Avatar Generation¶
Characters & Avatars
Open Source¶
-
AptAvatar Discovered by AI
2026
A 14-billion parameter long-form audio-driven avatar generation framework designed for production-ready, fast, and expressive inference. It utilizes endpoint-anchored distribution distillation and self-generated history replay to deliver vivid motion and high-fidelity video generation without quality degradation over long durations.
audio-drivendistillationtalking-headvideo-generation -
Avatar Forcing Discovered by AI
2026
A real-time interactive head avatar generation framework that models user-avatar interactions through diffusion forcing, enabling the avatar to process real-time multimodal inputs and mirror expressions.
avatarinteractivereal-timetalking-head -
Avatar-Forever Discovered by AI
2026
A decoupled parallel training framework for high-quality, real-time, and effectively unbounded audio-driven avatar generation. It treats generation efficiency and long-horizon robustness as independent capabilities trained in parallel, achieving 27.2 FPS on a single H100 using an LTX-Video backbone.
audio-drivenavatarreal-timetalking-head -
DAMA Discovered by AI
2026
A 3D clothed avatar reconstruction method that uses disentangled body-anchored Gaussians to enable controllable, multi-layered garments with clear boundaries and layer-ordering control.
avatarclothed-humangaussian-splatting -
DynaAvatar Discovered by AI
2026
A zero-shot framework that reconstructs animatable 3D human avatars with motion-dependent cloth dynamics from a single image. It employs a Transformer-based feed-forward architecture to predict dynamic 3D Gaussian deformations without subject-specific optimization.
3dgsavatarcloth-dynamicszero-shot -
GenLCA: 3D Diffusion for Full-Body Avatars from In-the-Wild Videos Discovered by AI
2026
A diffusion-based generative model developed by Meta and Zhejiang University for generating and editing photorealistic, animatable full-body 3D Gaussian avatars from text and image inputs, trained on large-scale in-the-wild video data.
3dgsavatardiffusionvideo-to-3d -
Hallo-Live Discovered by AI
2026
A real-time streaming framework for joint audio-visual avatar generation that combines asynchronous dual-stream diffusion with human-centric preference-guided distillation.
avatarreal-timetalking-head -
HumanNOVA Discovered by AI
2026
A photorealistic, universal, and rapid framework for generating animatable 3D human avatars from a single RGB image, leveraging a scalable data generation pipeline with rigged assets and extensive poses.
3d-humanavatargaussian-splatting -
LongCat-Video-Avatar Discovered by AI
2026
An open-source, production-ready framework for audio-driven talking avatar video generation, supporting long video generation with stable lip synchronization and consistent identity.
avatardiffusiontalking-head -
LUNA Discovered by AI
2026
An LBS-free universal neural animation model that directly maps diverse 2D controls, such as images, keypoints, and sketches, into 3D Gaussian deformations. It bypasses explicit body fitting and linear blend skinning to deliver realistic human motion and zero-shot cross-identity generalization.
3d-gaussiananimationavatarlbs-free -
Monocular Avatar Reconstruction via Cascaded Diffusion Priors and UV-Space Differentiable Shading Discovered by AI
2026
A monocular 3D avatar reconstruction framework that leverages cascaded diffusion priors and UV-space differentiable shading to generate high-fidelity, relightable 3D avatars from a single in-the-wild image.
3d-avatarmonocular-reconstructionrelightable -
MVCHead Discovered by AI
2026
A single-shot state space model that learns multi-view consistent 3D Gaussian head avatars from randomly sampled 2D images alone, without requiring multi-view datasets, 3D supervision, or intermediate view generation.
avatargaussian-splattinghead-avatar -
SplatShot Discovered by AI
2026
A framework for reconstructing photorealistic 3D face avatars from a single unconstrained photograph by combining feed-forward 3D Gaussian Splatting with 2D diffusion priors to ensure both geometric consistency and photorealism.
avatarface-avatargaussian-splatting -
UP2You: Fast Reconstruction of Yourself from Unconstrained Photo Collections Discovered by AI
2026
A tuning-free solution for reconstructing high-fidelity 3D clothed portraits from extremely unconstrained, raw in-the-wild 2D photos by rectifying them into orthogonal multi-view images and normal maps.
avatarmeshmulti-viewreconstruction -
AdaHuman Discovered by AI
2025
A framework that generates high-fidelity animatable 3D avatars from a single in-the-wild image using a pose-conditioned 3D joint diffusion model and 3D Gaussian Splatting reconstruction.
