We propose Heterogeneous Masked Autoregression (HMA) for modeling action-video dynamics to generate high-quality data and evaluation in scaling robot learning. Building interactive video world models and policies for robotics is difficult due to the challenge of handling diverse settings while maintaining computational efficiency to run in real time. HMA uses heterogeneous pre-training from observations and action sequences across different robotic embodiments, domains, and tasks. HMA uses masked autoregression to generate quantized or soft tokens for video predictions. \ourshort achieves better visual fidelity and controllability than the previous robotic video generation models with 15 times faster speed in the real world. After post-training, this model can be used as a video simulator from low-level action inputs for evaluating policies and generating synthetic data. See this link https://liruiw.github.io/hma for more information.
With the rise of diffusion models, audio-video generation has been revolutionized. However, most existing methods rely on separate modules for each modality, with limited exploration of unified generative architectures. In addition, many are confined to a single task and small-scale datasets. To address these limitations, we first propose UniForm, a unified multi-task diffusion transformer that jointly generates audio and visual modalities in a shared latent space. A single diffusion process models both audio and video, capturing the inherent correlations between sound and vision. Second, we introduce task-specific noise schemes and task tokens, enabling a single model to support multiple tasks, including text-to-audio-video, audio-to-video, and video-to-audio generation. Furthermore, by leveraging large language models and a large-scale text-audio-video combined dataset, UniForm achieves greater generative diversity than prior approaches. Extensive experiments show that UniForm achieves the state-of-the-art performance across audio-video generation tasks, producing content that is both well-aligned and close to real-world data distributions. Our demos are available at https://uniform-t2av.github.io/.
Text-driven video generation has advanced significantly due to developments in diffusion models. Beyond the training and sampling phases, recent studies have investigated noise priors of diffusion models, as improved noise priors yield better generation results. One recent approach employs the Fourier transform to manipulate noise, marking the initial exploration of frequency operations in this context. However, it often generates videos that lack motion dynamics and imaging details. In this work, we provide a comprehensive theoretical analysis of the variance decay issue present in existing methods, contributing to the loss of details and motion dynamics. Recognizing the critical impact of noise distribution on generation quality, we introduce FreqPrior, a novel noise initialization strategy that refines noise in the frequency domain. Our method features a novel filtering technique designed to address different frequency signals while maintaining the noise prior distribution that closely approximates a standard Gaussian distribution. Additionally, we propose a partial sampling process by perturbing the latent at an intermediate timestep during finding the noise prior, significantly reducing inference time without compromising quality. Extensive experiments on VBench demonstrate that our method achieves the highest scores in both quality and semantic assessments, resulting in the best overall total score. These results highlight the superiority of our proposed noise prior.
We propose MotionAgent, enabling fine-grained motion control for text-guided image-to-video generation. The key technique is the motion field agent that converts motion information in text prompts into explicit motion fields, providing flexible and precise motion guidance. Specifically, the agent extracts the object movement and camera motion described in the text and converts them into object trajectories and camera extrinsics, respectively. An analytical optical flow composition module integrates these motion representations in 3D space and projects them into a unified optical flow. An optical flow adapter takes the flow to control the base image-to-video diffusion model for generating fine-grained controlled videos. The significant improvement in the Video-Text Camera Motion metrics on VBench indicates that our method achieves precise control over camera motion. We construct a subset of VBench to evaluate the alignment of motion information in the text and the generated video, outperforming other advanced models on motion generation accuracy.
Music is a deeply personal experience and our aim is to enhance this with a fully-automated pipeline for personalized music video generation. Our work allows listeners to not just be consumers but co-creators in the music video generation process by creating personalized, consistent and context-driven visuals based on lyrics, rhythm and emotion in the music. The pipeline combines multimodal translation and generation techniques and utilizes low-rank adaptation on listeners' images to create immersive music videos that reflect both the music and the individual. To ensure the ethical use of users' identity, we also introduce CHARCHA (patent pending), a facial identity verification protocol that protects people against unauthorized use of their face while at the same time collecting authorized images from users for personalizing their videos. This paper thus provides a secure and innovative framework for creating deeply personalized music videos.
