Skip to content

3D Face / Head

Characters & Avatars

Open Source

  • AF

    Any3DAvatar: Fast and High-Quality Full-Head 3D Avatar Reconstruction from Single Portrait Image Discovered by AI

    2026

    A fast feed-forward method for single-image 3D Gaussian head avatar generation. It reconstructs a complete 3D head with fine geometry and appearance details from a single portrait in under one second, bridging the gap between speed and quality.

    3d-head gaussian-splatting real-time single-image-to-3d

    Project · GitHub · arXiv

  • E3

    EmoteGPT: 3D Human Facial Expressions from Natural Language Descriptions Discovered by AI

    2026

    A text-driven 3D facial expression generation framework that maps natural language descriptions (both explicit physical details and implicit situational contexts) to 3D Morphable Model (3DMM) parameters using a Multimodal Large Language Model.

    3d-face 3dmm expression text-to-3d

    arXiv · Website

  • FG

    FastGHA: Generalized Few-Shot 3D Gaussian Head Avatars with Real-Time Animation Discovered by AI

    2026

    A feed-forward method that generates high-quality 3D Gaussian head avatars from only a few input images (few-shot) while supporting real-time animation. It uses a novel architecture to generalize to unseen subjects without per-identity optimization.

    3d-head 3dgs avatar few-shot

    arXiv

  • GA

    Generalizable and Animatable 3D Full-Head Gaussian Avatar from a Single Image Discovered by AI

    2026

    A feed-forward framework for one-shot 3D full-head animatable avatar reconstruction from a single image. It enables real-time animation and simultaneous 360-degree rendering while preserving identity and expression details.

    3dgs head-avatar one-shot

    Project · GitHub · arXiv

  • GH

    GNM Head: A Generative aNthropometric Model of the human head Discovered by AI

    2026

    GNM Head is Google's open-source, highly detailed 3D parametric model of the human head, face, neck, eyeballs, teeth, and tongue. Trained on high-resolution 3D scans, it provides over 250 identity controls and 380 expression controls, serving as a robust conditioning signal for generative AI and 3D character creation.

    3d-head face-reconstruction google parametric-model

    GitHub · arXiv

  • HF

    High-Fidelity Single-Image Head Modeling with Industry-Grade Topology Discovered by AI

    2026

    This paper presents a single-image head mesh reconstruction framework that addresses the challenge of simultaneously preserving facial identity and producing industry-grade topology. It adopts a coarse-to-fine hierarchical optimization pipeline across rig, joint, and vertex stages to yield meshes with semantically meaningful edge flow suitable for downstream animation workflows.

    3d-head mesh-reconstruction topology

    arXiv

  • KA

    KaoLRM Discovered by AI

    2026

    A parametric 3D face reconstruction framework that repurposes pre-trained Large Reconstruction Models (LRM) and incorporates FLAME-based 2D Gaussian Splatting to achieve high viewpoint consistency.

    3d-face gaussian-splatting lrm reconstruction

    GitHub · arXiv · Paper

  • MF

    MATCH: Feed-forward Gaussian Registration for Head Avatar Creation and Editing Discovered by AI

    2026

    A multi-view Gaussian registration method that directly predicts topologically corresponding 3D Gaussian splat textures from calibrated multi-view images in 0.5 seconds, enabling rapid head avatar creation and editing.

    gaussian-splatting head-avatar multi-view reconstruction

    Project · GitHub · arXiv · Paper

  • RR

    RealDenseFace: Real-time Monocular 3D Face Reconstruction from Dense UV-space Priors Discovered by AI

    2026

    RealDenseFace is a real-time optimization-based monocular 3D face reconstruction method that fits a 3D Morphable Model (3DMM) to dense UV-space network predictions. It formulates 3DMM fitting as a nonlinear least-squares problem and solves it with a tailored Gauss-Newton solver, achieving high-fidelity reconstruction at real-time speeds.

    3dmm face-reconstruction real-time

    GitHub · arXiv

  • SK

    Skullptor Discovered by AI

    2026

    A high-fidelity 3D head reconstruction framework developed by Ubisoft LaForge that utilizes multi-view normal prediction to reconstruct detailed head geometry in seconds from a small number of views.

    3d-head normal-prediction reconstruction ubisoft

    GitHub · arXiv · Paper

  • SP

    Splat-Portrait: Generalizing Talking Heads with Gaussian Splatting Discovered by AI

    2026

    A 3D talking head generation framework based on 3D Gaussian Splatting that generalizes to unseen identities from a single portrait image, producing high-fidelity 3D head reconstructions and natural animations.

