We investigate how to enhance the physical fidelity of video generation models by leveraging synthetic videos derived from computer graphics pipelines. These rendered videos respect real-world physics, such as maintaining 3D consistency, and serve as a valuable resource that can potentially improve video generation models. To harness this potential, we propose a solution that curates and integrates synthetic data while introducing a method to transfer its physical realism to the model, significantly reducing unwanted artifacts. Through experiments on three representative tasks emphasizing physical consistency, we demonstrate its efficacy in enhancing physical fidelity. While our model still lacks a deep understanding of physics, our work offers one of the first empirical demonstrations that synthetic video enhances physical fidelity in video synthesis. Website: https://kevinz8866.github.io/simulation/
Diffusion Transformers for video generation have gained significant research interest since the impressive performance of SORA. Efficient deployment of such generative-AI models on GPUs has been demonstrated with dynamic quantization. However, resource-constrained devices cannot support dynamic quantization, and need static quantization of the models for their efficient deployment on AI processors. In this paper, we propose a novel method for the post-training quantization of OpenSora\cite{opensora}, a Video Diffusion Transformer, without relying on dynamic quantization techniques. Our approach employs static quantization, achieving video quality comparable to FP16 and dynamically quantized ViDiT-Q methods, as measured by CLIP, and VQA metrics. In particular, we utilize per-step calibration data to adequately provide a post-training statically quantized model for each time step, incorporating channel-wise quantization for weights and tensor-wise quantization for activations. By further applying the smooth-quantization technique, we can obtain high-quality video outputs with the statically quantized models. Extensive experimental results demonstrate that static quantization can be a viable alternative to dynamic quantization for video diffusion transformers, offering a more efficient approach without sacrificing performance.
The probabilistic diffusion model (DM), generating content by inferencing through a recursive chain structure, has emerged as a powerful framework for visual generation. After pre-training on enormous unlabeled data, the model needs to be properly aligned to meet requirements for downstream applications. How to efficiently align the foundation DM is a crucial task. Contemporary methods are either based on Reinforcement Learning (RL) or truncated Backpropagation (BP). However, RL and truncated BP suffer from low sample efficiency and biased gradient estimation respectively, resulting in limited improvement or, even worse, complete training failure. To overcome the challenges, we propose the Recursive Likelihood Ratio (RLR) optimizer, a zeroth-order informed fine-tuning paradigm for DM. The zeroth-order gradient estimator enables the computation graph rearrangement within the recursive diffusive chain, making the RLR's gradient estimator an unbiased one with the lower variance than other methods. We provide theoretical guarantees for the performance of the RLR. Extensive experiments are conducted on image and video generation tasks to validate the superiority of the RLR. Furthermore, we propose a novel prompt technique that is natural for the RLR to achieve a synergistic effect.
Current video generative foundation models primarily focus on text-to-video tasks, providing limited control for fine-grained video content creation. Although adapter-based approaches (e.g., ControlNet) enable additional controls with minimal fine-tuning, they encounter challenges when integrating multiple conditions, including: branch conflicts between independently trained adapters, parameter redundancy leading to increased computational cost, and suboptimal performance compared to full fine-tuning. To address these challenges, we introduce FullDiT, a unified foundation model for video generation that seamlessly integrates multiple conditions via unified full-attention mechanisms. By fusing multi-task conditions into a unified sequence representation and leveraging the long-context learning ability of full self-attention to capture condition dynamics, FullDiT reduces parameter overhead, avoids conditions conflict, and shows scalability and emergent ability. We further introduce FullBench for multi-task video generation evaluation. Experiments demonstrate that FullDiT achieves state-of-the-art results, highlighting the efficacy of full-attention in complex multi-task video generation.
