Hyeongwoo Kim

Kartik Teotia, Mallikarjun B R, Xingang Pan et al.

Multi-view volumetric rendering techniques have recently shown great potential in modeling and synthesizing high-quality head avatars. A common approach to capture full head dynamic performances is to track the underlying geometry using a mesh-based template or 3D cube-based graphics primitives. While these model-based approaches achieve promising results, they often fail to learn complex geometric details such as the mouth interior, hair, and topological changes over time. This paper presents a novel approach to building highly photorealistic digital head avatars. Our method learns a canonical space via an implicit function parameterized by a neural network. It leverages multiresolution hash encoding in the learned feature space, allowing for high-quality, faster training and high-resolution rendering. At test time, our method is driven by a monocular RGB video. Here, an image encoder extracts face-specific features that also condition the learnable canonical space. This encourages deformation-dependent texture variations during training. We also propose a novel optical flow based loss that ensures correspondences in the learned canonical space, thus encouraging artifact-free and temporally consistent renderings. We show results on challenging facial expressions and show free-viewpoint renderings at interactive real-time rates for medium image resolutions. Our method outperforms all existing approaches, both visually and numerically. We will release our multiple-identity dataset to encourage further research. Our Project page is available at: https://vcai.mpi-inf.mpg.de/projects/HQ3DAvatar/

Kartik Teotia, Hyeongwoo Kim, Pablo Garrido et al.

Real-time rendering of human head avatars is a cornerstone of many computer graphics applications, such as augmented reality, video games, and films, to name a few. Recent approaches address this challenge with computationally efficient geometry primitives in a carefully calibrated multi-view setup. Albeit producing photorealistic head renderings, it often fails to represent complex motion changes such as the mouth interior and strongly varying head poses. We propose a new method to generate highly dynamic and deformable human head avatars from multi-view imagery in real-time. At the core of our method is a hierarchical representation of head models that allows to capture the complex dynamics of facial expressions and head movements. First, with rich facial features extracted from raw input frames, we learn to deform the coarse facial geometry of the template mesh. We then initialize 3D Gaussians on the deformed surface and refine their positions in a fine step. We train this coarse-to-fine facial avatar model along with the head pose as a learnable parameter in an end-to-end framework. This enables not only controllable facial animation via video inputs, but also high-fidelity novel view synthesis of challenging facial expressions, such as tongue deformations and fine-grained teeth structure under large motion changes. Moreover, it encourages the learned head avatar to generalize towards new facial expressions and head poses at inference time. We demonstrate the performance of our method with comparisons against the related methods on different datasets, spanning challenging facial expression sequences across multiple identities. We also show the potential application of our approach by demonstrating a cross-identity facial performance transfer application.

Federico Nocentini, Kwanggyoon Seo, Qingju Liu et al.

Speech-Driven Facial Animation (SDFA) has gained significant attention due to its applications in movies, video games, and virtual reality. However, most existing models are trained on single-language data, limiting their effectiveness in real-world multilingual scenarios. In this work, we address multilingual SDFA, which is essential for realistic generation since language influences phonetics, rhythm, intonation, and facial expressions. Speaking style is also shaped by individual differences, not only by language. Existing methods typically rely on either language-specific or speaker-specific conditioning, but not both, limiting their ability to model their interaction. We introduce Polyglot, a unified diffusion-based architecture for personalized multilingual SDFA. Our method uses transcript embeddings to encode language information and style embeddings extracted from reference facial sequences to capture individual speaking characteristics. Polyglot does not require predefined language or speaker labels, enabling generalization across languages and speakers through self-supervised learning. By jointly conditioning on language and style, it captures expressive traits such as rhythm, articulation, and habitual facial movements, producing temporally coherent and realistic animations. Experiments show improved performance in both monolingual and multilingual settings, providing a unified framework for modeling language and personal style in SDFA.

Taewon Kim, Hyosoon Jang, Hyunjin Seo et al.

