Xiaodong Cun

Weihuang Liu, Xiaodong Cun, Chi-Man Pun et al. · tsinghua

Image inpainting aims to fill the missing hole of the input. It is hard to solve this task efficiently when facing high-resolution images due to two reasons: (1) Large reception field needs to be handled for high-resolution image inpainting. (2) The general encoder and decoder network synthesizes many background pixels synchronously due to the form of the image matrix. In this paper, we try to break the above limitations for the first time thanks to the recent development of continuous implicit representation. In detail, we down-sample and encode the degraded image to produce the spatial-adaptive parameters for each spatial patch via an attentional Fast Fourier Convolution(FFC)-based parameter generation network. Then, we take these parameters as the weights and biases of a series of multi-layer perceptron(MLP), where the input is the encoded continuous coordinates and the output is the synthesized color value. Thanks to the proposed structure, we only encode the high-resolution image in a relatively low resolution for larger reception field capturing. Then, the continuous position encoding will be helpful to synthesize the photo-realistic high-frequency textures by re-sampling the coordinate in a higher resolution. Also, our framework enables us to query the coordinates of missing pixels only in parallel, yielding a more efficient solution than the previous methods. Experiments show that the proposed method achieves real-time performance on the 2048$\times$2048 images using a single GTX 2080 Ti GPU and can handle 4096$\times$4096 images, with much better performance than existing state-of-the-art methods visually and numerically. The code is available at: https://github.com/NiFangBaAGe/CoordFill.

Haoxin Chen, Menghan Xia, Yingqing He et al.

Video generation has increasingly gained interest in both academia and industry. Although commercial tools can generate plausible videos, there is a limited number of open-source models available for researchers and engineers. In this work, we introduce two diffusion models for high-quality video generation, namely text-to-video (T2V) and image-to-video (I2V) models. T2V models synthesize a video based on a given text input, while I2V models incorporate an additional image input. Our proposed T2V model can generate realistic and cinematic-quality videos with a resolution of $1024 \times 576$, outperforming other open-source T2V models in terms of quality. The I2V model is designed to produce videos that strictly adhere to the content of the provided reference image, preserving its content, structure, and style. This model is the first open-source I2V foundation model capable of transforming a given image into a video clip while maintaining content preservation constraints. We believe that these open-source video generation models will contribute significantly to the technological advancements within the community.

Weihuang Liu, Xi Shen, Chi-Man Pun et al.

We consider the generic problem of detecting low-level structures in images, which includes segmenting the manipulated parts, identifying out-of-focus pixels, separating shadow regions, and detecting concealed objects. Whereas each such topic has been typically addressed with a domain-specific solution, we show that a unified approach performs well across all of them. We take inspiration from the widely-used pre-training and then prompt tuning protocols in NLP and propose a new visual prompting model, named Explicit Visual Prompting (EVP). Different from the previous visual prompting which is typically a dataset-level implicit embedding, our key insight is to enforce the tunable parameters focusing on the explicit visual content from each individual image, i.e., the features from frozen patch embeddings and the input's high-frequency components. The proposed EVP significantly outperforms other parameter-efficient tuning protocols under the same amount of tunable parameters (5.7% extra trainable parameters of each task). EVP also achieves state-of-the-art performances on diverse low-level structure segmentation tasks compared to task-specific solutions. Our code is available at: https://github.com/NiFangBaAGe/Explicit-Visual-Prompt.

Yaofang Liu, Xiaodong Cun, Xuebo Liu et al.

The vision and language generative models have been overgrown in recent years. For video generation, various open-sourced models and public-available services have been developed to generate high-quality videos. However, these methods often use a few metrics, e.g., FVD or IS, to evaluate the performance. We argue that it is hard to judge the large conditional generative models from the simple metrics since these models are often trained on very large datasets with multi-aspect abilities. Thus, we propose a novel framework and pipeline for exhaustively evaluating the performance of the generated videos. Our approach involves generating a diverse and comprehensive list of 700 prompts for text-to-video generation, which is based on an analysis of real-world user data and generated with the assistance of a large language model. Then, we evaluate the state-of-the-art video generative models on our carefully designed benchmark, in terms of visual qualities, content qualities, motion qualities, and text-video alignment with 17 well-selected objective metrics. To obtain the final leaderboard of the models, we further fit a series of coefficients to align the objective metrics to the users' opinions. Based on the proposed human alignment method, our final score shows a higher correlation than simply averaging the metrics, showing the effectiveness of the proposed evaluation method.

Chenyang Qi, Xiaodong Cun, Yong Zhang et al. · tsinghua

The diffusion-based generative models have achieved remarkable success in text-based image generation. However, since it contains enormous randomness in generation progress, it is still challenging to apply such models for real-world visual content editing, especially in videos. In this paper, we propose FateZero, a zero-shot text-based editing method on real-world videos without per-prompt training or use-specific mask. To edit videos consistently, we propose several techniques based on the pre-trained models. Firstly, in contrast to the straightforward DDIM inversion technique, our approach captures intermediate attention maps during inversion, which effectively retain both structural and motion information. These maps are directly fused in the editing process rather than generated during denoising. To further minimize semantic leakage of the source video, we then fuse self-attentions with a blending mask obtained by cross-attention features from the source prompt. Furthermore, we have implemented a reform of the self-attention mechanism in denoising UNet by introducing spatial-temporal attention to ensure frame consistency. Yet succinct, our method is the first one to show the ability of zero-shot text-driven video style and local attribute editing from the trained text-to-image model. We also have a better zero-shot shape-aware editing ability based on the text-to-video model. Extensive experiments demonstrate our superior temporal consistency and editing capability than previous works.

Zinuo Li, Xuhang Chen, Chi-Man Pun et al.

Shadows often occur when we capture the documents with casual equipment, which influences the visual quality and readability of the digital copies. Different from the algorithms for natural shadow removal, the algorithms in document shadow removal need to preserve the details of fonts and figures in high-resolution input. Previous works ignore this problem and remove the shadows via approximate attention and small datasets, which might not work in real-world situations. We handle high-resolution document shadow removal directly via a larger-scale real-world dataset and a carefully designed frequency-aware network. As for the dataset, we acquire over 7k couples of high-resolution (2462 x 3699) images of real-world document pairs with various samples under different lighting circumstances, which is 10 times larger than existing datasets. As for the design of the network, we decouple the high-resolution images in the frequency domain, where the low-frequency details and high-frequency boundaries can be effectively learned via the carefully designed network structure. Powered by our network and dataset, the proposed method clearly shows a better performance than previous methods in terms of visual quality and numerical results. The code, models, and dataset are available at: https://github.com/CXH-Research/DocShadow-SD7K

Kun Cheng, Xiaodong Cun, Yong Zhang et al. · tsinghua

We present VideoReTalking, a new system to edit the faces of a real-world talking head video according to input audio, producing a high-quality and lip-syncing output video even with a different emotion. Our system disentangles this objective into three sequential tasks: (1) face video generation with a canonical expression; (2) audio-driven lip-sync; and (3) face enhancement for improving photo-realism. Given a talking-head video, we first modify the expression of each frame according to the same expression template using the expression editing network, resulting in a video with the canonical expression. This video, together with the given audio, is then fed into the lip-sync network to generate a lip-syncing video. Finally, we improve the photo-realism of the synthesized faces through an identity-aware face enhancement network and post-processing. We use learning-based approaches for all three steps and all our modules can be tackled in a sequential pipeline without any user intervention. Furthermore, our system is a generic approach that does not need to be retrained to a specific person. Evaluations on two widely-used datasets and in-the-wild examples demonstrate the superiority of our framework over other state-of-the-art methods in terms of lip-sync accuracy and visual quality.