Despite tremendous recent progress, generative video models still struggle to capture real-world motion, dynamics, and physics. We show that this limitation arises from the conventional pixel reconstruction objective, which biases models toward appearance fidelity at the expense of motion coherence. To address this, we introduce VideoJAM, a novel framework that instills an effective motion prior to video generators, by encouraging the model to learn a joint appearance-motion representation. VideoJAM is composed of two complementary units. During training, we extend the objective to predict both the generated pixels and their corresponding motion from a single learned representation. During inference, we introduce Inner-Guidance, a mechanism that steers the generation toward coherent motion by leveraging the model's own evolving motion prediction as a dynamic guidance signal. Notably, our framework can be applied to any video model with minimal adaptations, requiring no modifications to the training data or scaling of the model. VideoJAM achieves state-of-the-art performance in motion coherence, surpassing highly competitive proprietary models while also enhancing the perceived visual quality of the generations. These findings emphasize that appearance and motion can be complementary and, when effectively integrated, enhance both the visual quality and the coherence of video generation. Project website: https://hila-chefer.github.io/videojam-paper.github.io/
This paper considers an efficient video modeling process called Video Latent Flow Matching (VLFM). Unlike prior works, which randomly sampled latent patches for video generation, our method relies on current strong pre-trained image generation models, modeling a certain caption-guided flow of latent patches that can be decoded to time-dependent video frames. We first speculate multiple images of a video are differentiable with respect to time in some latent space. Based on this conjecture, we introduce the HiPPO framework to approximate the optimal projection for polynomials to generate the probability path. Our approach gains the theoretical benefits of the bounded universal approximation error and timescale robustness. Moreover, VLFM processes the interpolation and extrapolation abilities for video generation with arbitrary frame rates. We conduct experiments on several text-to-video datasets to showcase the effectiveness of our method.
Video foundation models have achieved significant advancement with the help of network upgrade as well as model scale-up. However, they are still hard to meet requirements of applications due to unsatisfied generation quality. To solve this problem, we propose to align video foundation models with human preferences from the perspective of post-training in this paper. Consequently, we introduce an Iterative Preference Optimization strategy to enhance generated video quality by incorporating human feedback. Specifically, IPO exploits a critic model to justify video generations for pairwise ranking as in Direct Preference Optimization or point-wise scoring as in Kahneman-Tversky Optimization. Given this, IPO optimizes video foundation models with guidance of signals from preference feedback, which helps improve generated video quality in subject consistency, motion smoothness and aesthetic quality, etc. In addition, IPO incorporates the critic model with the multi-modality large language model, which enables it to automatically assign preference labels without need of retraining or relabeling. In this way, IPO can efficiently perform multi-round preference optimization in an iterative manner, without the need of tediously manual labeling. Comprehensive experiments demonstrate that the proposed IPO can effectively improve the video generation quality of a pretrained model and help a model with only 2B parameters surpass the one with 5B parameters. Besides, IPO achieves new state-of-the-art performance on VBench benchmark. We will release our source codes, models as well as dataset to advance future research and applications.
Diffusion Transformers (DiTs) dominate video generation but their high computational cost severely limits real-world applicability, usually requiring tens of minutes to generate a few seconds of video even on high-performance GPUs. This inefficiency primarily arises from the quadratic computational complexity of 3D Full Attention with respect to the context length. In this paper, we propose a training-free framework termed Sparse VideoGen (SVG) that leverages the inherent sparsity in 3D Full Attention to boost inference efficiency. We reveal that the attention heads can be dynamically classified into two groups depending on distinct sparse patterns: (1) Spatial Head, where only spatially-related tokens within each frame dominate the attention output, and (2) Temporal Head, where only temporally-related tokens across different frames dominate. Based on this insight, SVG proposes an online profiling strategy to capture the dynamic sparse patterns and predicts the type of attention head. Combined with a novel hardware-efficient tensor layout transformation and customized kernel implementations, SVG achieves up to 2.28x and 2.33x end-to-end speedup on CogVideoX-v1.5 and HunyuanVideo, respectively, while preserving generation quality.
