Driving a high-quality and photorealistic full-body virtual human from a few RGB cameras is a challenging problem that has become increasingly relevant with emerging virtual reality technologies. A promising solution to democratize such technology would be a generalizable method that takes sparse multi-view images of any person and then generates photoreal free-view renderings of them. However, the state-of-the-art approaches are not scalable to very large datasets and, thus, lack diversity and photorealism. To address this problem, we propose GIGA, a novel, generalizable full-body model for rendering photoreal humans in free viewpoint, driven by a single-view or sparse multi-view video. Notably, GIGA can scale training to a few thousand subjects while maintaining high photorealism and synthesizing dynamic appearance. At the core, we introduce a MultiHeadUNet architecture, which takes an approximate RGB texture accumulated from a single or multiple sparse views and predicts 3D Gaussian primitives represented as 2D texels on top of a human body mesh. At test time, our method performs novel view synthesis of a virtual 3D Gaussian-based human from 1 to 4 input views and a tracked body template for unseen identities. Our method excels over prior works by a significant margin in terms of identity generalization capability and photorealism.

We present MATCH (Multi-view Avatars from Topologically Corresponding Heads), a multi-view Gaussian registration method for high-quality head avatar creation and editing. State-of-the-art multi-view head avatar methods require time-consuming head tracking followed by expensive avatar optimization, often resulting in a total creation time of more than one day. MATCH, in contrast, directly predicts Gaussian splat textures in correspondence from calibrated multi-view images in just 0.5 seconds per frame, without requiring data preprocessing. The learned intra-subject correspondence across frames enables fast creation of personalized head avatars, while correspondence across subjects supports applications such as expression transfer, optimization-free tracking, semantic editing, and identity interpolation. We establish these correspondences end-to-end using a transformer-based model that predicts Gaussian splat textures in the fixed UV layout of a template mesh. To achieve this, we introduce a novel registration-guided attention block, where each UV-map token attends exclusively to image tokens depicting its corresponding mesh region. This design improves efficiency and performance compared to dense cross-view attention. MATCH outperforms existing methods in novel-view synthesis, geometry registration, and head avatar generation, while making avatar creation 10 times faster than the closest competing baseline. The code and model weights are available on the project website.

We present SynShot, a novel method for the few-shot inversion of a drivable head avatar based on a synthetic prior. We tackle three major challenges. First, training a controllable 3D generative network requires a large number of diverse sequences, for which pairs of images and high-quality tracked meshes are not always available. Second, the use of real data is strictly regulated (e.g., under the General Data Protection Regulation, which mandates frequent deletion of models and data to accommodate a situation when participant’s consent is withdrawn). Synthetic data, free from these constraints, is an appealing alternative. Third, state-of-the-art monocular avatar models struggle to generalize to new views and expressions, lacking a strong prior and often overfitting to a specific viewpoint distribution. Inspired by machine learning models trained solely on synthetic data, we propose a method that learns a prior model from a large dataset of synthetic heads with diverse identities, expressions, and viewpoints. With few input images, SynShot fine-tunes the pretrained synthetic prior to bridge the domain gap, modeling a photorealistic head avatar that generalizes to novel expressions and viewpoints. We model the head avatar using 3D Gaussian splatting and a convolutional encoder-decoder that outputs Gaussian parameters in UV texture space. To account for the different modeling complexities over parts of the head (e.g., skin vs hair), we embed the prior with explicit control for upsampling the number of per-part primitives. Compared to SOTA monocular and GAN-based methods, SynShot significantly improves novel view and expression synthesis.

Sparse volumetric reconstruction and rendering via 3D Gaussian splatting have recently enabled animatable 3D head avatars that are rendered under arbitrary viewpoints with impressive photorealism. Today, such photoreal avatars are seen as a key component in emerging applications in telepresence, extended reality, and entertainment. Building a photoreal avatar requires estimating the complex non-rigid motion of different facial components as seen in input video images; due to inaccurate motion estimation, animatable models typically present a loss of fidelity and detail when compared to their non-animatable counterparts, built from an individual facial expression. Also, recent state-of-the-art models are often affected by memory limitations that reduce the number of 3D Gaussians used for modeling, leading to lower detail and quality. To address these problems, we present a new high-detail 3D head avatar model that improves upon the state of the art, largely increasing the number of 3D Gaussians and modeling quality for rendering at 4K resolution. Our high-quality model is reconstructed from multiview input video and builds on top of a mesh-based 3D morphable model, which provides a coarse deformation layer for the head. Photoreal appearance is modelled by 3D Gaussians embedded within the continuous UVD tangent space of this mesh, allowing for more effective densification where most needed. Additionally, these Gaussians are warped by a novel UVD deformation field to capture subtle, localized motion. Our key contribution is the novel deformable Gaussian encoding and overall fitting procedure that allows our head model to preserve appearance detail, while capturing facial motion and other transient high-frequency features such as skin wrinkling.