Jinbo Xing, Menghan Xia, Yuxin Liu et al. · tsinghua

Creating a vivid video from the event or scenario in our imagination is a truly fascinating experience. Recent advancements in text-to-video synthesis have unveiled the potential to achieve this with prompts only. While text is convenient in conveying the overall scene context, it may be insufficient to control precisely. In this paper, we explore customized video generation by utilizing text as context description and motion structure (e.g. frame-wise depth) as concrete guidance. Our method, dubbed Make-Your-Video, involves joint-conditional video generation using a Latent Diffusion Model that is pre-trained for still image synthesis and then promoted for video generation with the introduction of temporal modules. This two-stage learning scheme not only reduces the computing resources required, but also improves the performance by transferring the rich concepts available in image datasets solely into video generation. Moreover, we use a simple yet effective causal attention mask strategy to enable longer video synthesis, which mitigates the potential quality degradation effectively. Experimental results show the superiority of our method over existing baselines, particularly in terms of temporal coherence and fidelity to users' guidance. In addition, our model enables several intriguing applications that demonstrate potential for practical usage.

Ge Yuan, Xiaodong Cun, Yong Zhang et al. · tsinghua

Exquisite demand exists for customizing the pretrained large text-to-image model, $\textit{e.g.}$, Stable Diffusion, to generate innovative concepts, such as the users themselves. However, the newly-added concept from previous customization methods often shows weaker combination abilities than the original ones even given several images during training. We thus propose a new personalization method that allows for the seamless integration of a unique individual into the pre-trained diffusion model using just $\textbf{one facial photograph}$ and only $\textbf{1024 learnable parameters}$ under $\textbf{3 minutes}$. So as we can effortlessly generate stunning images of this person in any pose or position, interacting with anyone and doing anything imaginable from text prompts. To achieve this, we first analyze and build a well-defined celeb basis from the embedding space of the pre-trained large text encoder. Then, given one facial photo as the target identity, we generate its own embedding by optimizing the weight of this basis and locking all other parameters. Empowered by the proposed celeb basis, the new identity in our customized model showcases a better concept combination ability than previous personalization methods. Besides, our model can also learn several new identities at once and interact with each other where the previous customization model fails to. The code will be released.

Yingqing He, Menghan Xia, Haoxin Chen et al. · tsinghua

Generating videos for visual storytelling can be a tedious and complex process that typically requires either live-action filming or graphics animation rendering. To bypass these challenges, our key idea is to utilize the abundance of existing video clips and synthesize a coherent storytelling video by customizing their appearances. We achieve this by developing a framework comprised of two functional modules: (i) Motion Structure Retrieval, which provides video candidates with desired scene or motion context described by query texts, and (ii) Structure-Guided Text-to-Video Synthesis, which generates plot-aligned videos under the guidance of motion structure and text prompts. For the first module, we leverage an off-the-shelf video retrieval system and extract video depths as motion structure. For the second module, we propose a controllable video generation model that offers flexible controls over structure and characters. The videos are synthesized by following the structural guidance and appearance instruction. To ensure visual consistency across clips, we propose an effective concept personalization approach, which allows the specification of the desired character identities through text prompts. Extensive experiments demonstrate that our approach exhibits significant advantages over various existing baselines.

Wenxuan Zhang, Xiaodong Cun, Xuan Wang et al.

Generating talking head videos through a face image and a piece of speech audio still contains many challenges. ie, unnatural head movement, distorted expression, and identity modification. We argue that these issues are mainly because of learning from the coupled 2D motion fields. On the other hand, explicitly using 3D information also suffers problems of stiff expression and incoherent video. We present SadTalker, which generates 3D motion coefficients (head pose, expression) of the 3DMM from audio and implicitly modulates a novel 3D-aware face render for talking head generation. To learn the realistic motion coefficients, we explicitly model the connections between audio and different types of motion coefficients individually. Precisely, we present ExpNet to learn the accurate facial expression from audio by distilling both coefficients and 3D-rendered faces. As for the head pose, we design PoseVAE via a conditional VAE to synthesize head motion in different styles. Finally, the generated 3D motion coefficients are mapped to the unsupervised 3D keypoints space of the proposed face render, and synthesize the final video. We conducted extensive experiments to demonstrate the superiority of our method in terms of motion and video quality.

Yue Ma, Yingqing He, Xiaodong Cun et al.

Generating text-editable and pose-controllable character videos have an imperious demand in creating various digital human. Nevertheless, this task has been restricted by the absence of a comprehensive dataset featuring paired video-pose captions and the generative prior models for videos. In this work, we design a novel two-stage training scheme that can utilize easily obtained datasets (i.e.,image pose pair and pose-free video) and the pre-trained text-to-image (T2I) model to obtain the pose-controllable character videos. Specifically, in the first stage, only the keypoint-image pairs are used only for a controllable text-to-image generation. We learn a zero-initialized convolutional encoder to encode the pose information. In the second stage, we finetune the motion of the above network via a pose-free video dataset by adding the learnable temporal self-attention and reformed cross-frame self-attention blocks. Powered by our new designs, our method successfully generates continuously pose-controllable character videos while keeps the editing and concept composition ability of the pre-trained T2I model. The code and models will be made publicly available.

Jianrong Zhang, Yangsong Zhang, Xiaodong Cun et al.

In this work, we investigate a simple and must-known conditional generative framework based on Vector Quantised-Variational AutoEncoder (VQ-VAE) and Generative Pre-trained Transformer (GPT) for human motion generation from textural descriptions. We show that a simple CNN-based VQ-VAE with commonly used training recipes (EMA and Code Reset) allows us to obtain high-quality discrete representations. For GPT, we incorporate a simple corruption strategy during the training to alleviate training-testing discrepancy. Despite its simplicity, our T2M-GPT shows better performance than competitive approaches, including recent diffusion-based approaches. For example, on HumanML3D, which is currently the largest dataset, we achieve comparable performance on the consistency between text and generated motion (R-Precision), but with FID 0.116 largely outperforming MotionDiffuse of 0.630. Additionally, we conduct analyses on HumanML3D and observe that the dataset size is a limitation of our approach. Our work suggests that VQ-VAE still remains a competitive approach for human motion generation.

Fei Yin, Yong Zhang, Xiaodong Cun et al.