    3dgs portrait-animation talking-head

    arXiv

  • TH

    TOPOS: High-Fidelity and Efficient Industry-Grade 3D Head Generation Discovered by AI

    2026

    A framework for single-image conditioned 3D head generation that recovers geometry and appearance under an industry-standard topology (such as MetaHuman), making it directly compatible with production-level rigging and animation pipelines.

    3d-head mesh metahuman

    arXiv

  • UF

    UVFaceFusion: Fast Multi-view Topologically Consistent Face Reconstruction in the Wild via UV-space Neural Fusion Discovered by AI

    2026

    A feed-forward framework for fast, topologically consistent multi-view face reconstruction that replaces heuristic optimization with learnable neural fusion in a canonical UV space.

    face-reconstruction multi-view neural-fusion uv-space

    GitHub · arXiv · Paper

  • 3G

    3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations Discovered by AI

    2025

    A novel method for generating high-quality 3D Gaussian head avatars with expressive dynamic appearances. It introduces a compact tensorial representation combining static triplanes and a lightweight 1D feature line to encode dynamic textures, achieving high-fidelity rendering and fast inference.

    3d-head 3dmm avatar gaussian-splatting

    Project · GitHub · arXiv

  • 33

    3DGH: 3D Head Generation with Composable Hair and Face Discovered by AI

    2025

    An unconditional generative model for 3D human heads that separates hair and face components using a template-based 3D Gaussian Splatting representation.

    3d-head gaussian-splatting hair-generation

    GitHub · arXiv

  • AG

    Arc2Avatar: Generating Expressive 3D Avatars from a Single Image via ID Guidance Discovered by AI

    2025

    A score distillation sampling (SDS) based method that generates a complete, detailed 3D head avatar from a single image. It integrates with 3D Morphable Models (3DMM) and a face foundation model to enable blendshape-based expression control and natural color fidelity.

    3d-head 3dgs 3dmm avatar

    Project · GitHub · arXiv

  • AR

    ARTalk Discovered by AI

    2025

    An autoregressive model for speech-driven 3D head animation that generates highly synchronized lip movements, realistic head poses, and eye blinks from arbitrary audio in real-time.

    3d-head facial-animation real-time speech-driven

    GitHub · arXiv · Paper

  • AD

    Audio-Driven Universal Gaussian Head Avatars Discovered by AI

    2025

    An audio-driven universal photorealistic avatar synthesis method combining a person-agnostic speech model with a Universal Head Avatar Prior (UHAP) trained on cross-identity multi-view videos.

    audio-driven gaussian-splatting talking-head

    GitHub · arXiv

  • FI

    FastAvatar: Instant 3D Gaussian Splatting for Faces from Single Unconstrained Poses Discovered by AI

    2025

    A feed-forward framework that generates high-quality 3D Gaussian Splatting (3DGS) face models from a single unconstrained face image in under 10ms. It enables real-time novel view synthesis and expression manipulation without test-time optimization.

    3d-face 3dgs real-time reconstruction

    Project · GitHub · arXiv

  • FT

    FastAvatar: Towards Unified and Fast 3D Avatar Reconstruction with Large Gaussian Reconstruction Transformers Discovered by AI

    2025

    A unified feed-forward 3D avatar reconstruction framework that can process diverse inputs (including a single image, sparse multi-view observations, or monocular video) to reconstruct a high-quality 3D Gaussian Splatting (3DGS) head avatar within seconds.

    3d-head 3dgs avatar transformer

    Project · GitHub · arXiv

  • LL

    LAM: Large Avatar Model for One-shot Animatable Gaussian Head Discovered by AI

    2025

    A large avatar model that generates animatable and high-fidelity 3D Gaussian head avatars from a single image in a single forward pass, enabling real-time reenactment and rendering.

    3d-head avatar gaussian-splatting

    GitHub · arXiv

  • PE

    PercHead Discovered by AI

    2025

    A perceptual head model for single-image 3D head reconstruction and editing that leverages a novel perceptual loss based on DINOv2 and SAM 2.1 to achieve deep visual understanding and generalized supervision.

    3d-head editing perceptual-loss reconstruction

    GitHub · arXiv · Paper

  • PI

    Pixel3DMM Discovered by AI

    2025

    A set of highly-generalized vision transformers that predict per-pixel geometric cues (surface normals and UV coordinates) to constrain the optimization of a 3D morphable face model (3DMM) from a single RGB image.