Eye gaze and expressions are crucial non-verbal signals in face-to-face communication. Visual effects and telepresence demand significant improvements in personalized tracking, animation, and synthesis of the eye region to achieve true immersion. Morphable face models, in combination with coordinate-based neural volumetric representations, show promise in solving the difficult problem of reconstructing intricate geometry (eyelashes) and synthesizing photorealistic appearance variations (wrinkles and specularities) of eye performances. We propose a novel hybrid representation - ShellNeRF - that builds a discretized volume around a 3DMM face mesh using concentric surfaces to model the deformable 'periocular' region. We define a canonical space using the UV layout of the shells that constrains the space of dense correspondence search. Combined with an explicit eyeball mesh for modeling corneal light-transport, our model allows for animatable photorealistic 3D synthesis of the whole eye region. Using multi-view video input, we demonstrate significant improvements over state-of-the-art in expression re-enactment and transfer for high-resolution close-up views of the eye region.

Volumetric modeling and neural radiance field representations have revolutionized 3D face capture and photorealistic novel view synthesis. However, these methods often require hundreds of multi-view input images and are thus inapplicable to cases with less than a handful of inputs. We present a novel volumetric prior on human faces that allows for high-fidelity expressive face modeling from as few as three input views captured in the wild. Our key insight is that an implicit prior trained on synthetic data alone can generalize to extremely challenging real-world identities and expressions and render novel views with fine idiosyncratic details like wrinkles and eyelashes. We leverage a 3D Morphable Face Model to synthesize a large training set, rendering each identity with different expressions, hair, clothing, and other assets. We then train a conditional Neural Radiance Field prior on this synthetic dataset and, at inference time, fine-tune the model on a very sparse set of real images of a single subject. On average, the fine-tuning requires only three inputs to cross the synthetic-to-real domain gap. The resulting personalized 3D model reconstructs strong idiosyncratic facial expressions and outperforms the state-of-the-art in high-quality novel view synthesis of faces from sparse inputs in terms of perceptual and photo-metric quality.

In this paper, we examine 3 important issues in the practical use of state-of-the-art facial landmark detectors and show how a combination of specific architectural modifications can directly improve their accuracy and temporal stability. First, many facial landmark detectors require face normalization as a preprocessing step, which is accomplished by a separately-trained neural network that crops and resizes the face in the input image. There is no guarantee that this pre-trained network performs the optimal face normalization for landmark detection. We instead analyze the use of a spatial transformer network that is trained alongside the landmark detector in an unsupervised manner, and jointly learn optimal face normalization and landmark detection. Second, we show that modifying the output head of the landmark predictor to infer landmarks in a canonical 3D space can further improve accuracy. To convert the predicted 3D landmarks into screen-space, we additionally predict the camera intrinsics and head pose from the input image. As a side benefit, this allows to predict the 3D face shape from a given image only using 2D landmarks as supervision, which is useful in determining landmark visibility among other things. Finally, when training a landmark detector on multiple datasets at the same time, annotation inconsistencies across datasets forces the network to produce a suboptimal average. We propose to add a semantic correction network to address this issue. This additional lightweight neural network is trained alongside the landmark detector, without requiring any additional supervision. While the insights of this paper can be applied to most common landmark detectors, we specifically target a recently-proposed continuous 2D landmark detector to demonstrate how each of our additions leads to meaningful improvements over the state-of-the-art on standard benchmarks.

We introduce a novel framework for 3D human avatar generation and personalization, leveraging text prompts to enhance user engagement and customization. Central to our approach are key innovations aimed at overcoming the challenges in photo-realistic avatar synthesis. Firstly, we utilize a conditional Neural Radiance Fields (NeRF) model, trained on a large-scale unannotated multi-view dataset, to create a versatile initial solution space that accelerates and diversifies avatar generation. Secondly, we develop a geometric prior, leveraging the capabilities of Text-to-Image Diffusion Models, to ensure superior view invariance and enable direct optimization of avatar geometry. These foundational ideas are complemented by our optimization pipeline built on Variational Score Distillation (VSD), which mitigates texture loss and over-saturation issues. As supported by our extensive experiments, these strategies collectively enable the creation of custom avatars with unparalleled visual quality and better adherence to input text prompts. You can find more results and videos in our website: syntec-research.github.io/MagicMirror .