Recent advances in generative modeling show that pretrained representations can improve generation as conditioning features or alignment targets. Motivated by this, we study protein representations for predicting structures beyond conventional function annotation. We propose TriProRep, a structure-aware pretraining method that jointly models three aligned residue-level views: amino-acid identity, backbone geometry, and local full-atom geometry, discretely encoded via VQ-VAE tokenizers. By pretraining to recover original tokens from generator-corrupted views, TriProRep learns to distinguish plausible but incorrect cross-view augmentations from the original protein. We further introduce RepSP, a benchmark for evaluating protein representations in structure-predictive settings. RepSP tests three uses of representations: homodimer co-folding from apo-chain representations, residue-level prediction of homodimer-derived interaction properties, and representation-aligned monomer structure prediction. Across these tasks, TriProRep improves over sequence-only and prior structure-aware representation models, while maintaining competitive performance on conventional benchmarks.

Akin Caliskan, Berkay Kicanaoglu, Hyeongwoo Kim

We propose PAV, Personalized Head Avatar for the synthesis of human faces under arbitrary viewpoints and facial expressions. PAV introduces a method that learns a dynamic deformable neural radiance field (NeRF), in particular from a collection of monocular talking face videos of the same character under various appearance and shape changes. Unlike existing head NeRF methods that are limited to modeling such input videos on a per-appearance basis, our method allows for learning multi-appearance NeRFs, introducing appearance embedding for each input video via learnable latent neural features attached to the underlying geometry. Furthermore, the proposed appearance-conditioned density formulation facilitates the shape variation of the character, such as facial hair and soft tissues, in the radiance field prediction. To the best of our knowledge, our approach is the first dynamic deformable NeRF framework to model appearance and shape variations in a single unified network for multi-appearances of the same subject. We demonstrate experimentally that PAV outperforms the baseline method in terms of visual rendering quality in our quantitative and qualitative studies on various subjects.

Pinxin Liu, Pengfei Zhang, Hyeongwoo Kim et al.

Co-speech gesture generation is crucial for creating lifelike avatars and enhancing human-computer interactions by synchronizing gestures with speech. Despite recent advancements, existing methods struggle with accurately identifying the rhythmic or semantic triggers from audio for generating contextualized gesture patterns and achieving pixel-level realism. To address these challenges, we introduce Contextual Gesture, a framework that improves co-speech gesture video generation through three innovative components: (1) a chronological speech-gesture alignment that temporally connects two modalities, (2) a contextualized gesture tokenization that incorporate speech context into motion pattern representation through distillation, and (3) a structure-aware refinement module that employs edge connection to link gesture keypoints to improve video generation. Our extensive experiments demonstrate that Contextual Gesture not only produces realistic and speech-aligned gesture videos but also supports long-sequence generation and video gesture editing applications, shown in Fig.1.

Pengfei Zhang, Pinxin Liu, Pablo Garrido et al.

Current human motion synthesis frameworks rely on global action descriptions, creating a modality gap that limits both motion understanding and generation capabilities. A single coarse description, such as run, fails to capture details such as variations in speed, limb positioning, and kinematic dynamics, leading to ambiguities between text and motion modalities. To address this challenge, we introduce KinMo, a unified framework built on a hierarchical describable motion representation that extends beyond global actions by incorporating kinematic group movements and their interactions. We design an automated annotation pipeline to generate high-quality, fine-grained descriptions for this decomposition, resulting in the KinMo dataset and offering a scalable and cost-efficient solution for dataset enrichment. To leverage these structured descriptions, we propose Hierarchical Text-Motion Alignment that progressively integrates additional motion details, thereby improving semantic motion understanding. Furthermore, we introduce a coarse-to-fine motion generation procedure to leverage enhanced spatial understanding to improve motion synthesis. Experimental results show that KinMo significantly improves motion understanding, demonstrated by enhanced text-motion retrieval performance and enabling more fine-grained motion generation and editing capabilities. Project Page: https://andypinxinliu.github.io/KinMo

Tuomas Varanka, Juan Luis Gonzalez, Hyeongwoo Kim et al.