Sora has unveiled the immense potential of the Diffusion Transformer (DiT) architecture in single-scene video generation. However, the more challenging task of multi-scene video generation, which offers broader applications, remains relatively underexplored. To bridge this gap, we propose Mask$^2$DiT, a novel approach that establishes fine-grained, one-to-one alignment between video segments and their corresponding text annotations. Specifically, we introduce a symmetric binary mask at each attention layer within the DiT architecture, ensuring that each text annotation applies exclusively to its respective video segment while preserving temporal coherence across visual tokens. This attention mechanism enables precise segment-level textual-to-visual alignment, allowing the DiT architecture to effectively handle video generation tasks with a fixed number of scenes. To further equip the DiT architecture with the ability to generate additional scenes based on existing ones, we incorporate a segment-level conditional mask, which conditions each newly generated segment on the preceding video segments, thereby enabling auto-regressive scene extension. Both qualitative and quantitative experiments confirm that Mask$^2$DiT excels in maintaining visual consistency across segments while ensuring semantic alignment between each segment and its corresponding text description. Our project page is https://tianhao-qi.github.io/Mask2DiTProject.
Collecting real-world vehicle accident videos for autonomous driving research is challenging due to their rarity and complexity. While existing driving video generation methods may produce visually realistic videos, they often fail to deliver physically realistic simulations because they lack the capability to generate accurate post-collision trajectories. In this paper, we introduce AccidentSim, a novel framework that generates physically realistic vehicle collision videos by extracting and utilizing the physical clues and contextual information available in real-world vehicle accident reports. Specifically, AccidentSim leverages a reliable physical simulator to replicate post-collision vehicle trajectories from the physical and contextual information in the accident reports and to build a vehicle collision trajectory dataset. This dataset is then used to fine-tune a language model, enabling it to respond to user prompts and predict physically consistent post-collision trajectories across various driving scenarios based on user descriptions. Finally, we employ Neural Radiance Fields (NeRF) to render high-quality backgrounds, merging them with the foreground vehicles that exhibit physically realistic trajectories to generate vehicle collision videos. Experimental results demonstrate that the videos produced by AccidentSim excel in both visual and physical authenticity.
This report presents Wan, a comprehensive and open suite of video foundation models designed to push the boundaries of video generation. Built upon the mainstream diffusion transformer paradigm, Wan achieves significant advancements in generative capabilities through a series of innovations, including our novel VAE, scalable pre-training strategies, large-scale data curation, and automated evaluation metrics. These contributions collectively enhance the model's performance and versatility. Specifically, Wan is characterized by four key features: Leading Performance: The 14B model of Wan, trained on a vast dataset comprising billions of images and videos, demonstrates the scaling laws of video generation with respect to both data and model size. It consistently outperforms the existing open-source models as well as state-of-the-art commercial solutions across multiple internal and external benchmarks, demonstrating a clear and significant performance superiority. Comprehensiveness: Wan offers two capable models, i.e., 1.3B and 14B parameters, for efficiency and effectiveness respectively. It also covers multiple downstream applications, including image-to-video, instruction-guided video editing, and personal video generation, encompassing up to eight tasks. Consumer-Grade Efficiency: The 1.3B model demonstrates exceptional resource efficiency, requiring only 8.19 GB VRAM, making it compatible with a wide range of consumer-grade GPUs. Openness: We open-source the entire series of Wan, including source code and all models, with the goal of fostering the growth of the video generation community. This openness seeks to significantly expand the creative possibilities of video production in the industry and provide academia with high-quality video foundation models. All the code and models are available at https://github.com/Wan-Video/Wan2.1.
The progress on generative models has led to significant advances on text-to-video (T2V) generation, yet the motion controllability of generated videos remains limited. Existing motion transfer methods explored the motion representations of reference videos to guide generation. Nevertheless, these methods typically rely on sample-specific optimization strategy, resulting in high computational burdens. In this paper, we propose EfficientMT, a novel and efficient end-to-end framework for video motion transfer. By leveraging a small set of synthetic paired motion transfer samples, EfficientMT effectively adapts a pretrained T2V model into a general motion transfer framework that can accurately capture and reproduce diverse motion patterns. Specifically, we repurpose the backbone of the T2V model to extract temporal information from reference videos, and further propose a scaler module to distill motion-related information. Subsequently, we introduce a temporal integration mechanism that seamlessly incorporates reference motion features into the video generation process. After training on our self-collected synthetic paired samples, EfficientMT enables general video motion transfer without requiring test-time optimization. Extensive experiments demonstrate that our EfficientMT outperforms existing methods in efficiency while maintaining flexible motion controllability. Our code will be available https://github.com/PrototypeNx/EfficientMT.