One-shot talking face generation aims at synthesizing a high-quality talking face video from an arbitrary portrait image, driven by a video or an audio segment. One challenging quality factor is the resolution of the output video: higher resolution conveys more details. In this work, we investigate the latent feature space of a pre-trained StyleGAN and discover some excellent spatial transformation properties. Upon the observation, we explore the possibility of using a pre-trained StyleGAN to break through the resolution limit of training datasets. We propose a novel unified framework based on a pre-trained StyleGAN that enables a set of powerful functionalities, i.e., high-resolution video generation, disentangled control by driving video or audio, and flexible face editing. Our framework elevates the resolution of the synthesized talking face to 1024*1024 for the first time, even though the training dataset has a lower resolution. We design a video-based motion generation module and an audio-based one, which can be plugged into the framework either individually or jointly to drive the video generation. The predicted motion is used to transform the latent features of StyleGAN for visual animation. To compensate for the transformation distortion, we propose a calibration network as well as a domain loss to refine the features. Moreover, our framework allows two types of facial editing, i.e., global editing via GAN inversion and intuitive editing based on 3D morphable models. Comprehensive experiments show superior video quality, flexible controllability, and editability over state-of-the-art methods.

Yuan Gong, Yong Zhang, Xiaodong Cun et al. · tsinghua

We target cross-domain face reenactment in this paper, i.e., driving a cartoon image with the video of a real person and vice versa. Recently, many works have focused on one-shot talking face generation to drive a portrait with a real video, i.e., within-domain reenactment. Straightforwardly applying those methods to cross-domain animation will cause inaccurate expression transfer, blur effects, and even apparent artifacts due to the domain shift between cartoon and real faces. Only a few works attempt to settle cross-domain face reenactment. The most related work AnimeCeleb requires constructing a dataset with pose vector and cartoon image pairs by animating 3D characters, which makes it inapplicable anymore if no paired data is available. In this paper, we propose a novel method for cross-domain reenactment without paired data. Specifically, we propose a transformer-based framework to align the motions from different domains into a common latent space where motion transfer is conducted via latent code addition. Two domain-specific motion encoders and two learnable motion base memories are used to capture domain properties. A source query transformer and a driving one are exploited to project domain-specific motion to the canonical space. The edited motion is projected back to the domain of the source with a transformer. Moreover, since no paired data is provided, we propose a novel cross-domain training scheme using data from two domains with the designed analogy constraint. Besides, we contribute a cartoon dataset in Disney style. Extensive evaluations demonstrate the superiority of our method over competing methods.

Jinbo Xing, Menghan Xia, Yuechen Zhang et al.

Speech-driven 3D facial animation has been widely studied, yet there is still a gap to achieving realism and vividness due to the highly ill-posed nature and scarcity of audio-visual data. Existing works typically formulate the cross-modal mapping into a regression task, which suffers from the regression-to-mean problem leading to over-smoothed facial motions. In this paper, we propose to cast speech-driven facial animation as a code query task in a finite proxy space of the learned codebook, which effectively promotes the vividness of the generated motions by reducing the cross-modal mapping uncertainty. The codebook is learned by self-reconstruction over real facial motions and thus embedded with realistic facial motion priors. Over the discrete motion space, a temporal autoregressive model is employed to sequentially synthesize facial motions from the input speech signal, which guarantees lip-sync as well as plausible facial expressions. We demonstrate that our approach outperforms current state-of-the-art methods both qualitatively and quantitatively. Also, a user study further justifies our superiority in perceptual quality.

Wenbo Hu, Xiangjun Gao, Xiaoyu Li et al.

Estimating video depth in open-world scenarios is challenging due to the diversity of videos in appearance, content motion, camera movement, and length. We present DepthCrafter, an innovative method for generating temporally consistent long depth sequences with intricate details for open-world videos, without requiring any supplementary information such as camera poses or optical flow. The generalization ability to open-world videos is achieved by training the video-to-depth model from a pre-trained image-to-video diffusion model, through our meticulously designed three-stage training strategy. Our training approach enables the model to generate depth sequences with variable lengths at one time, up to 110 frames, and harvest both precise depth details and rich content diversity from realistic and synthetic datasets. We also propose an inference strategy that can process extremely long videos through segment-wise estimation and seamless stitching. Comprehensive evaluations on multiple datasets reveal that DepthCrafter achieves state-of-the-art performance in open-world video depth estimation under zero-shot settings. Furthermore, DepthCrafter facilitates various downstream applications, including depth-based visual effects and conditional video generation.

Youxin Pang, Yong Zhang, Weize Quan et al.

One-shot video-driven talking face generation aims at producing a synthetic talking video by transferring the facial motion from a video to an arbitrary portrait image. Head pose and facial expression are always entangled in facial motion and transferred simultaneously. However, the entanglement sets up a barrier for these methods to be used in video portrait editing directly, where it may require to modify the expression only while maintaining the pose unchanged. One challenge of decoupling pose and expression is the lack of paired data, such as the same pose but different expressions. Only a few methods attempt to tackle this challenge with the feat of 3D Morphable Models (3DMMs) for explicit disentanglement. But 3DMMs are not accurate enough to capture facial details due to the limited number of Blenshapes, which has side effects on motion transfer. In this paper, we introduce a novel self-supervised disentanglement framework to decouple pose and expression without 3DMMs and paired data, which consists of a motion editing module, a pose generator, and an expression generator. The editing module projects faces into a latent space where pose motion and expression motion can be disentangled, and the pose or expression transfer can be performed in the latent space conveniently via addition. The two generators render the modified latent codes to images, respectively. Moreover, to guarantee the disentanglement, we propose a bidirectional cyclic training strategy with well-designed constraints. Evaluations demonstrate our method can control pose or expression independently and be used for general video editing.

Fei Yin, Yong Zhang, Xuan Wang et al.

Recently, a surge of high-quality 3D-aware GANs have been proposed, which leverage the generative power of neural rendering. It is natural to associate 3D GANs with GAN inversion methods to project a real image into the generator's latent space, allowing free-view consistent synthesis and editing, referred as 3D GAN inversion. Although with the facial prior preserved in pre-trained 3D GANs, reconstructing a 3D portrait with only one monocular image is still an ill-pose problem. The straightforward application of 2D GAN inversion methods focuses on texture similarity only while ignoring the correctness of 3D geometry shapes. It may raise geometry collapse effects, especially when reconstructing a side face under an extreme pose. Besides, the synthetic results in novel views are prone to be blurry. In this work, we propose a novel method to promote 3D GAN inversion by introducing facial symmetry prior. We design a pipeline and constraints to make full use of the pseudo auxiliary view obtained via image flipping, which helps obtain a robust and reasonable geometry shape during the inversion process. To enhance texture fidelity in unobserved viewpoints, pseudo labels from depth-guided 3D warping can provide extra supervision. We design constraints aimed at filtering out conflict areas for optimization in asymmetric situations. Comprehensive quantitative and qualitative evaluations on image reconstruction and editing demonstrate the superiority of our method.

Yingqing He, Shaoshu Yang, Haoxin Chen et al.