Existing video deraining methods are often trained on paired datasets, either synthetic, which limits their ability to generalize to real-world rain, or captured by static cameras, which restricts their effectiveness in dynamic scenes with background and camera motion. Furthermore, recent works in fine-tuning diffusion models have shown promising results, but the fine-tuning tends to weaken the generative prior, limiting generalization to unseen cases. In this paper, we introduce the first zero-shot video deraining method for complex dynamic scenes that does not require synthetic data nor model fine-tuning, by leveraging a pretrained text-to-video diffusion model that demonstrates strong generalization capabilities. By inverting an input video into the latent space of diffusion models, its reconstruction process can be intervened and pushed away from the model's concept of rain using negative prompting. At the core of our approach is an attention switching mechanism that we found is crucial for maintaining dynamic backgrounds as well as structural consistency between the input and the derained video, mitigating artifacts introduced by naive negative prompting. Our approach is validated through extensive experiments on real-world rain datasets, demonstrating substantial improvements over prior methods and showcasing robust generalization without the need for supervised training.

Kartik Teotia, Helge Rhodin, Mohit Mendiratta et al.

We introduce the first method for audio-driven universal photorealistic avatar synthesis, combining a person-agnostic speech model with our novel Universal Head Avatar Prior (UHAP). UHAP is trained on cross-identity multi-view videos. In particular, our UHAP is supervised with neutral scan data, enabling it to capture the identity-specific details at high fidelity. In contrast to previous approaches, which predominantly map audio features to geometric deformations only while ignoring audio-dependent appearance variations, our universal speech model directly maps raw audio inputs into the UHAP latent expression space. This expression space inherently encodes, both, geometric and appearance variations. For efficient personalization to new subjects, we employ a monocular encoder, which enables lightweight regression of dynamic expression variations across video frames. By accounting for these expression-dependent changes, it enables the subsequent model fine-tuning stage to focus exclusively on capturing the subject's global appearance and geometry. Decoding these audio-driven expression codes via UHAP generates highly realistic avatars with precise lip synchronization and nuanced expressive details, such as eyebrow movement, gaze shifts, and realistic mouth interior appearance as well as motion. Extensive evaluations demonstrate that our method is not only the first generalizable audio-driven avatar model that can account for detailed appearance modeling and rendering, but it also outperforms competing (geometry-only) methods across metrics measuring lip-sync accuracy, quantitative image quality, and perceptual realism.

Eray Eren, Qingju Liu, Hyeongwoo Kim et al.

Prosody conveys rich emotional and semantic information of the speech signal as well as individual idiosyncrasies. We propose a stand-alone model that maps text-to-prosodic features such as F0 and energy and can be used in downstream tasks such as TTS. The ProMode encoder takes as input acoustic features and time-aligned textual content, both are partially masked, and obtains a fixed-length latent prosodic embedding. The decoder predicts acoustics in the masked region using both the encoded prosody input and unmasked textual content. Trained on the GigaSpeech dataset, we compare our method with state-of-the-art style encoders. For F0 and energy predictions, we show consistent improvements for our model at different levels of granularity. We also integrate these predicted prosodic features into a TTS system and conduct perceptual tests, which show higher prosody preference compared to the baselines, demonstrating the model's potential in tasks where prosody modeling is important.

Nick Yiwen Huang, Akin Caliskan, Berkay Kicanaoglu et al.

We consider the problem of disentangling 3D from large vision-language models, which we show on generative 3D portraits. This allows free-form text control of appearance attributes like age, hair style, and glasses, and 3D geometry control of face expression and camera pose. In this setting, we assume we use a pre-trained large vision-language model (LVLM; CLIP) to generate from a smaller 2D dataset with no additional paired labels and with a pre-defined 3D morphable model (FLAME). First, we disentangle using canonicalization to a 2D reference frame from a deformable neural 3D triplane representation. But another form of entanglement arises from the significant noise in the LVLM's embedding space that describes irrelevant features. This damages output quality and diversity, but we overcome this with a Jacobian regularization that can be computed efficiently with a stochastic approximator. Compared to existing methods, our approach produces portraits with added text and 3D control, where portraits remain consistent when either control is changed. Broadly, this approach lets creators control 3D generators on their own 2D face data without needing resources to label large data or train large models.

