latent decoder to maintain the visual fidelity of the generated video. Watermarks are embedded into layers with minimal impact on visual appearance using an importance-based weight modulation strategy. We optimize both the watermark decoder and the latent decoder of diffusion model, effectively balancing the trade-off between visual quality and bit accuracy. Our experiments show that our method embeds invisible watermarks into video diffusion models, ensuring robust decoding accuracy with 512-bit capacity, even under distortions.

The rapid development of diffusion models (DMs) has significantly advanced image and video applications, making "what you want is what you see" a reality. Among these, video editing has gained substantial attention and seen a swift rise in research activity, necessitating a comprehensive and systematic review of the existing literature. This paper reviews diffusion model-based video editing techniques, including theoretical foundations and practical applications. We begin by overviewing the mathematical formulation and image domain's key methods. Subsequently, we categorize video editing approaches by the inherent connections of their core technologies, depicting evolutionary trajectory. This paper also dives into novel applications, including point-based editing and pose-guided human video editing. Additionally, we present a comprehensive comparison using our newly introduced V2VBench. Building on the progress achieved to date, the paper concludes with ongoing challenges and potential directions for future research.

Video diffusion models can enable embodied agents to anticipate plausible futures from the recent past, but they are typically trained offline on curated datasets--a mismatch with the agents' learning setup at deployment: online, from a single video stream that sequentially outputs one frame at a time. We bridge this training gap and demonstrate that training autoregressive video diffusion models from such a stream, resembling the experience of embodied agents, is not only possible but can also perform comparably to standard offline training given the same number of gradient steps. We find that this robustness to video stream autocorrelation and nonstationarity can be achieved using experience replay methods that retain a subset of the video stream. To support training and evaluation in this setting, we introduce five new datasets for streaming lifelong generative video modeling: Lifelong Bouncing Balls (O), Lifelong Bouncing Balls (C), Lifelong 3D Maze, Lifelong Drive, and Lifelong PLAICraft, each consisting of one million consecutive frames from environments of increasing complexity. Together, our datasets and experiments lay the groundwork for video generative models and world models that continuously learn from single-sensor video streams rather than fixed datasets.

Given the remarkable achievements in image generation through diffusion models, the research community has shown increasing interest in extending these models to video generation. Recent diffusion models for video generation have predominantly utilized attention layers to extract temporal features. However, attention layers are limited by their computational costs, which increase quadratically with the sequence length. This limitation presents significant challenges when generating longer video sequences using diffusion models. To overcome this challenge, we propose leveraging state-space models (SSMs) as temporal feature extractors. SSMs (e.g., Mamba) have recently gained attention as promising alternatives due to their linear-time memory consumption relative to sequence length. In line with previous research suggesting that using bidirectional SSMs is effective for understanding spatial features in image generation, we found that bidirectionality is also beneficial for capturing temporal features in video data, rather than relying on traditional unidirectional SSMs. We conducted comprehensive evaluations on multiple long-term video datasets, such as MineRL Navigate, across various model sizes. For sequences up to 256 frames, SSM-based models require less memory to achieve the same FVD as attention-based models. Moreover, SSM-based models often deliver better performance with comparable GPU memory usage. Our codes are available at https://github.com/shim0114/SSM-Meets-Video-Diffusion-Models.