    3d-face 3dmm flame reconstruction

    GitHub · arXiv · Paper

  • RR

    RGBAvatar: Reduced Gaussian Blendshapes for Online Modeling of Head Avatars Discovered by AI

    2025

    A method for reconstructing photorealistic, animatable head avatars on the fly by mapping tracked 3DMM parameters into reduced blendshape weights with an MLP, creating a compact set of 3D Gaussian blendshape bases.

    blendshapes gaussian-splatting head-avatar real-time

    GitHub · arXiv · Paper

  • SS

    StyleMM: Stylized 3D Morphable Face Model via Text-Driven Aligned Image Translation Discovered by AI

    2025

    A novel framework that constructs a stylized 3D Morphable Model (3DMM) based on user-defined text descriptions specifying a target style, enabling high-quality stylized 3D face generation and animation with consistent geometry and texture.

    3dmm face stylization text-to-3d

    Project · GitHub · arXiv

  • VT

    VGGTFace: Topologically Consistent Facial Geometry Reconstruction in the Wild Discovered by AI

    2025

    An automatic approach that leverages a 3D foundation model (VGGT) to reconstruct topologically consistent 3D facial geometry from in-the-wild multi-view images captured by everyday users in less than 10 seconds.

    face-reconstruction mesh multi-view topology

    GitHub · arXiv · Paper

  • 3G

    3D Gaussian Blendshapes for Head Avatar Animation NEW Discovered by AI

    2024

    A framework that learns 3D Gaussian blendshapes from monocular video to capture high-frequency facial expressions and render photorealistic head avatars in real time.

    3d-head avatar-animation blendshapes gaussian-splatting

    GitHub · arXiv

  • FA

    FaceLift Discovered by AI

    2024

    A feed-forward approach for generalizable 360-degree 3D head reconstruction from a single image, using a multi-view latent diffusion model and a transformer-based reconstructor to output 3D Gaussian splats.

    3d-head gaussian-splatting reconstruction

    GitHub · arXiv

  • GG

    GAGAvatar: Generalizable and Animatable Gaussian Head Avatar Discovered by AI

    2024

    A generalizable and animatable 3D Gaussian head avatar framework that generates high-fidelity 3D Gaussians from a single image in a single forward pass, capturing identity-preserving details.

    3d-head avatar gaussian-splatting

    GitHub · arXiv

  • GD

    Gaussian Déjà-vu: Creating Controllable 3D Gaussian Head-Avatars with Enhanced Generalization and Personalization Abilities Discovered by AI

    2024

    A framework that accelerates controllable 3D Gaussian head avatar creation by first training a generalized model on large 2D datasets and then personalizing it with few-shot target images, improving generalization and expression control.

    3d-head avatar few-shot gaussian-splatting

    Project · GitHub · arXiv

  • GR

    GaussianTalker: Real-Time High-Fidelity Talking Head Synthesis with Audio-Driven 3D Gaussian Splatting Discovered by AI

    2024

    A real-time, pose-controllable talking head synthesis framework that deforms a canonical 3D Gaussian Splatting representation in sync with speech audio by encoding Gaussian attributes into a shared implicit feature space.

    3dgs real-time talking-head

    arXiv

  • GF

    GGHead: Fast and Generalizable 3D Gaussian Heads Discovered by AI

    2024

    An unconditional 3D generative model for human heads based on 3D Gaussian Splatting. It generates high-quality, view-consistent 3D heads at high resolution in real-time without requiring 2D super-resolution networks.

    3dgs head-generation real-time

    Project · GitHub · arXiv

  • GG

    GPHM: Gaussian Parametric Head Model for Monocular Head Avatar Reconstruction Discovered by AI

    2024

    A parametric head model based on 3D Gaussian Splatting designed for monocular head avatar reconstruction. It enables rapid, high-quality, and photorealistic 3D head reconstruction using only a small amount of monocular video data.

    3dgs head-reconstruction parametric-model

    Project · GitHub · arXiv

  • HF

    HeadGAP: Few-shot 3D Head Avatar via Generalizable Gaussian Priors Discovered by AI

    2024

    A few-shot 3D head avatar creation method that learns generalizable 3D Gaussian head priors from large-scale data, enabling high-quality personalized head avatar reconstruction and real-time animation.

    3d-head avatar-generation few-shot gaussian-splatting

    Project · arXiv

  • HT

    HeadStudio: Text to Animatable Head Avatars with 3D Gaussian Splatting NEW Discovered by AI

    2024

    HeadStudio generates text-driven animatable 3D head avatars using 3D Gaussian Splatting combined with FLAME parametric expression priors for real-time rendering.

    3d-head flame gaussian-splatting text-to-3d

    GitHub · arXiv

  • IS

    ID-Sculpt Discovered by AI

    2024

    An optimization-based framework for high-fidelity 3D head generation from a single portrait image that preserves identity using ID-aware score distillation and landmark-guided optimization.