Recent advancements in video generation have significantly improved the ability to synthesize videos from text instructions. However, existing models still struggle with key challenges such as instruction misalignment, content hallucination, safety concerns, and bias. Addressing these limitations, we introduce MJ-BENCH-VIDEO, a large-scale video preference benchmark designed to evaluate video generation across five critical aspects: Alignment, Safety, Fineness, Coherence & Consistency, and Bias & Fairness. This benchmark incorporates 28 fine-grained criteria to provide a comprehensive evaluation of video preference. Building upon this dataset, we propose MJ-VIDEO, a Mixture-of-Experts (MoE)-based video reward model designed to deliver fine-grained reward. MJ-VIDEO can dynamically select relevant experts to accurately judge the preference based on the input text-video pair. This architecture enables more precise and adaptable preference judgments. Through extensive benchmarking on MJ-BENCH-VIDEO, we analyze the limitations of existing video reward models and demonstrate the superior performance of MJ-VIDEO in video preference assessment, achieving 17.58% and 15.87% improvements in overall and fine-grained preference judgments, respectively. Additionally, introducing MJ-VIDEO for preference tuning in video generation enhances the alignment performance.
With the rapid development of AIGC technology, significant progress has been made in diffusion model-based technologies for text-to-image (T2I) and text-to-video (T2V). In recent years, a few studies have introduced the strategy of Direct Preference Optimization (DPO) into T2I tasks, significantly enhancing human preferences in generated images. However, existing T2V generation methods lack a well-formed pipeline with exact loss function to guide the alignment of generated videos with human preferences using DPO strategies. Additionally, challenges such as the scarcity of paired video preference data hinder effective model training. At the same time, the lack of training datasets poses a risk of insufficient flexibility and poor video generation quality in the generated videos. Based on those problems, our work proposes three targeted solutions in sequence. 1) Our work is the first to introduce the DPO strategy into the T2V tasks. By deriving a carefully structured loss function, we utilize human feedback to align video generation with human preferences. We refer to this new method as HuViDPO. 2) Our work constructs small-scale human preference datasets for each action category and fine-tune this model, improving the aesthetic quality of the generated videos while reducing training costs. 3) We adopt a First-Frame-Conditioned strategy, leveraging the rich in formation from the first frame to guide the generation of subsequent frames, enhancing flexibility in video generation. At the same time, we employ a SparseCausal Attention mechanism to enhance the quality of the generated videos.More details and examples can be accessed on our website: https://tankowa.github.io/HuViDPO. github.io/.
In the era of generative AI, integrating video generation models into robotics opens new possibilities for the general-purpose robot agent. This paper introduces imitation learning with latent video planning (VILP). We propose a latent video diffusion model to generate predictive robot videos that adhere to temporal consistency to a good degree. Our method is able to generate highly time-aligned videos from multiple views, which is crucial for robot policy learning. Our video generation model is highly time-efficient. For example, it can generate videos from two distinct perspectives, each consisting of six frames with a resolution of 96x160 pixels, at a rate of 5 Hz. In the experiments, we demonstrate that VILP outperforms the existing video generation robot policy across several metrics: training costs, inference speed, temporal consistency of generated videos, and the performance of the policy. We also compared our method with other imitation learning methods. Our findings indicate that VILP can rely less on extensive high-quality task-specific robot action data while still maintaining robust performance. In addition, VILP possesses robust capabilities in representing multi-modal action distributions. Our paper provides a practical example of how to effectively integrate video generation models into robot policies, potentially offering insights for related fields and directions. For more details, please refer to our open-source repository https://github.com/ZhengtongXu/VILP.