Monocular 3D face reconstruction is a wide-spread topic, and existing approaches tackle the problem either through fast neural network inference or offline iterative reconstruction of face geometry. In either case carefully-designed energy functions are minimized, commonly including loss terms like a photometric loss, a landmark reprojection loss, and others. In this work we propose a new loss function for monocular face capture, inspired by how humans would perceive the quality of a 3D face reconstruction given a particular image. It is widely known that shading provides a strong indicator for 3D shape in the human visual system. As such, our new 'perceptual' shape loss aims to judge the quality of a 3D face estimate using only shading cues. Our loss is implemented as a discriminator-style neural network that takes an input face image and a shaded render of the geometry estimate, and then predicts a score that perceptually evaluates how well the shaded render matches the given image. This 'critic' network operates on the RGB image and geometry render alone, without requiring an estimate of the albedo or illumination in the scene. Furthermore, our loss operates entirely in image space and is thus agnostic to mesh topology. We show how our new perceptual shape loss can be combined with traditional energy terms for monocular 3D face optimization and deep neural network regression, improving upon current state-of-the-art results.

High-fidelity, photorealistic 3D capture of a human face is a long-standing problem in computer graphics – the complex material of skin, intricate geometry of hair, and fine scale textural details make it challenging. Traditional techniques rely on very large and expensive capture rigs to reconstruct explicit mesh geometry and appearance maps, and are limited by the accuracy of hand-crafted reflectance models. More recent volumetric methods (e.g., NeRFs) have enabled view-synthesis and sometimes relighting by learning an implicit representation of the density and reflectance basis, but suffer from artifacts and blurriness due to the inherent ambiguities in volumetric modeling. These problems are further exacerbated when capturing with few cameras and light sources. We present a novel technique for high-quality capture of a human face for 3D view synthesis and relighting using a sparse, compact capture rig consisting of 15 cameras and 15 lights. Our method combines a neural volumetric representation with traditional mesh reconstruction from multiview stereo. The proxy geometry allows us to anchor the 3D density field to prevent artifacts and guide the disentanglement of intrinsic radiance components of the face appearance such as diffuse and specular reflectance, and incident radiance (shadowing) fields. Our hybrid representation significantly improves the state-of-the-art quality for arbitrarily dense renders of a face from desired camera viewpoint as well as environmental, directional, and near-field lighting.

Neural networks for facial landmark detection are notoriously limited to a fixed set of landmarks in a dedicated layout, which must be specified at training time. Dedicated datasets must also be hand-annotated with the corresponding landmark configuration for training. We propose the first facial landmark detection network that can predict continuous, unlimited landmarks, allowing to specify the number and location of the desired landmarks at inference time. Our method combines a simple image feature extractor with a queried landmark predictor, and the user can specify any continuous query points relative to a 3D template face mesh as input. As it is not tied to a fixed set of landmarks, our method is able to leverage all pre-existing 2D landmark datasets for training, even if they have inconsistent landmark configurations. As a result, we present a very powerful facial landmark detector that can be trained once, and can be used readily for numerous applications like 3D face reconstruction, arbitrary face segmentation, and is even compatible with helmeted mounted cameras, and therefore could vastly simplify face tracking workflows for media and entertainment applications.

3D facial animation is often produced by manipulating facial deformation models (or rigs), that are traditionally parameterized by expression controls. A key component that is usually overlooked is expression 'style', as in, how a particular expression is performed. Although it is common to define a semantic basis of expressions that characters can perform, most characters perform each expression in their own style. To date, style is usually entangled with the expression, and it is not possible to transfer the style of one character to another when considering facial animation. We present a new face model, based on a data-driven implicit neural physics model, that can be driven by both expression and style separately. At the core, we present a framework for learning implicit physics-based actuations for multiple subjects simultaneously, trained on a few arbitrary performance capture sequences from a small set of identities. Once trained, our method allows generalized physics-based facial animation for any of the trained identities, extending to unseen performances. Furthermore, it grants control over the animation style, enabling style transfer from one character to another or blending styles of different characters. Lastly, as a physics-based model, it is capable of synthesizing physical effects, such as collision handling, setting our method apart from conventional approaches.