In this work, we investigate the capability of generating images from pre-trained diffusion models at much higher resolutions than the training image sizes. In addition, the generated images should have arbitrary image aspect ratios. When generating images directly at a higher resolution, 1024 x 1024, with the pre-trained Stable Diffusion using training images of resolution 512 x 512, we observe persistent problems of object repetition and unreasonable object structures. Existing works for higher-resolution generation, such as attention-based and joint-diffusion approaches, cannot well address these issues. As a new perspective, we examine the structural components of the U-Net in diffusion models and identify the crucial cause as the limited perception field of convolutional kernels. Based on this key observation, we propose a simple yet effective re-dilation that can dynamically adjust the convolutional perception field during inference. We further propose the dispersed convolution and noise-damped classifier-free guidance, which can enable ultra-high-resolution image generation (e.g., 4096 x 4096). Notably, our approach does not require any training or optimization. Extensive experiments demonstrate that our approach can address the repetition issue well and achieve state-of-the-art performance on higher-resolution image synthesis, especially in texture details. Our work also suggests that a pre-trained diffusion model trained on low-resolution images can be directly used for high-resolution visual generation without further tuning, which may provide insights for future research on ultra-high-resolution image and video synthesis.

Xiaodong Cun, Zhendong Wang, Chi-Man Pun et al.

Color constancy aims to restore the constant colors of a scene under different illuminants. However, due to the existence of camera spectral sensitivity, the network trained on a certain sensor, cannot work well on others. Also, since the training datasets are collected in certain environments, the diversity of illuminants is limited for complex real world prediction. In this paper, we tackle these problems via two aspects. First, we propose cross-sensor self-supervised training to train the network. In detail, we consider both the general sRGB images and the white-balanced RAW images from current available datasets as the white-balanced agents. Then, we train the network by randomly sampling the artificial illuminants in a sensor-independent manner for scene relighting and supervision. Second, we analyze a previous cascaded framework and present a more compact and accurate model by sharing the backbone parameters with learning attention specifically. Experiments show that our cross-sensor model and single-sensor model outperform other state-of-the-art methods by a large margin on cross and single sensor evaluations, respectively, with only 16% parameters of the previous best model.

Xuhang Chen, Xiaodong Cun, Chi-Man Pun et al.

Shadow removal improves the visual quality and legibility of digital copies of documents. However, document shadow removal remains an unresolved subject. Traditional techniques rely on heuristics that vary from situation to situation. Given the quality and quantity of current public datasets, the majority of neural network models are ill-equipped for this task. In this paper, we propose a Transformer-based model for document shadow removal that utilizes shadow context encoding and decoding in both shadow and shadow-free regions. Additionally, shadow detection and pixel-level enhancement are included in the whole coarse-to-fine process. On the basis of comprehensive benchmark evaluations, it is competitive with state-of-the-art methods.

Yihao Zhi, Xiaodong Cun, Xuelin Chen et al.

Gestures are non-verbal but important behaviors accompanying people's speech. While previous methods are able to generate speech rhythm-synchronized gestures, the semantic context of the speech is generally lacking in the gesticulations. Although semantic gestures do not occur very regularly in human speech, they are indeed the key for the audience to understand the speech context in a more immersive environment. Hence, we introduce LivelySpeaker, a framework that realizes semantics-aware co-speech gesture generation and offers several control handles. In particular, our method decouples the task into two stages: script-based gesture generation and audio-guided rhythm refinement. Specifically, the script-based gesture generation leverages the pre-trained CLIP text embeddings as the guidance for generating gestures that are highly semantically aligned with the script. Then, we devise a simple but effective diffusion-based gesture generation backbone simply using pure MLPs, that is conditioned on only audio signals and learns to gesticulate with realistic motions. We utilize such powerful prior to rhyme the script-guided gestures with the audio signals, notably in a zero-shot setting. Our novel two-stage generation framework also enables several applications, such as changing the gesticulation style, editing the co-speech gestures via textual prompting, and controlling the semantic awareness and rhythm alignment with guided diffusion. Extensive experiments demonstrate the advantages of the proposed framework over competing methods. In addition, our core diffusion-based generative model also achieves state-of-the-art performance on two benchmarks. The code and model will be released to facilitate future research.

Sijie Zhao, Wenbo Hu, Xiaodong Cun et al.

This paper presents a novel framework for converting 2D videos to immersive stereoscopic 3D, addressing the growing demand for 3D content in immersive experience. Leveraging foundation models as priors, our approach overcomes the limitations of traditional methods and boosts the performance to ensure the high-fidelity generation required by the display devices. The proposed system consists of two main steps: depth-based video splatting for warping and extracting occlusion mask, and stereo video inpainting. We utilize pre-trained stable video diffusion as the backbone and introduce a fine-tuning protocol for the stereo video inpainting task. To handle input video with varying lengths and resolutions, we explore auto-regressive strategies and tiled processing. Finally, a sophisticated data processing pipeline has been developed to reconstruct a large-scale and high-quality dataset to support our training. Our framework demonstrates significant improvements in 2D-to-3D video conversion, offering a practical solution for creating immersive content for 3D devices like Apple Vision Pro and 3D displays. In summary, this work contributes to the field by presenting an effective method for generating high-quality stereoscopic videos from monocular input, potentially transforming how we experience digital media.

Yudian Zheng, Xiaodong Cun, Menghan Xia et al.

Understanding semantic intricacies and high-level concepts is essential in image sketch generation, and this challenge becomes even more formidable when applied to the domain of videos. To address this, we propose a novel optimization-based framework for sketching videos represented by the frame-wise Bézier curve. In detail, we first propose a cross-frame stroke initialization approach to warm up the location and the width of each curve. Then, we optimize the locations of these curves by utilizing a semantic loss based on CLIP features and a newly designed consistency loss using the self-decomposed 2D atlas network. Built upon these design elements, the resulting sketch video showcases impressive visual abstraction and temporal coherence. Furthermore, by transforming a video into SVG lines through the sketching process, our method unlocks applications in sketch-based video editing and video doodling, enabled through video composition, as exemplified in the teaser.

Qinyu Yang, Haoxin Chen, Yong Zhang et al.

In order to improve the quality of synthesized videos, currently, one predominant method involves retraining an expert diffusion model and then implementing a noising-denoising process for refinement. Despite the significant training costs, maintaining consistency of content between the original and enhanced videos remains a major challenge. To tackle this challenge, we propose a novel formulation that considers both visual quality and consistency of content. Consistency of content is ensured by a proposed loss function that maintains the structure of the input, while visual quality is improved by utilizing the denoising process of pretrained diffusion models. To address the formulated optimization problem, we have developed a plug-and-play noise optimization strategy, referred to as Noise Calibration. By refining the initial random noise through a few iterations, the content of original video can be largely preserved, and the enhancement effect demonstrates a notable improvement. Extensive experiments have demonstrated the effectiveness of the proposed method.

Shihua Huang, Zhichao Lu, Xiaodong Cun et al.