Gereon Fox, Wentao Liu, Hyeongwoo Kim et al.

There are concerns that new approaches to the synthesis of high quality face videos may be misused to manipulate videos with malicious intent. The research community therefore developed methods for the detection of modified footage and assembled benchmark datasets for this task. In this paper, we examine how the performance of forgery detectors depends on the presence of artefacts that the human eye can see. We introduce a new benchmark dataset for face video forgery detection, of unprecedented quality. It allows us to demonstrate that existing detection techniques have difficulties detecting fakes that reliably fool the human eye. We thus introduce a new family of detectors that examine combinations of spatial and temporal features and outperform existing approaches both in terms of detection accuracy and generalization.

Lingjie Liu, Weipeng Xu, Marc Habermann et al.

Synthesizing realistic videos of humans using neural networks has been a popular alternative to the conventional graphics-based rendering pipeline due to its high efficiency. Existing works typically formulate this as an image-to-image translation problem in 2D screen space, which leads to artifacts such as over-smoothing, missing body parts, and temporal instability of fine-scale detail, such as pose-dependent wrinkles in the clothing. In this paper, we propose a novel human video synthesis method that approaches these limiting factors by explicitly disentangling the learning of time-coherent fine-scale details from the embedding of the human in 2D screen space. More specifically, our method relies on the combination of two convolutional neural networks (CNNs). Given the pose information, the first CNN predicts a dynamic texture map that contains time-coherent high-frequency details, and the second CNN conditions the generation of the final video on the temporally coherent output of the first CNN. We demonstrate several applications of our approach, such as human reenactment and novel view synthesis from monocular video, where we show significant improvement over the state of the art both qualitatively and quantitatively.

Hyeongwoo Kim, Mohamed Elgharib, Michael Zollhöfer et al.

Dubbing is a technique for translating video content from one language to another. However, state-of-the-art visual dubbing techniques directly copy facial expressions from source to target actors without considering identity-specific idiosyncrasies such as a unique type of smile. We present a style-preserving visual dubbing approach from single video inputs, which maintains the signature style of target actors when modifying facial expressions, including mouth motions, to match foreign languages. At the heart of our approach is the concept of motion style, in particular for facial expressions, i.e., the person-specific expression change that is yet another essential factor beyond visual accuracy in face editing applications. Our method is based on a recurrent generative adversarial network that captures the spatiotemporal co-activation of facial expressions, and enables generating and modifying the facial expressions of the target actor while preserving their style. We train our model with unsynchronized source and target videos in an unsupervised manner using cycle-consistency and mouth expression losses, and synthesize photorealistic video frames using a layered neural face renderer. Our approach generates temporally coherent results, and handles dynamic backgrounds. Our results show that our dubbing approach maintains the idiosyncratic style of the target actor better than previous approaches, even for widely differing source and target actors.

Mohamed Elgharib, Mallikarjun BR, Ayush Tewari et al.

Face performance capture and reenactment techniques use multiple cameras and sensors, positioned at a distance from the face or mounted on heavy wearable devices. This limits their applications in mobile and outdoor environments. We present EgoFace, a radically new lightweight setup for face performance capture and front-view videorealistic reenactment using a single egocentric RGB camera. Our lightweight setup allows operations in uncontrolled environments, and lends itself to telepresence applications such as video-conferencing from dynamic environments. The input image is projected into a low dimensional latent space of the facial expression parameters. Through careful adversarial training of the parameter-space synthetic rendering, a videorealistic animation is produced. Our problem is challenging as the human visual system is sensitive to the smallest face irregularities that could occur in the final results. This sensitivity is even stronger for video results. Our solution is trained in a pre-processing stage, through a supervised manner without manual annotations. EgoFace captures a wide variety of facial expressions, including mouth movements and asymmetrical expressions. It works under varying illuminations, background, movements, handles people from different ethnicities and can operate in real time.

Lingjie Liu, Weipeng Xu, Michael Zollhoefer et al.