Long-context video modeling is essential for enabling generative models to function as world simulators, as they must maintain temporal coherence over extended time spans. However, most existing models are trained on short clips, limiting their ability to capture long-range dependencies, even with test-time extrapolation. While training directly on long videos is a natural solution, the rapid growth of vision tokens makes it computationally prohibitive. To support exploring efficient long-context video modeling, we first establish a strong autoregressive baseline called Frame AutoRegressive (FAR). FAR models temporal dependencies between continuous frames, converges faster than video diffusion transformers, and outperforms token-level autoregressive models. Based on this baseline, we observe context redundancy in video autoregression. Nearby frames are critical for maintaining temporal consistency, whereas distant frames primarily serve as context memory. To eliminate this redundancy, we propose the long short-term context modeling using asymmetric patchify kernels, which apply large kernels to distant frames to reduce redundant tokens, and standard kernels to local frames to preserve fine-grained detail. This significantly reduces the training cost of long videos. Our method achieves state-of-the-art results on both short and long video generation, providing an effective baseline for long-context autoregressive video modeling.
Score-based or diffusion models generate high-quality tabular data, surpassing GAN-based and VAE-based models. However, these methods require substantial training time. In this paper, we introduce RecTable, which uses the rectified flow modeling, applied in such as text-to-image generation and text-to-video generation. RecTable features a simple architecture consisting of a few stacked gated linear unit blocks. Additionally, our training strategies are also simple, incorporating a mixed-type noise distribution and a logit-normal timestep distribution. Our experiments demonstrate that RecTable achieves competitive performance compared to the several state-of-the-art diffusion and score-based models while reducing the required training time. Our code is available at https://github.com/fmp453/rectable.
Video generation models have achieved remarkable progress in text-to-video tasks. These models are typically trained on text-video pairs with highly detailed and carefully crafted descriptions, while real-world user inputs during inference are often concise, vague, or poorly structured. This gap makes prompt optimization crucial for generating high-quality videos. Current methods often rely on large language models (LLMs) to refine prompts through in-context learning, but suffer from several limitations: they may distort user intent, omit critical details, or introduce safety risks. Moreover, they optimize prompts without considering the impact on the final video quality, which can lead to suboptimal results. To address these issues, we introduce VPO, a principled framework that optimizes prompts based on three core principles: harmlessness, accuracy, and helpfulness. The generated prompts faithfully preserve user intents and, more importantly, enhance the safety and quality of generated videos. To achieve this, VPO employs a two-stage optimization approach. First, we construct and refine a supervised fine-tuning (SFT) dataset based on principles of safety and alignment. Second, we introduce both text-level and video-level feedback to further optimize the SFT model with preference learning. Our extensive experiments demonstrate that VPO significantly improves safety, alignment, and video quality compared to baseline methods. Moreover, VPO shows strong generalization across video generation models. Furthermore, we demonstrate that VPO could outperform and be combined with RLHF methods on video generation models, underscoring the effectiveness of VPO in aligning video generation models. Our code and data are publicly available at https://github.com/thu-coai/VPO.
Classifier-Free Guidance (CFG) is a fundamental technique in training conditional diffusion models. The common practice for CFG-based training is to use a single network to learn both conditional and unconditional noise prediction, with a small dropout rate for conditioning. However, we observe that the joint learning of unconditional noise with limited bandwidth in training results in poor priors for the unconditional case. More importantly, these poor unconditional noise predictions become a serious reason for degrading the quality of conditional generation. Inspired by the fact that most CFG-based conditional models are trained by fine-tuning a base model with better unconditional generation, we first show that simply replacing the unconditional noise in CFG with that predicted by the base model can significantly improve conditional generation. Furthermore, we show that a diffusion model other than the one the fine-tuned model was trained on can be used for unconditional noise replacement. We experimentally verify our claim with a range of CFG-based conditional models for both image and video generation, including Zero-1-to-3, Versatile Diffusion, DiT, DynamiCrafter, and InstructPix2Pix.