We introduce DEIM, an innovative and efficient training framework designed to accelerate convergence in real-time object detection with Transformer-based architectures (DETR). To mitigate the sparse supervision inherent in one-to-one (O2O) matching in DETR models, DEIM employs a Dense O2O matching strategy. This approach increases the number of positive samples per image by incorporating additional targets, using standard data augmentation techniques. While Dense O2O matching speeds up convergence, it also introduces numerous low-quality matches that could affect performance. To address this, we propose the Matchability-Aware Loss (MAL), a novel loss function that optimizes matches across various quality levels, enhancing the effectiveness of Dense O2O. Extensive experiments on the COCO dataset validate the efficacy of DEIM. When integrated with RT-DETR and D-FINE, it consistently boosts performance while reducing training time by 50%. Notably, paired with RT-DETRv2, DEIM achieves 53.2% AP in a single day of training on an NVIDIA 4090 GPU. Additionally, DEIM-trained real-time models outperform leading real-time object detectors, with DEIM-D-FINE-L and DEIM-D-FINE-X achieving 54.7% and 56.5% AP at 124 and 78 FPS on an NVIDIA T4 GPU, respectively, without the need for additional data. We believe DEIM sets a new baseline for advancements in real-time object detection. Our code and pre-trained models are available at https://github.com/ShihuaHuang95/DEIM.

Weihuang Liu, Xi Shen, Haolun Li et al.

Zero-shot Video Object Segmentation (ZSVOS) aims at segmenting the primary moving object without any human annotations. Mainstream solutions mainly focus on learning a single model on large-scale video datasets, which struggle to generalize to unseen videos. In this work, we introduce a test-time training (TTT) strategy to address the problem. Our key insight is to enforce the model to predict consistent depth during the TTT process. In detail, we first train a single network to perform both segmentation and depth prediction tasks. This can be effectively learned with our specifically designed depth modulation layer. Then, for the TTT process, the model is updated by predicting consistent depth maps for the same frame under different data augmentations. In addition, we explore different TTT weight updating strategies. Our empirical results suggest that the momentum-based weight initialization and looping-based training scheme lead to more stable improvements. Experiments show that the proposed method achieves clear improvements on ZSVOS. Our proposed video TTT strategy provides significant superiority over state-of-the-art TTT methods. Our code is available at: https://nifangbaage.github.io/DATTT.

Yue Ma, Xiaodong Cun, Sen Liang et al.

Text-based video editing has recently attracted considerable interest in changing the style or replacing the objects with a similar structure. Beyond this, we demonstrate that properties such as shape, size, location, motion, etc., can also be edited in videos. Our key insight is that the keyframe transformations of the specific internal feature (e.g., edge maps of objects or human pose), can easily propagate to other frames to provide generation guidance. We thus propose MagicStick, a controllable video editing method that edits the video properties by utilizing the transformation on the extracted internal control signals. In detail, to keep the appearance, we inflate both the pretrained image diffusion model and ControlNet to the temporal dimension and train low-rank adaptions (LORA) layers to fit the specific scenes. Then, in editing, we perform an inversion and editing framework. Differently, finetuned ControlNet is introduced in both inversion and generation for attention guidance with the proposed attention remix between the spatial attention maps of inversion and editing. Yet succinct, our method is the first method to show the ability of video property editing from the pre-trained text-to-image model. We present experiments on numerous examples within our unified framework. We also compare with shape-aware text-based editing and handcrafted motion video generation, demonstrating our superior temporal consistency and editing capability than previous works. The code and models are available on https://github.com/mayuelala/MagicStick.

Ziyao Huang, Fan Tang, Yong Zhang et al.

Despite the remarkable process of talking-head-based avatar-creating solutions, directly generating anchor-style videos with full-body motions remains challenging. In this study, we propose Make-Your-Anchor, a novel system necessitating only a one-minute video clip of an individual for training, subsequently enabling the automatic generation of anchor-style videos with precise torso and hand movements. Specifically, we finetune a proposed structure-guided diffusion model on input video to render 3D mesh conditions into human appearances. We adopt a two-stage training strategy for the diffusion model, effectively binding movements with specific appearances. To produce arbitrary long temporal video, we extend the 2D U-Net in the frame-wise diffusion model to a 3D style without additional training cost, and a simple yet effective batch-overlapped temporal denoising module is proposed to bypass the constraints on video length during inference. Finally, a novel identity-specific face enhancement module is introduced to improve the visual quality of facial regions in the output videos. Comparative experiments demonstrate the effectiveness and superiority of the system in terms of visual quality, temporal coherence, and identity preservation, outperforming SOTA diffusion/non-diffusion methods. Project page: \url{https://github.com/ICTMCG/Make-Your-Anchor}.

Ziyi Xu, Ziyao Huang, Juan Cao et al.

The generation of anchor-style product promotion videos presents promising opportunities in e-commerce, advertising, and consumer engagement. Despite advancements in pose-guided human video generation, creating product promotion videos remains challenging. In addressing this challenge, we identify the integration of human-object interactions (HOI) into pose-guided human video generation as a core issue. To this end, we introduce AnchorCrafter, a novel diffusion-based system designed to generate 2D videos featuring a target human and a customized object, achieving high visual fidelity and controllable interactions. Specifically, we propose two key innovations: the HOI-appearance perception, which enhances object appearance recognition from arbitrary multi-view perspectives and disentangles object and human appearance, and the HOI-motion injection, which enables complex human-object interactions by overcoming challenges in object trajectory conditioning and inter-occlusion management. Extensive experiments show that our system improves object appearance preservation by 7.5\% and doubles the object localization accuracy compared to existing state-of-the-art approaches. It also outperforms existing approaches in maintaining human motion consistency and high-quality video generation. Project page including data, code, and Huggingface demo: https://github.com/cangcz/AnchorCrafter.

Haoxin Chen, Yong Zhang, Xiaodong Cun et al.

Text-to-video generation aims to produce a video based on a given prompt. Recently, several commercial video models have been able to generate plausible videos with minimal noise, excellent details, and high aesthetic scores. However, these models rely on large-scale, well-filtered, high-quality videos that are not accessible to the community. Many existing research works, which train models using the low-quality WebVid-10M dataset, struggle to generate high-quality videos because the models are optimized to fit WebVid-10M. In this work, we explore the training scheme of video models extended from Stable Diffusion and investigate the feasibility of leveraging low-quality videos and synthesized high-quality images to obtain a high-quality video model. We first analyze the connection between the spatial and temporal modules of video models and the distribution shift to low-quality videos. We observe that full training of all modules results in a stronger coupling between spatial and temporal modules than only training temporal modules. Based on this stronger coupling, we shift the distribution to higher quality without motion degradation by finetuning spatial modules with high-quality images, resulting in a generic high-quality video model. Evaluations are conducted to demonstrate the superiority of the proposed method, particularly in picture quality, motion, and concept composition.

Yaofang Liu, Yumeng Ren, Aitor Artola et al.