We propose ``Shape from Semantics'', which is able to create 3D models whose geometry and appearance match given semantics when observed from different views. Traditional ``Shape from X'' tasks usually use visual input (e.g., RGB images or depth maps) to reconstruct geometry, imposing strict constraints that limit creative explorations. As applications, works like Shadow Art and Wire Art often struggle to grasp the embedded semantics of their design through direct observation and rely heavily on specific setups for proper display. To address these limitations, our framework uses semantics as input, greatly expanding the design space to create objects that integrate multiple semantic elements and are easily discernible by observers. Considering that this task requires a rich imagination, we adopt various generative models and structure-to-detail pipelines. Specifically, we adopt multi-semantics Score Distillation Sampling (SDS) to distill 3D geometry and appearance from 2D diffusion models, ensuring that the initial shape is consistent with the semantic input. We then use image restoration and video generation models to add more details as supervision. Finally, we introduce neural signed distance field (SDF) representation to achieve detailed shape reconstruction. Our framework generates meshes with complex details, well-structured geometry, coherent textures, and smooth transitions, resulting in visually appealing and eye-catching designs. Project page: https://shapefromsemantics.github.io
The remarkable progress in text-to-video diffusion models enables photorealistic generations, although the contents of the generated video often include unnatural movement or deformation, reverse playback, and motionless scenes. Recently, an alignment problem has attracted huge attention, where we steer the output of diffusion models based on some quantity on the goodness of the content. Because there is a large room for improvement of perceptual quality along the frame direction, we should address which metrics we should optimize and how we can optimize them in the video generation. In this paper, we propose diffusion latent beam search with lookahead estimator, which can select better diffusion latent to maximize a given alignment reward, at inference time. We then point out that the improvement of perceptual video quality considering the alignment to prompts requires reward calibration by weighting existing metrics. When evaluating outputs by using vision language models as a proxy of humans, many previous metrics to quantify the naturalness of video do not always correlate with evaluation and also depend on the degree of dynamic descriptions in evaluation prompts. We demonstrate that our method improves the perceptual quality based on the calibrated reward, without model parameter update, and outputs the best generation compared to greedy search and best-of-N sampling. We provide practical guidelines on which axes, among search budget, lookahead steps for reward estimate, and denoising steps, in the reverse diffusion process, we should allocate the inference-time computation.
Image animation has become a promising area in multimodal research, with a focus on generating videos from reference images. While prior work has largely emphasized generic video generation guided by text, music-driven dance video generation remains underexplored. In this paper, we introduce MuseDance, an innovative end-to-end model that animates reference images using both music and text inputs. This dual input enables MuseDance to generate personalized videos that follow text descriptions and synchronize character movements with the music. Unlike existing approaches, MuseDance eliminates the need for complex motion guidance inputs, such as pose or depth sequences, making flexible and creative video generation accessible to users of all expertise levels. To advance research in this field, we present a new multimodal dataset comprising 2,904 dance videos with corresponding background music and text descriptions. Our approach leverages diffusion-based methods to achieve robust generalization, precise control, and temporal consistency, setting a new baseline for the music-driven image animation task.
Recently, with the tremendous success of diffusion models in the field of text-to-image (T2I) generation, increasing attention has been directed toward their potential in text-to-video (T2V) applications. However, the computational demands of diffusion models pose significant challenges, particularly in generating high-resolution videos with high frame rates. In this paper, we propose CascadeV, a cascaded latent diffusion model (LDM), that is capable of producing state-of-the-art 2K resolution videos. Experiments demonstrate that our cascaded model achieves a higher compression ratio, substantially reducing the computational challenges associated with high-quality video generation. We also implement a spatiotemporal alternating grid 3D attention mechanism, which effectively integrates spatial and temporal information, ensuring superior consistency across the generated video frames. Furthermore, our model can be cascaded with existing T2V models, theoretically enabling a 4$\times$ increase in resolution or frames per second without any fine-tuning. Our code is available at https://github.com/bytedance/CascadeV.
With the rapid advancement of artificial intelligence technologies such as ChatGPT, AI agents and video generation,contemporary mobile systems have begun integrating these AI capabilities on local devices to enhance privacy and reduce response latency. To meet the computational demands of AI tasks, current mobile SoCs are equipped with diverse AI accelerators, including GPUs and Neural Processing Units (NPUs). However, there has not been a comprehensive characterization of these heterogeneous processors, and existing designs typically only leverage a single AI accelerator for LLM inference, leading to suboptimal use of computational resources and memory bandwidth. In this paper, we first summarize key performance characteristics of mobile SoC, including heterogeneous processors, unified memory, synchronization, etc. Drawing on these observations, we propose different tensor partition strategies to fulfill the distinct requirements of the prefill and decoding phases. We further design a fast synchronization mechanism that leverages the unified memory address provided by mobile SoCs. By employing these techniques, we present HeteroLLM, the fastest LLM inference engine in mobile devices which supports both layer-level and tensor-level heterogeneous execution. Evaluation results show that HeteroLLM achieves 9.99 and 4.36 performance improvement over other mobile-side LLM inference engines: MLC and MNN.