We present Video Motion Graphs, a system designed to generate realistic human motion videos. Using a reference video and conditional signals such as music or motion tags, the system synthesizes new videos by first retrieving video clips with gestures matching the conditions and then generating interpolation frames to seamlessly connect clip boundaries. The core of our approach is HMInterp, a robust Video Frame Interpolation (VFI) model that enables seamless interpolation of discontinuous frames, even for complex motion scenarios like dancing. HMInterp i) employs a dual-branch interpolation approach, combining a Motion Diffusion Model for human skeleton motion interpolation with a diffusion-based video frame interpolation model for final frame generation. ii) adopts condition progressive training to effectively leverage identity strong and weak conditions, such as images and pose. These designs ensure both high video texture quality and accurate motion trajectory. Results show that our Video Motion Graphs outperforms existing generative- and retrieval-based methods for multi-modal conditioned human motion video generation. Project page can be found at https://h-liu1997.github.io/Video-Motion-Graphs/
Estimating motion in videos is an essential computer vision problem with many downstream applications, including controllable video generation and robotics. Current solutions are primarily trained using synthetic data or require tuning of situation-specific heuristics, which inherently limits these models' capabilities in real-world contexts. Despite recent developments in large-scale self-supervised learning from videos, leveraging such representations for motion estimation remains relatively underexplored. In this work, we develop Opt-CWM, a self-supervised technique for flow and occlusion estimation from a pre-trained next-frame prediction model. Opt-CWM works by learning to optimize counterfactual probes that extract motion information from a base video model, avoiding the need for fixed heuristics while training on unrestricted video inputs. We achieve state-of-the-art performance for motion estimation on real-world videos while requiring no labeled data.
Similar to conventional video generation, current deep learning-based weather prediction frameworks often lack explicit physical constraints, leading to unphysical outputs that limit their reliability for operational forecasting. Among various physical processes requiring proper representation, radiation plays a fundamental role as it drives Earth's weather and climate systems. However, accurate simulation of radiative transfer processes remains challenging for traditional numerical weather prediction (NWP) models due to their inherent complexity and high computational costs. Here, we propose FuXi-RTM, a hybrid physics-guided deep learning framework designed to enhance weather forecast accuracy while enforcing physical consistency. FuXi-RTM integrates a primary forecasting model (FuXi) with a fixed deep learning-based radiative transfer model (DLRTM) surrogate that efficiently replaces conventional radiation parameterization schemes. This represents the first deep learning-based weather forecasting framework to explicitly incorporate physical process modeling. Evaluated over a comprehensive 5-year dataset, FuXi-RTM outperforms its unconstrained counterpart in 88.51% of 3320 variable and lead time combinations, with improvements in radiative flux predictions. By incorporating additional physical processes, FuXi-RTM paves the way for next-generation weather forecasting systems that are both accurate and physically consistent.
Generative models offer a scalable and flexible paradigm for simulating complex environments, yet current approaches fall short in addressing the domain-specific requirements of autonomous driving - such as multi-agent interactions, fine-grained control, and multi-camera consistency. We introduce GAIA-2, Generative AI for Autonomy, a latent diffusion world model that unifies these capabilities within a single generative framework. GAIA-2 supports controllable video generation conditioned on a rich set of structured inputs: ego-vehicle dynamics, agent configurations, environmental factors, and road semantics. It generates high-resolution, spatiotemporally consistent multi-camera videos across geographically diverse driving environments (UK, US, Germany). The model integrates both structured conditioning and external latent embeddings (e.g., from a proprietary driving model) to facilitate flexible and semantically grounded scene synthesis. Through this integration, GAIA-2 enables scalable simulation of both common and rare driving scenarios, advancing the use of generative world models as a core tool in the development of autonomous systems. Videos are available at https://wayve.ai/thinking/gaia-2.