The rapid advancement of video diffusion models has been hindered by fundamental limitations in temporal modeling, particularly the rigid synchronization of frame evolution imposed by conventional scalar timestep variables. While task-specific adaptations and autoregressive models have sought to address these challenges, they remain constrained by computational inefficiency, catastrophic forgetting, or narrow applicability. In this work, we present Pusa, a groundbreaking paradigm that leverages vectorized timestep adaptation (VTA) to enable fine-grained temporal control within a unified video diffusion framework. Besides, VTA is a non-destructive adaptation, which means it fully preserves the capabilities of the base model. By finetuning the SOTA Wan2.1-T2V-14B model with VTA, we achieve unprecedented efficiency -- surpassing the performance of Wan-I2V-14B with $\leq$ 1/200 of the training cost (\$500 vs. $\geq$ \$100,000) and $\leq$ 1/2500 of the dataset size (4K vs. $\geq$ 10M samples). Pusa not only sets a new standard for image-to-video (I2V) generation, achieving a VBench-I2V total score of 87.32\% (vs. 86.86\% of Wan-I2V-14B), but also unlocks many zero-shot multi-task capabilities such as start-end frames and video extension -- all without task-specific training. Meanwhile, Pusa can still perform text-to-video generation. Mechanistic analyses reveal that our approach preserves the foundation model's generative priors while surgically injecting temporal dynamics, avoiding the combinatorial explosion inherent to vectorized timesteps. This work establishes a scalable, efficient, and versatile paradigm for next-generation video synthesis, democratizing high-fidelity video generation for research and industry alike. Code is open-sourced at https://github.com/Yaofang-Liu/Pusa-VidGen

Yuzhou Huang, Liangbin Xie, Xintao Wang et al. · tencent-ai

Current instruction-based editing methods, such as InstructPix2Pix, often fail to produce satisfactory results in complex scenarios due to their dependence on the simple CLIP text encoder in diffusion models. To rectify this, this paper introduces SmartEdit, a novel approach to instruction-based image editing that leverages Multimodal Large Language Models (MLLMs) to enhance their understanding and reasoning capabilities. However, direct integration of these elements still faces challenges in situations requiring complex reasoning. To mitigate this, we propose a Bidirectional Interaction Module that enables comprehensive bidirectional information interactions between the input image and the MLLM output. During training, we initially incorporate perception data to boost the perception and understanding capabilities of diffusion models. Subsequently, we demonstrate that a small amount of complex instruction editing data can effectively stimulate SmartEdit's editing capabilities for more complex instructions. We further construct a new evaluation dataset, Reason-Edit, specifically tailored for complex instruction-based image editing. Both quantitative and qualitative results on this evaluation dataset indicate that our SmartEdit surpasses previous methods, paving the way for the practical application of complex instruction-based image editing.

Weihuang Liu, Xi Shen, Chi-Man Pun et al.

Foreground segmentation is a fundamental problem in computer vision, which includes salient object detection, forgery detection, defocus blur detection, shadow detection, and camouflage object detection. Previous works have typically relied on domain-specific solutions to address accuracy and robustness issues in those applications. In this paper, we present a unified framework for a number of foreground segmentation tasks without any task-specific designs. We take inspiration from the widely-used pre-training and then prompt tuning protocols in NLP and propose a new visual prompting model, named Explicit Visual Prompting (EVP). Different from the previous visual prompting which is typically a dataset-level implicit embedding, our key insight is to enforce the tunable parameters focusing on the explicit visual content from each individual image, i.e., the features from frozen patch embeddings and high-frequency components. Our method freezes a pre-trained model and then learns task-specific knowledge using a few extra parameters. Despite introducing only a small number of tunable parameters, EVP achieves superior performance than full fine-tuning and other parameter-efficient fine-tuning methods. Experiments in fourteen datasets across five tasks show the proposed method outperforms other task-specific methods while being considerably simple. The proposed method demonstrates the scalability in different architectures, pre-trained weights, and tasks. The code is available at: https://github.com/NiFangBaAGe/Explicit-Visual-Prompt.

Jingtang Liang, Xiaodong Cun, Chi-Man Pun et al.

Image harmonization aims to modify the color of the composited region with respect to the specific background. Previous works model this task as a pixel-wise image-to-image translation using UNet family structures. However, the model size and computational cost limit the ability of their models on edge devices and higher-resolution images. To this end, we propose a novel spatial-separated curve rendering network(S$^2$CRNet) for efficient and high-resolution image harmonization for the first time. In S$^2$CRNet, we firstly extract the spatial-separated embeddings from the thumbnails of the masked foreground and background individually. Then, we design a curve rendering module(CRM), which learns and combines the spatial-specific knowledge using linear layers to generate the parameters of the piece-wise curve mapping in the foreground region. Finally, we directly render the original high-resolution images using the learned color curve. Besides, we also make two extensions of the proposed framework via the Cascaded-CRM and Semantic-CRM for cascaded refinement and semantic guidance, respectively. Experiments show that the proposed method reduces more than 90% parameters compared with previous methods but still achieves the state-of-the-art performance on both synthesized iHarmony4 and real-world DIH test sets. Moreover, our method can work smoothly on higher resolution images(eg., $2048\times2048$) in 0.1 seconds with much lower GPU computational resources than all existing methods. The code will be made available at \url{http://github.com/stefanLeong/S2CRNet}.

Zhendong Wang, Xiaodong Cun, Jianmin Bao et al.

In this paper, we present Uformer, an effective and efficient Transformer-based architecture for image restoration, in which we build a hierarchical encoder-decoder network using the Transformer block. In Uformer, there are two core designs. First, we introduce a novel locally-enhanced window (LeWin) Transformer block, which performs nonoverlapping window-based self-attention instead of global self-attention. It significantly reduces the computational complexity on high resolution feature map while capturing local context. Second, we propose a learnable multi-scale restoration modulator in the form of a multi-scale spatial bias to adjust features in multiple layers of the Uformer decoder. Our modulator demonstrates superior capability for restoring details for various image restoration tasks while introducing marginal extra parameters and computational cost. Powered by these two designs, Uformer enjoys a high capability for capturing both local and global dependencies for image restoration. To evaluate our approach, extensive experiments are conducted on several image restoration tasks, including image denoising, motion deblurring, defocus deblurring and deraining. Without bells and whistles, our Uformer achieves superior or comparable performance compared with the state-of-the-art algorithms. The code and models are available at https://github.com/ZhendongWang6/Uformer.

Xiaodong Cun, Chi-Man Pun

Digital watermark is a commonly used technique to protect the copyright of medias. Simultaneously, to increase the robustness of watermark, attacking technique, such as watermark removal, also gets the attention from the community. Previous watermark removal methods require to gain the watermark location from users or train a multi-task network to recover the background indiscriminately. However, when jointly learning, the network performs better on watermark detection than recovering the texture. Inspired by this observation and to erase the visible watermarks blindly, we propose a novel two-stage framework with a stacked attention-guided ResUNets to simulate the process of detection, removal and refinement. In the first stage, we design a multi-task network called SplitNet. It learns the basis features for three sub-tasks altogether while the task-specific features separately use multiple channel attentions. Then, with the predicted mask and coarser restored image, we design RefineNet to smooth the watermarked region with a mask-guided spatial attention. Besides network structure, the proposed algorithm also combines multiple perceptual losses for better quality both visually and numerically. We extensively evaluate our algorithm over four different datasets under various settings and the experiments show that our approach outperforms other state-of-the-art methods by a large margin. The code is available at http://github.com/vinthony/deep-blind-watermark-removal.