We propose a method for generating video-realistic animations of real humans under user control. In contrast to conventional human character rendering, we do not require the availability of a production-quality photo-realistic 3D model of the human, but instead rely on a video sequence in conjunction with a (medium-quality) controllable 3D template model of the person. With that, our approach significantly reduces production cost compared to conventional rendering approaches based on production-quality 3D models, and can also be used to realistically edit existing videos. Technically, this is achieved by training a neural network that translates simple synthetic images of a human character into realistic imagery. For training our networks, we first track the 3D motion of the person in the video using the template model, and subsequently generate a synthetically rendered version of the video. These images are then used to train a conditional generative adversarial network that translates synthetic images of the 3D model into realistic imagery of the human. We evaluate our method for the reenactment of another person that is tracked in order to obtain the motion data, and show video results generated from artist-designed skeleton motion. Our results outperform the state-of-the-art in learning-based human image synthesis. Project page: http://gvv.mpi-inf.mpg.de/projects/wxu/HumanReenactment/

Hyeongwoo Kim, Pablo Garrido, Ayush Tewari et al.

We present a novel approach that enables photo-realistic re-animation of portrait videos using only an input video. In contrast to existing approaches that are restricted to manipulations of facial expressions only, we are the first to transfer the full 3D head position, head rotation, face expression, eye gaze, and eye blinking from a source actor to a portrait video of a target actor. The core of our approach is a generative neural network with a novel space-time architecture. The network takes as input synthetic renderings of a parametric face model, based on which it predicts photo-realistic video frames for a given target actor. The realism in this rendering-to-video transfer is achieved by careful adversarial training, and as a result, we can create modified target videos that mimic the behavior of the synthetically-created input. In order to enable source-to-target video re-animation, we render a synthetic target video with the reconstructed head animation parameters from a source video, and feed it into the trained network -- thus taking full control of the target. With the ability to freely recombine source and target parameters, we are able to demonstrate a large variety of video rewrite applications without explicitly modeling hair, body or background. For instance, we can reenact the full head using interactive user-controlled editing, and realize high-fidelity visual dubbing. To demonstrate the high quality of our output, we conduct an extensive series of experiments and evaluations, where for instance a user study shows that our video edits are hard to detect.

Ayush Tewari, Michael Zollhöfer, Pablo Garrido et al.

The reconstruction of dense 3D models of face geometry and appearance from a single image is highly challenging and ill-posed. To constrain the problem, many approaches rely on strong priors, such as parametric face models learned from limited 3D scan data. However, prior models restrict generalization of the true diversity in facial geometry, skin reflectance and illumination. To alleviate this problem, we present the first approach that jointly learns 1) a regressor for face shape, expression, reflectance and illumination on the basis of 2) a concurrently learned parametric face model. Our multi-level face model combines the advantage of 3D Morphable Models for regularization with the out-of-space generalization of a learned corrective space. We train end-to-end on in-the-wild images without dense annotations by fusing a convolutional encoder with a differentiable expert-designed renderer and a self-supervised training loss, both defined at multiple detail levels. Our approach compares favorably to the state-of-the-art in terms of reconstruction quality, better generalizes to real world faces, and runs at over 250 Hz.

Hyeongwoo Kim, Michael Zollhöfer, Ayush Tewari et al.

We introduce InverseFaceNet, a deep convolutional inverse rendering framework for faces that jointly estimates facial pose, shape, expression, reflectance and illumination from a single input image. By estimating all parameters from just a single image, advanced editing possibilities on a single face image, such as appearance editing and relighting, become feasible in real time. Most previous learning-based face reconstruction approaches do not jointly recover all dimensions, or are severely limited in terms of visual quality. In contrast, we propose to recover high-quality facial pose, shape, expression, reflectance and illumination using a deep neural network that is trained using a large, synthetically created training corpus. Our approach builds on a novel loss function that measures model-space similarity directly in parameter space and significantly improves reconstruction accuracy. We further propose a self-supervised bootstrapping process in the network training loop, which iteratively updates the synthetic training corpus to better reflect the distribution of real-world imagery. We demonstrate that this strategy outperforms completely synthetically trained networks. Finally, we show high-quality reconstructions and compare our approach to several state-of-the-art approaches.