Artificial Intelligence Generated Content (AIGC) has advanced significantly, particularly with the development of video generation models such as text-to-video (T2V) models and image-to-video (I2V) models. However, like other AIGC types, video generation requires robust content control. A common approach is to embed watermarks, but most research has focused on images, with limited attention given to videos. Traditional methods, which embed watermarks frame-by-frame in a post-processing manner, often degrade video quality. In this paper, we propose VideoShield, a novel watermarking framework specifically designed for popular diffusion-based video generation models. Unlike post-processing methods, VideoShield embeds watermarks directly during video generation, eliminating the need for additional training. To ensure video integrity, we introduce a tamper localization feature that can detect changes both temporally (across frames) and spatially (within individual frames). Our method maps watermark bits to template bits, which are then used to generate watermarked noise during the denoising process. Using DDIM Inversion, we can reverse the video to its original watermarked noise, enabling straightforward watermark extraction. Additionally, template bits allow precise detection for potential temporal and spatial modification. Extensive experiments across various video models (both T2V and I2V models) demonstrate that our method effectively extracts watermarks and detects tamper without compromising video quality. Furthermore, we show that this approach is applicable to image generation models, enabling tamper detection in generated images as well. Codes and models are available at \href{https://github.com/hurunyi/VideoShield}{https://github.com/hurunyi/VideoShield}.
Recent advancements in video generation have significantly impacted various downstream applications, particularly in identity-preserving video generation (IPT2V). However, existing methods struggle with "copy-paste" artifacts and low similarity issues, primarily due to their reliance on low-level facial image information. This dependence can result in rigid facial appearances and artifacts reflecting irrelevant details. To address these challenges, we propose EchoVideo, which employs two key strategies: (1) an Identity Image-Text Fusion Module (IITF) that integrates high-level semantic features from text, capturing clean facial identity representations while discarding occlusions, poses, and lighting variations to avoid the introduction of artifacts; (2) a two-stage training strategy, incorporating a stochastic method in the second phase to randomly utilize shallow facial information. The objective is to balance the enhancements in fidelity provided by shallow features while mitigating excessive reliance on them. This strategy encourages the model to utilize high-level features during training, ultimately fostering a more robust representation of facial identities. EchoVideo effectively preserves facial identities and maintains full-body integrity. Extensive experiments demonstrate that it achieves excellent results in generating high-quality, controllability and fidelity videos.
Video generation has achieved significant advances through rectified flow techniques, but issues like unsmooth motion and misalignment between videos and prompts persist. In this work, we develop a systematic pipeline that harnesses human feedback to mitigate these problems and refine the video generation model. Specifically, we begin by constructing a large-scale human preference dataset focused on modern video generation models, incorporating pairwise annotations across multi-dimensions. We then introduce VideoReward, a multi-dimensional video reward model, and examine how annotations and various design choices impact its rewarding efficacy. From a unified reinforcement learning perspective aimed at maximizing reward with KL regularization, we introduce three alignment algorithms for flow-based models by extending those from diffusion models. These include two training-time strategies: direct preference optimization for flow (Flow-DPO) and reward weighted regression for flow (Flow-RWR), and an inference-time technique, Flow-NRG, which applies reward guidance directly to noisy videos. Experimental results indicate that VideoReward significantly outperforms existing reward models, and Flow-DPO demonstrates superior performance compared to both Flow-RWR and standard supervised fine-tuning methods. Additionally, Flow-NRG lets users assign custom weights to multiple objectives during inference, meeting personalized video quality needs. Project page: https://gongyeliu.github.io/videoalign.