Human-object interaction (HOI) synthesis is important for various applications, ranging from virtual reality to robotics. However, acquiring 3D HOI data is challenging due to its complexity and high cost, limiting existing methods to the narrow diversity of object types and interaction patterns in training datasets. This paper proposes a novel zero-shot HOI synthesis framework without relying on end-to-end training on currently limited 3D HOI datasets. The core idea of our method lies in leveraging extensive HOI knowledge from pre-trained Multimodal Models. Given a text description, our system first obtains temporally consistent 2D HOI image sequences using image or video generation models, which are then uplifted to 3D HOI milestones of human and object poses. We employ pre-trained human pose estimation models to extract human poses and introduce a generalizable category-level 6-DoF estimation method to obtain the object poses from 2D HOI images. Our estimation method is adaptive to various object templates obtained from text-to-3D models or online retrieval. A physics-based tracking of the 3D HOI kinematic milestone is further applied to refine both body motions and object poses, yielding more physically plausible HOI generation results. The experimental results demonstrate that our method is capable of generating open-vocabulary HOIs with physical realism and semantic diversity.
Despite the recent progress of audio-driven video generation, existing methods mostly focus on driving facial movements, leading to non-coherent head and body dynamics. Moving forward, it is desirable yet challenging to generate holistic human videos with both accurate lip-sync and delicate co-speech gestures w.r.t. given audio. In this work, we propose AudCast, a generalized audio-driven human video generation framework adopting a cascade Diffusion-Transformers (DiTs) paradigm, which synthesizes holistic human videos based on a reference image and a given audio. 1) Firstly, an audio-conditioned Holistic Human DiT architecture is proposed to directly drive the movements of any human body with vivid gesture dynamics. 2) Then to enhance hand and face details that are well-knownly difficult to handle, a Regional Refinement DiT leverages regional 3D fitting as the bridge to reform the signals, producing the final results. Extensive experiments demonstrate that our framework generates high-fidelity audio-driven holistic human videos with temporal coherence and fine facial and hand details. Resources can be found at https://guanjz20.github.io/projects/AudCast.
The motion transfer task involves transferring motion from a source video to newly generated videos, requiring the model to decouple motion from appearance. Previous diffusion-based methods primarily rely on separate spatial and temporal attention mechanisms within 3D U-Net. In contrast, state-of-the-art video Diffusion Transformers (DiT) models use 3D full attention, which does not explicitly separate temporal and spatial information. Thus, the interaction between spatial and temporal dimensions makes decoupling motion and appearance more challenging for DiT models. In this paper, we propose DeT, a method that adapts DiT models to improve motion transfer ability. Our approach introduces a simple yet effective temporal kernel to smooth DiT features along the temporal dimension, facilitating the decoupling of foreground motion from background appearance. Meanwhile, the temporal kernel effectively captures temporal variations in DiT features, which are closely related to motion. Moreover, we introduce explicit supervision along dense trajectories in the latent feature space to further enhance motion consistency. Additionally, we present MTBench, a general and challenging benchmark for motion transfer. We also introduce a hybrid motion fidelity metric that considers both the global and local motion similarity. Therefore, our work provides a more comprehensive evaluation than previous works. Extensive experiments on MTBench demonstrate that DeT achieves the best trade-off between motion fidelity and edit fidelity.
Despite substantial progress in text-to-video generation, achieving precise and flexible control over fine-grained spatiotemporal attributes remains a significant unresolved challenge in video generation research. To address these limitations, we introduce VCtrl (also termed PP-VCtrl), a novel framework designed to enable fine-grained control over pre-trained video diffusion models in a unified manner. VCtrl integrates diverse user-specified control signals-such as Canny edges, segmentation masks, and human keypoints-into pretrained video diffusion models via a generalizable conditional module capable of uniformly encoding multiple types of auxiliary signals without modifying the underlying generator. Additionally, we design a unified control signal encoding pipeline and a sparse residual connection mechanism to efficiently incorporate control representations. Comprehensive experiments and human evaluations demonstrate that VCtrl effectively enhances controllability and generation quality. The source code and pre-trained models are publicly available and implemented using the PaddlePaddle framework at http://github.com/PaddlePaddle/PaddleMIX/tree/develop/ppdiffusers/examples/ppvctrl.