Xiaodong Cun, Chi-Man Pun

Defocus Blur Detection(DBD) aims to separate in-focus and out-of-focus regions from a single image pixel-wisely. This task has been paid much attention since bokeh effects are widely used in digital cameras and smartphone photography. However, identifying obscure homogeneous regions and borderline transitions in partially defocus images is still challenging. To solve these problems, we introduce depth information into DBD for the first time. When the camera parameters are fixed, we argue that the accuracy of DBD is highly related to scene depth. Hence, we consider the depth information as the approximate soft label of DBD and propose a joint learning framework inspired by knowledge distillation. In detail, we learn the defocus blur from ground truth and the depth distilled from a well-trained depth estimation network at the same time. Thus, the sharp region will provide a strong prior for depth estimation while the blur detection also gains benefits from the distilled depth. Besides, we propose a novel decoder in the fully convolutional network(FCN) as our network structure. In each level of the decoder, we design the Selective Reception Field Block(SRFB) for merging multi-scale features efficiently and reuse the side outputs as Supervision-guided Attention Block(SAB). Unlike previous methods, the proposed decoder builds reception field pyramids and emphasizes salient regions simply and efficiently. Experiments show that our approach outperforms 11 other state-of-the-art methods on two popular datasets. Our method also runs at over 30 fps on a single GPU, which is 2x faster than previous works. The code is available at: https://github.com/vinthony/depth-distillation

Xiaodong Cun, Chi-Man Pun, Cheng Shi

Shadow removal is an essential task for scene understanding. Many studies consider only matching the image contents, which often causes two types of ghosts: color in-consistencies in shadow regions or artifacts on shadow boundaries. In this paper, we tackle these issues in two ways. First, to carefully learn the border artifacts-free image, we propose a novel network structure named the dual hierarchically aggregation network~(DHAN). It contains a series of growth dilated convolutions as the backbone without any down-samplings, and we hierarchically aggregate multi-context features for attention and prediction, respectively. Second, we argue that training on a limited dataset restricts the textural understanding of the network, which leads to the shadow region color in-consistencies. Currently, the largest dataset contains 2k+ shadow/shadow-free image pairs. However, it has only 0.1k+ unique scenes since many samples share exactly the same background with different shadow positions. Thus, we design a shadow matting generative adversarial network~(SMGAN) to synthesize realistic shadow mattings from a given shadow mask and shadow-free image. With the help of novel masks or scenes, we enhance the current datasets using synthesized shadow images. Experiments show that our DHAN can erase the shadows and produce high-quality ghost-free images. After training on the synthesized and real datasets, our network outperforms other state-of-the-art methods by a large margin. The code is available: http://github.com/vinthony/ghost-free-shadow-removal/

Xiaodong Cun, Chi-Man Pun

Image composition is one of the most important applications in image processing. However, the inharmonious appearance between the spliced region and background degrade the quality of the image. Thus, we address the problem of Image Harmonization: Given a spliced image and the mask of the spliced region, we try to harmonize the "style" of the pasted region with the background (non-spliced region). Previous approaches have been focusing on learning directly by the neural network. In this work, we start from an empirical observation: the differences can only be found in the spliced region between the spliced image and the harmonized result while they share the same semantic information and the appearance in the non-spliced region. Thus, in order to learn the feature map in the masked region and the others individually, we propose a novel attention module named Spatial-Separated Attention Module (S2AM). Furthermore, we design a novel image harmonization framework by inserting the S2AM in the coarser low-level features of the Unet structure in two different ways. Besides image harmonization, we make a big step for harmonizing the composite image without the specific mask under previous observation. The experiments show that the proposed S2AM performs better than other state-of-the-art attention modules in our task. Moreover, we demonstrate the advantages of our model against other state-of-the-art image harmonization methods via criteria from multiple points of view. Code is available at https://github.com/vinthony/s2am

Jay Zhangjie Wu, Guian Fang, Haoning Wu et al. · tencent-ai

Generative models have demonstrated remarkable capability in synthesizing high-quality text, images, and videos. For video generation, contemporary text-to-video models exhibit impressive capabilities, crafting visually stunning videos. Nonetheless, evaluating such videos poses significant challenges. Current research predominantly employs automated metrics such as FVD, IS, and CLIP Score. However, these metrics provide an incomplete analysis, particularly in the temporal assessment of video content, thus rendering them unreliable indicators of true video quality. Furthermore, while user studies have the potential to reflect human perception accurately, they are hampered by their time-intensive and laborious nature, with outcomes that are often tainted by subjective bias. In this paper, we investigate the limitations inherent in existing metrics and introduce a novel evaluation pipeline, the Text-to-Video Score (T2VScore). This metric integrates two pivotal criteria: (1) Text-Video Alignment, which scrutinizes the fidelity of the video in representing the given text description, and (2) Video Quality, which evaluates the video's overall production caliber with a mixture of experts. Moreover, to evaluate the proposed metrics and facilitate future improvements on them, we present the TVGE dataset, collecting human judgements of 2,543 text-to-video generated videos on the two criteria. Experiments on the TVGE dataset demonstrate the superiority of the proposed T2VScore on offering a better metric for text-to-video generation.

Lanqing Guo, Yingqing He, Haoxin Chen et al.

Diffusion models have proven to be highly effective in image and video generation; however, they encounter challenges in the correct composition of objects when generating images of varying sizes due to single-scale training data. Adapting large pre-trained diffusion models to higher resolution demands substantial computational and optimization resources, yet achieving generation capabilities comparable to low-resolution models remains challenging. This paper proposes a novel self-cascade diffusion model that leverages the knowledge gained from a well-trained low-resolution image/video generation model, enabling rapid adaptation to higher-resolution generation. Building on this, we employ the pivot replacement strategy to facilitate a tuning-free version by progressively leveraging reliable semantic guidance derived from the low-resolution model. We further propose to integrate a sequence of learnable multi-scale upsampler modules for a tuning version capable of efficiently learning structural details at a new scale from a small amount of newly acquired high-resolution training data. Compared to full fine-tuning, our approach achieves a $5\times$ training speed-up and requires only 0.002M tuning parameters. Extensive experiments demonstrate that our approach can quickly adapt to higher-resolution image and video synthesis by fine-tuning for just $10k$ steps, with virtually no additional inference time.

Minghong Cai, Xiaodong Cun, Xiaoyu Li et al.