    3d-head identity-preserving reconstruction

    Project · GitHub · arXiv

  • ME

    MeGA Discovered by AI

    2024

    A Hybrid Mesh-Gaussian Head Avatar framework that models different head components (such as skin and hair) with their most suitable representations. It combines a deformable mesh for the face with 3D Gaussian Splatting for the hair to achieve high-fidelity rendering and head editing.

    gaussian-splatting head-avatar mesh

    GitHub · arXiv

  • RP

    Real3D-Portrait Discovered by AI

    2024

    An audio-driven 3D talking portrait synthesis framework that reconstructs a 3D avatar from a single image and animates it with realistic head motion and facial expressions.

    audio-driven avatar talking-head

    Project · GitHub · arXiv

  • SI

    Single Image, Any Face: Generalisable 3D Face Generation Discovered by AI

    2024

    A generalizable 3D face generation framework (Gen3D-Face) designed to reconstruct photorealistic 3D human face avatars from a single unconstrained, out-of-domain portrait image.

    3d-face avatar-generation reconstruction single-view

    arXiv

  • S3

    SMIRK: 3D Facial Expressions through Analysis-by-Neural-Synthesis Discovered by AI

    2024

    A framework for highly expressive 3D face reconstruction from in-the-wild images. It utilizes an analysis-by-neural-synthesis approach to faithfully capture subtle, extreme, asymmetric, or rarely observed facial expressions.

    3d-face expression flame reconstruction

    GitHub · arXiv

  • SP

    SphereHead Discovered by AI

    2024

    A 3D-aware GAN framework for full-head synthesis that introduces a spherical tri-plane representation. By aligning the tri-plane projection planes with spherical coordinates, it matches the geometric characteristics of the human head, reducing artifacts and improving rendering quality from all viewing angles.

    3d-generative-model 3d-head-synthesis gan tri-plane

    Project · GitHub · arXiv

  • CO

    CVTHead: One-shot Controllable Head Avatar with Vertex-feature Transformer Discovered by AI

    2023

    CVTHead is a one-shot controllable head avatar generation method that utilizes a Vertex-feature Transformer to achieve explicit face control of 3D Morphable Models (3DMM). It generates controllable neural head avatars from a single image and enables real-time animation with high-fidelity rendering.

    3dmm head-avatar transformer

    arXiv

  • DD

    DiffusionAvatars: Deferred Diffusion for High-fidelity 3D Head Avatars Discovered by AI

    2023

    A diffusion-based neural rendering framework for high-fidelity 3D head avatars that offers intuitive control over both pose and expression by leveraging 2D diffusion priors and a neural parametric head model.

    3d-head avatar diffusion expression

    GitHub · arXiv

  • EE

    EMOTE: Emotional Speech-Driven Animation with Content-Emotion Disentanglement Discovered by AI

    2023

    An emotional speech-driven 3D facial animation model that disentangles speech content from emotional expression, enabling explicit control over the expressed emotion while maintaining accurate lip synchronization.

    emotion-control facial-animation talking-head

    GitHub · arXiv

  • FA

    FaceG2E Discovered by AI

    2023

    A unified text-guided framework for 3D face synthesis and editing. It decouples geometry and texture generation to preserve fine-grained geometric details, and introduces a self-guided consistency preservation method to enable precise, text-controlled local editing of the generated 3D faces.

    3d-face-editing 3d-face-generation diffusion text-to-3d

    GitHub · arXiv

  • FH

    FlashAvatar: High-fidelity Head Avatar with Efficient Gaussian Embedding Discovered by AI

    2023

    A lightweight 3D animatable head avatar representation that reconstructs a digital avatar from a short monocular video sequence in minutes and renders high-fidelity photo-realistic images at 300FPS on consumer GPUs using 3D Gaussian Splatting.

    3d-head avatar gaussian-splatting real-time

    GitHub · arXiv

  • GH

    Gaussian Head Avatar Discovered by AI

    2023

    A high-fidelity 3D head avatar representation that utilizes controllable dynamic 3D Gaussians to model expressive human heads. It introduces a neural deformation field to map canonical Gaussians to the deformed space based on expression and pose, enabling ultra-high-fidelity rendering at 2K resolution.

    3d-gaussian-splatting deformation-field face-reconstruction head-avatar

    Project · GitHub · arXiv

  • GA

    GaussianHead Discovered by AI

    2023

    A method for constructing high-fidelity 3D head avatars using anisotropic 3D Gaussians. It introduces a learnable Gaussian derivation mechanism that dynamically adjusts Gaussian attributes based on head pose and expression, capturing fine-grained dynamic details like wrinkles and hair movement.