Recent work on radiance fields and volumetric inverse rendering (e.g., NeRFs) has provided excellent results in building data-driven models of real scenes for novel view synthesis with high photorealism. While full control over viewpoint is achieved, scene lighting is typically "baked" into the model and cannot be changed; other methods only capture limited variation in lighting or make restrictive assumptions about the captured scene. These limitations prevent the application on arbitrary materials and novel 3D environments with complex, distinct lighting. In this paper, we target the application scenario of capturing high-fidelity assets for neural relighting in controlled studio conditions, but without requiring a dense light stage. Instead, we leverage a small number of area lights commonly used in photogrammetry. We propose ReNeRF, a relightable radiance field model based on the intuitive and powerful approach of image-based relighting, which implicitly captures global light transport (for arbitrary objects) without complex, error-prone simulations. Thus, our new method is simple and provides full control over viewpoint and lighting, without simplistic assumptions about how light interacts with the scene. In addition, ReNeRF does not rely on the usual assumption of distant lighting - during training, we explicitly account for the distance between 3D points in the volume and point samples on the light sources. Thus, at test time, we achieve better generalization to novel, continuous lighting directions, including nearfield lighting effects.

Presurgical orthopedic plates are widely used for the treatment of cleft lip and palate, which is the most common craniofacial birth defect. For the traditional plate fabrication, an impression is taken under airway-endangering conditions, which recent digital alternatives overcome via intraoral scanners. However, these alternatives demand proficiency in 3D modeling software in addition to the generally required clinical knowledge of plate design. We address these limitations with a data-driven and fully automated digital pipeline, endowed with a graphical user interface. The pipeline adopts a deep learning model to landmark raw intraoral scans of arbitrary mesh topology and orientation, which guides the nonrigid surface registration subsequently employed to segment the scans. The plates that are individually fit to these segmented scans are 3D-printable and offer optional customization. With the distance to the alveolar ridges closely centered around the targeted 0.1 mm, our pipeline computes tightly fitting plates in less than 3 min. The plates were approved in 12 out of 12 cases by two cleft care professionals in a printed-model-based evaluation. Moreover, since the pipeline was implemented in clinical routine in two hospitals, 19 patients have been undergoing treatment utilizing our automated designs. The results demonstrate that our automated pipeline meets the high precision requirements of the medical setting employed in cleft lip and palate care while substantially reducing the design time and required clinical expertise, which could facilitate access to this presurgical treatment, especially in low-income countries.

We present a novel graph-based simulation approach for generating micro wrinkle geometry on human skin, which can easily scale up to the micro-meter range and millions of wrinkles. The simulation first samples pores on the skin and treats them as nodes in a graph. These nodes are then connected and the resulting edges become candidate wrinkles. An iterative optimization inspired by pedestrian trail formation is then used to assign weights to those edges, i.e., to carve out the wrinkles. Finally, we convert the graph to a detailed skin displacement map using novel shape functions implemented in graphics shaders. Our simulation and displacement map creation steps expose fine controls over the appearance at real-time framerates suitable for interactive exploration and design. We demonstrate the effectiveness of the generated wrinkles by enhancing state-of-art 3D reconstructions of real human subjects with simulated micro wrinkles, and furthermore propose an artist-driven design flow for adding micro wrinkles to fictional characters.

Recent research work has developed powerful generative models (e.g., StyleGAN2) that can synthesize complete human head images with impressive photorealism, enabling applications such as photorealistically editing real photographs. While these models can be trained on large collections of unposed images, their lack of explicit 3D knowledge makes it difficult to achieve even basic control over 3D viewpoint without unintentionally altering identity. On the other hand, recent Neural Radiance Field (NeRF) methods have already achieved multiview-consistent, photorealistic renderings but they are so far limited to a single facial identity. In this paper, we propose a new Morphable Radiance Field (MoRF) method that extends a NeRF into a generative neural model that can realistically synthesize multiview-consistent images of complete human heads, with variable and controllable identity. MoRF allows for morphing between particular identities, synthesizing arbitrary new identities, or quickly generating a NeRF from few images of a new subject, all while providing realistic and consistent rendering under novel viewpoints. We train MoRF in a supervised fashion by leveraging a high-quality database of multiview portrait images of several people, captured in studio with polarization-based separation of diffuse and specular reflection. Here, we demonstrate how MoRF is a strong new step forwards towards generative NeRFs for 3D neural head modeling.