Ayush Tewari, Michael Zollhöfer, Hyeongwoo Kim et al.

In this work we propose a novel model-based deep convolutional autoencoder that addresses the highly challenging problem of reconstructing a 3D human face from a single in-the-wild color image. To this end, we combine a convolutional encoder network with an expert-designed generative model that serves as decoder. The core innovation is our new differentiable parametric decoder that encapsulates image formation analytically based on a generative model. Our decoder takes as input a code vector with exactly defined semantic meaning that encodes detailed face pose, shape, expression, skin reflectance and scene illumination. Due to this new way of combining CNN-based with model-based face reconstruction, the CNN-based encoder learns to extract semantically meaningful parameters from a single monocular input image. For the first time, a CNN encoder and an expert-designed generative model can be trained end-to-end in an unsupervised manner, which renders training on very large (unlabeled) real world data feasible. The obtained reconstructions compare favorably to current state-of-the-art approaches in terms of quality and richness of representation.

Hyeongwoo Kim, Christian Richardt, Christian Theobalt

Many compelling video post-processing effects, in particular aesthetic focus editing and refocusing effects, are feasible if per-frame depth information is available. Existing computational methods to capture RGB and depth either purposefully modify the optics (coded aperture, light-field imaging), or employ active RGB-D cameras. Since these methods are less practical for users with normal cameras, we present an algorithm to capture all-in-focus RGB-D video of dynamic scenes with an unmodified commodity video camera. Our algorithm turns the often unwanted defocus blur into a valuable signal. The input to our method is a video in which the focus plane is continuously moving back and forth during capture, and thus defocus blur is provoked and strongly visible. This can be achieved by manually turning the focus ring of the lens during recording. The core algorithmic ingredient is a new video-based depth-from-defocus algorithm that computes space-time-coherent depth maps, deblurred all-in-focus video, and the focus distance for each frame. We extensively evaluate our approach, and show that it enables compelling video post-processing effects, such as different types of refocusing.

Christian Richardt, Hyeongwoo Kim, Levi Valgaerts et al.

We propose a new technique for computing dense scene flow from two handheld videos with wide camera baselines and different photometric properties due to different sensors or camera settings like exposure and white balance. Our technique innovates in two ways over existing methods: (1) it supports independently moving cameras, and (2) it computes dense scene flow for wide-baseline scenarios.We achieve this by combining state-of-the-art wide-baseline correspondence finding with a variational scene flow formulation. First, we compute dense, wide-baseline correspondences using DAISY descriptors for matching between cameras and over time. We then detect and replace occluded pixels in the correspondence fields using a novel edge-preserving Laplacian correspondence completion technique. We finally refine the computed correspondence fields in a variational scene flow formulation. We show dense scene flow results computed from challenging datasets with independently moving, handheld cameras of varying camera settings.

Tae-Hyun Oh, Yu-Wing Tai, Jean-Charles Bazin et al.

Robust Principal Component Analysis (RPCA) via rank minimization is a powerful tool for recovering underlying low-rank structure of clean data corrupted with sparse noise/outliers. In many low-level vision problems, not only it is known that the underlying structure of clean data is low-rank, but the exact rank of clean data is also known. Yet, when applying conventional rank minimization for those problems, the objective function is formulated in a way that does not fully utilize a priori target rank information about the problems. This observation motivates us to investigate whether there is a better alternative solution when using rank minimization. In this paper, instead of minimizing the nuclear norm, we propose to minimize the partial sum of singular values, which implicitly encourages the target rank constraint. Our experimental analyses show that, when the number of samples is deficient, our approach leads to a higher success rate than conventional rank minimization, while the solutions obtained by the two approaches are almost identical when the number of samples is more than sufficient. We apply our approach to various low-level vision problems, e.g. high dynamic range imaging, motion edge detection, photometric stereo, image alignment and recovery, and show that our results outperform those obtained by the conventional nuclear norm rank minimization method.