    3d-gaussian-splatting face-reconstruction head-avatar tvcg

    Project · GitHub · arXiv

  • HT

    HeadArtist: Text-conditioned 3D Head Generation with Self Score Distillation Discovered by AI

    2023

    A text-conditioned 3D head generation framework that utilizes a landmark-guided ControlNet as a generative prior and optimizes a parameterized 3D head model via self score distillation (SSD).

    3d-head controlnet score-distillation text-to-3d

    GitHub · arXiv

  • HR

    HeadGaS: Real-Time Animatable Head Avatars via 3D Gaussian Splatting NEW Discovered by AI

    2023

    HeadGaS represents 3D animatable head avatars using 3D Gaussian Splatting driven by explicit expression features for real-time, high-fidelity facial animation.

    3d-head avatar gaussian-splatting real-time

    arXiv

  • TR

    Towards Realistic Generative 3D Face Models Discovered by AI

    2023

    A 3D generative face model that combines 2D face generative models with semantic face manipulation, enabling the generation and text-guided editing of detailed 3D rendered faces with high-quality mesh and texture.

    3d-face face-editing mesh-generation texture-generation

    Project · GitHub · arXiv

  • EE

    EMOCA: Emotion Driven Monocular Face Capture and Animation Discovered by AI

    2022

    A monocular 3D face reconstruction method that focuses on capturing a wide spectrum of facial expressions and emotional states from in-the-wild images using a novel emotional consistency loss.

    3dmm emotion-capture face-reconstruction

    GitHub · arXiv

  • SP

    SPECTRE Discovered by AI

    2022

    A visual speech-aware perceptual 3D facial expression reconstruction framework from monocular videos. It ensures that speech perception and mouth movements are accurately preserved in the reconstructed 3D talking head.

    face-reconstruction facial-expression speech-aware

    GitHub · arXiv

  • EG

    EG3D

    2021

    A geometry-aware 3D GAN using a hybrid explicit-implicit tri-plane representation to synthesize multi-view-consistent face images and 3D geometry from single-view photo collections.

    face-head gan neural-rendering

    Project · GitHub · arXiv

  • DE

    DECA Discovered by AI

    2020

    A monocular 3D face reconstruction framework that regresses 3D face shape and animatable details (such as person-specific wrinkles) that change dynamically with expression. It enables highly realistic facial animation from in-the-wild images.

    face-reconstruction facial-expression wrinkles

    GitHub · arXiv

Proprietary / Industry Tools

  • FA

    FaceGen Discovered by AI

    2024

    A commercial 3D face creation software that generates realistic 3D human heads from one or more photos, with automatic generation of animation targets and compatibility with 3D scan data.

    3d-face blendshapes commercial head-generation

    Website

  • H3

    Headshot 3 Discovered by AI

    2024

    An AI-powered 3D head generator plugin for Reallusion Character Creator that automatically transforms single portrait photos into highly detailed, animatable 3D digital human heads.

    3d-head character-creator image-to-3d proprietary

    Website

  • KF

    KeenTools FaceBuilder for Blender NEW Discovered by AI

    2024

    A Blender plugin that creates custom 3D human head models with texture projection from photos using a parametric 3D face model.

    3d-head 3dmm blender-addon face-reconstruction

    Website

  • FW

    Faceform Wrap Discovered by AI

    2023

    An industry-standard topology transfer and 3D scan processing tool that converts 3D scans of actor facial expressions into a consistent topology, widely used for facial rigging and blendshape creation in games and VFX.

    3d-scan blendshapes topology-transfer

    Website

  • KF

    KeenTools FaceBuilder Discovered by AI

    2023

    A popular plugin for Blender and Nuke that allows artists to create photorealistic 3D human faces and heads from a single or multiple photos, featuring AI-powered auto-alignment, texture blending, and ARKit-compatible FACS blendshapes.

    blender-plugin blendshapes face-reconstruction

    Website

  • RH

    Reallusion Headshot Discovered by AI

    2020

    An AI-powered 3D head generator plugin for Character Creator that creates fully rigged, animation-ready 3D heads from a single photo or imported 3D mesh.

    character-creator photo-to-3d plugin

    Website

  • FF

    FaceBuilder for Blender (KeenTools) Discovered by AI

    2019

    A popular commercial add-on for Blender and Nuke that allows artists to easily sculpt 3D human faces and heads based on a few photographs, producing game-ready, animatable meshes.

    blender-addon commercial face-modeling photogrammetry

    Website