Face swapping is the process of applying a source actor's appearance to a target actor's performance in a video. This is a challenging visual effect that has seen increasing demand in film and television production. Recent work has shown that data-driven methods based on deep learning can produce compelling effects at production quality in a fraction of the time required for a traditional 3D pipeline. However, the dominant approach operates only on 2D imagery without reference to the underlying facial geometry or texture, resulting in poor generalization under novel viewpoints and little artistic control. Methods that do incorporate geometry rely on pre-learned facial priors that do not adapt well to particular geometric features of the source and target faces. We approach the problem of face swapping from the perspective of learning simultaneous convolutional facial autoencoders for the source and target identities, using a shared encoder network with identity-specific decoders. The key novelty in our approach is that each decoder first lifts the latent code into a 3D representation, comprising a dynamic face texture and a deformable 3D face shape, before projecting this 3D face back onto the input image using a differentiable renderer. The coupled autoencoders are trained only on videos of the source and target identities, without requiring 3D supervision. By leveraging the learned 3D geometry and texture, our method achieves face swapping with higher quality than when using off-the-shelf monocular 3D face reconstruction, and overall lower FID score than state-of-the-art 2D methods. Furthermore, our 3D representation allows for efficient artistic control over the result, which can be hard to achieve with existing 2D approaches.

Facial hair is a largely overlooked topic in facial performance capture. Most production pipelines in the entertainment industry do not have a way to automatically capture facial hair or track the skin underneath it. Thus, actors are asked to shave clean before face capture, which is very often undesirable. Capturing the geometry of individual facial hairs is very challenging, and their presence makes it harder to capture the deforming shape of the underlying skin surface. Some attempts have already been made at automating this task, but only for static faces with relatively sparse 3D hair reconstructions. In particular, current methods lack the temporal correspondence needed when capturing a sequence of video frames depicting facial performance. The problem of robustly tracking the skin underneath also remains unaddressed. In this paper, we propose the first multiview reconstruction pipeline that tracks both the dense 3D facial hair, as well as the underlying 3D skin for entire performances. Our method operates with standard setups for face photogrammetry, without requiring dense camera arrays. For a given capture subject, our algorithm first reconstructs a dense, high-quality neutral 3D facial hairstyle by registering sparser hair reconstructions over multiple frames that depict a neutral face under quasi-rigid motion. This custom-built, reference facial hairstyle is then tracked throughout a variety of changing facial expressions in a captured performance, and the result is used to constrain the tracking of the 3D skin surface underneath. We demonstrate the proposed capture pipeline on a variety of different facial hairstyles and lengths, ranging from sparse and short to dense full-beards.

Facial appearance capture techniques estimate geometry and reflectance properties of facial skin by performing a computationally intensive inverse rendering optimization in which one or more images are re-rendered a large number of times and compared to real images coming from multiple cameras. Due to the high computational burden, these techniques often make several simplifying assumptions to tame complexity and make the problem more tractable. For example, it is common to assume that the scene consists of only distant light sources, and ignore indirect bounces of light (on the surface and within the surface). Also, methods based on polarized lighting often simplify the light interaction with the surface and assume perfect separation of diffuse and specular reflectance. In this paper, we move in the opposite direction and demonstrate the impact on facial appearance capture quality when departing from these idealized conditions towards models that seek to more accurately represent the lighting, while at the same time minimally increasing computational burden. We compare the results obtained with a state-of-the-art appearance capture method [RGB ∗ 20], with and without our proposed improvements to the lighting model.

Photorealistic digital re-aging of faces in video is becoming increasingly common in entertainment and advertising. But the predominant 2D painting workflow often requires frame-by-frame manual work that can take days to accomplish, even by skilled artists. Although research on facial image re-aging has attempted to automate and solve this problem, current techniques are of little practical use as they typically suffer from facial identity loss, poor resolution, and unstable results across subsequent video frames. In this paper, we present the first practical, fully-automatic and production-ready method for re-aging faces in video images. Our first key insight is in addressing the problem of collecting longitudinal training data for learning to re-age faces over extended periods of time, a task that is nearly impossible to accomplish for a large number of real people. We show how such a longitudinal dataset can be constructed by leveraging the current state-of-the-art in facial re-aging that, although failing on real images, does provide photoreal re-aging results on synthetic faces. Our second key insight is then to leverage such synthetic data and formulate facial re-aging as a practical image-to-image translation task that can be performed by training a well-understood U-Net architecture, without the need for more complex network designs. We demonstrate how the simple U-Net, surprisingly, allows us to advance the state of the art for re-aging real faces on video, with unprecedented temporal stability and preservation of facial identity across variable expressions, viewpoints, and lighting conditions. Finally, our new face re-aging network (FRAN) incorporates simple and intuitive mechanisms that provides artists with localized control and creative freedom to direct and fine-tune the re-aging effect, a feature that is largely important in real production pipelines and often overlooked in related research work.