Sora-like video generation models have achieved remarkable progress with a Multi-Modal Diffusion Transformer MM-DiT architecture. However, the current video generation models predominantly focus on single-prompt, struggling to generate coherent scenes with multiple sequential prompts that better reflect real-world dynamic scenarios. While some pioneering works have explored multi-prompt video generation, they face significant challenges including strict training data requirements, weak prompt following, and unnatural transitions. To address these problems, we propose DiTCtrl, a training-free multi-prompt video generation method under MM-DiT architectures for the first time. Our key idea is to take the multi-prompt video generation task as temporal video editing with smooth transitions. To achieve this goal, we first analyze MM-DiT's attention mechanism, finding that the 3D full attention behaves similarly to that of the cross/self-attention blocks in the UNet-like diffusion models, enabling mask-guided precise semantic control across different prompts with attention sharing for multi-prompt video generation. Based on our careful design, the video generated by DiTCtrl achieves smooth transitions and consistent object motion given multiple sequential prompts without additional training. Besides, we also present MPVBench, a new benchmark specially designed for multi-prompt video generation to evaluate the performance of multi-prompt generation. Extensive experiments demonstrate that our method achieves state-of-the-art performance without additional training.

Jiwen Yu, Xiaodong Cun, Chenyang Qi et al.

Large-scale text-to-video (T2V) diffusion models have great progress in recent years in terms of visual quality, motion and temporal consistency. However, the generation process is still a black box, where all attributes (e.g., appearance, motion) are learned and generated jointly without precise control ability other than rough text descriptions. Inspired by image animation which decouples the video as one specific appearance with the corresponding motion, we propose AnimateZero to unveil the pre-trained text-to-video diffusion model, i.e., AnimateDiff, and provide more precise appearance and motion control abilities for it. For appearance control, we borrow intermediate latents and their features from the text-to-image (T2I) generation for ensuring the generated first frame is equal to the given generated image. For temporal control, we replace the global temporal attention of the original T2V model with our proposed positional-corrected window attention to ensure other frames align with the first frame well. Empowered by the proposed methods, AnimateZero can successfully control the generating progress without further training. As a zero-shot image animator for given images, AnimateZero also enables multiple new applications, including interactive video generation and real image animation. The detailed experiments demonstrate the effectiveness of the proposed method in both T2V and related applications.

Tao Wu, Yong Zhang, Xiaodong Cun et al.

Zero-shot customized video generation has gained significant attention due to its substantial application potential. Existing methods rely on additional models to extract and inject reference subject features, assuming that the Video Diffusion Model (VDM) alone is insufficient for zero-shot customized video generation. However, these methods often struggle to maintain consistent subject appearance due to suboptimal feature extraction and injection techniques. In this paper, we reveal that VDM inherently possesses the force to extract and inject subject features. Departing from previous heuristic approaches, we introduce a novel framework that leverages VDM's inherent force to enable high-quality zero-shot customized video generation. Specifically, for feature extraction, we directly input reference images into VDM and use its intrinsic feature extraction process, which not only provides fine-grained features but also significantly aligns with VDM's pre-trained knowledge. For feature injection, we devise an innovative bidirectional interaction between subject features and generated content through spatial self-attention within VDM, ensuring that VDM has better subject fidelity while maintaining the diversity of the generated video. Experiments on both customized human and object video generation validate the effectiveness of our framework.

Xiuli Bi, Jian Lu, Bo Liu et al.

Benefiting from large-scale pre-training of text-video pairs, current text-to-video (T2V) diffusion models can generate high-quality videos from the text description. Besides, given some reference images or videos, the parameter-efficient fine-tuning method, i.e. LoRA, can generate high-quality customized concepts, e.g., the specific subject or the motions from a reference video. However, combining the trained multiple concepts from different references into a single network shows obvious artifacts. To this end, we propose CustomTTT, where we can joint custom the appearance and the motion of the given video easily. In detail, we first analyze the prompt influence in the current video diffusion model and find the LoRAs are only needed for the specific layers for appearance and motion customization. Besides, since each LoRA is trained individually, we propose a novel test-time training technique to update parameters after combination utilizing the trained customized models. We conduct detailed experiments to verify the effectiveness of the proposed methods. Our method outperforms several state-of-the-art works in both qualitative and quantitative evaluations.

Zichao Yu, Zhen Zou, Guojiang Shao et al.

Diffusion models have demonstrated remarkable success in generative tasks, yet their iterative denoising process results in slow inference, limiting their practicality. While existing acceleration methods exploit the well-known U-shaped similarity pattern between adjacent steps through caching mechanisms, they lack theoretical foundation and rely on simplistic computation reuse, often leading to performance degradation. In this work, we provide a theoretical understanding by analyzing the denoising process through the second-order Adams-Bashforth method, revealing a linear relationship between the outputs of consecutive steps. This analysis explains why the outputs of adjacent steps exhibit a U-shaped pattern. Furthermore, extending Adams-Bashforth method to higher order, we propose a novel caching-based acceleration approach for diffusion models, instead of directly reusing cached results, with a truncation error bound of only \(O(h^k)\) where $h$ is the step size. Extensive validation across diverse image and video diffusion models (including HunyuanVideo and FLUX.1-dev) with various schedulers demonstrates our method's effectiveness in achieving nearly $3\times$ speedup while maintaining original performance levels, offering a practical real-time solution without compromising generation quality.

Yaowei Li, Lingen Li, Zhaoyang Zhang et al.

As user expectations for image editing continue to rise, the demand for flexible, fine-grained manipulation of specific visual elements presents a challenge for current diffusion-based methods. In this work, we present BlobCtrl, a framework for element-level image editing based on a probabilistic blob-based representation. Treating blobs as visual primitives, BlobCtrl disentangles layout from appearance, affording fine-grained, controllable object-level manipulation. Our key contributions are twofold: (1) an in-context dual-branch diffusion model that separates foreground and background processing, incorporating blob representations to explicitly decouple layout and appearance, and (2) a self-supervised disentangle-then-reconstruct training paradigm with an identity-preserving loss function, along with tailored strategies to efficiently leverage blob-image pairs. To foster further research, we introduce BlobData for large-scale training and BlobBench, a benchmark for systematic evaluation. Experimental results demonstrate that BlobCtrl achieves state-of-the-art performance in a variety of element-level editing tasks, such as object addition, removal, scaling, and replacement, while maintaining computational efficiency. Project Webpage: https://liyaowei-stu.github.io/project/BlobCtrl/

Shaoshu Yang, Yong Zhang, Xiaodong Cun et al.

Video generation has made remarkable progress in recent years, especially since the advent of the video diffusion models. Many video generation models can produce plausible synthetic videos, e.g., Stable Video Diffusion (SVD). However, most video models can only generate low frame rate videos due to the limited GPU memory as well as the difficulty of modeling a large set of frames. The training videos are always uniformly sampled at a specified interval for temporal compression. Previous methods promote the frame rate by either training a video interpolation model in pixel space as a postprocessing stage or training an interpolation model in latent space for a specific base video model. In this paper, we propose a training-free video interpolation method for generative video diffusion models, which is generalizable to different models in a plug-and-play manner. We investigate the non-linearity in the feature space of video diffusion models and transform a video model into a self-cascaded video diffusion model with incorporating the designed hidden state correction modules. The self-cascaded architecture and the correction module are proposed to retain the temporal consistency between key frames and the interpolated frames. Extensive evaluations are preformed on multiple popular video models to demonstrate the effectiveness of the propose method, especially that our training-free method is even comparable to trained interpolation models supported by huge compute resources and large-scale datasets.