3d-humanavatargaussian-splatting -
AnimPortrait3D Discovered by AI
2025
A framework for text-based realistic animatable 3DGS head avatar generation that aligns the 3D Gaussian representation with a morphable head model to ensure precise facial animation control.
3dgsavatarheadtext-to-3d -
Avat3r Discovered by AI
2025
A large animatable Gaussian reconstruction model that regresses high-quality and animatable 3D head avatars from just a few input images, significantly reducing test-time optimization and compute requirements.
avatargaussian-splattinghead-avatar -
GSAC: Leveraging Gaussian Splatting for Photorealistic Avatar Creation with Unity Integration Discovered by AI
2025
A framework that leverages 3D Gaussian Splatting to create photorealistic, animatable avatars from smartphone videos, with seamless integration into the Unity game engine for real-time interactive applications.
3d-avatargaussian-splattingreal-timeunity -
InfiniHuman Discovered by AI
2025
A framework for generating diverse and highly controllable 3D human avatars from multiple modalities (text, clothing, body pose, or shape) using a scalable 3D human dataset generated via 2D diffusion models and a robust 3D reconstruction pipeline.
3d-avataravatar-generationcontrollable-generationsiggraph-asia -
JoyStreamer-Flash Discovered by AI
2025
An audio-driven autoregressive generative model synthesizing real-time, infinite-length avatar videos via advanced diffusion Transformers, ensuring superior lip-sync, temporal coherence, and visual fidelity.
autoregressiveavatarreal-timetalking-head -
LHM: Large Animatable Human Reconstruction Model from a Single Image in Seconds Discovered by AI
2025
LHM is a feed-forward large animatable human reconstruction model that generates fully animatable 3D human avatars from a single input image in seconds. It decouples geometry, appearance, and deformation to produce production-ready 3D character models without lengthy optimization.
3d-avataranimatable-humangaussian-splattingsingle-image-to-3d -
LiveAvatar Discovered by AI
2025
An algorithm-system co-designed framework for real-time, streaming, and infinite-length audio-driven avatar video generation based on a 14B diffusion model.
audio-drivenavatarreal-timestreaming -
OmniHuman-1.5 Discovered by AI
2025
An advanced video avatar generation framework by ByteDance that integrates cognitive simulation to produce semantically coherent and expressive character animations. By leveraging Multimodal Large Language Models for high-level semantic guidance, it generates realistic full-body movements and facial expressions aligned with the speech's rhythm and emotional context.
avatarbytedancediffusion-transformertalking-head -
RMAvatar: Photorealistic Human Avatar Reconstruction from Monocular Video Based on Rectified Mesh-embedded Gaussians Discovered by AI
2025
A novel human avatar representation that embeds 3D Gaussian Splatting onto explicit mesh geometry to reconstruct clothed avatars from monocular video. It features a pose-related Gaussian rectification module to capture complex non-rigid deformations, improving realism and expressiveness.
3d-gaussianavatarmonocular-videoreconstruction -
TaoAvatar: Real-Time Lifelike Full-Body Talking Avatars for Augmented Reality via 3D Gaussian Splatting Discovered by AI
2025
A high-fidelity, lightweight, 3DGS-based full-body talking avatar framework designed for real-time performance on mobile and AR devices, driven by various facial and body motion signals.
3dgsaugmented-realityavatarreal-time -
TeRA: Rethinking Text-guided Realistic 3D Avatar Generation Discovered by AI
2025
A text-guided 3D realistic avatar generation framework based on a latent diffusion model. It employs a two-stage training strategy, distilling a decoder to derive a structured latent space from a large human reconstruction model, enabling fast feed-forward generation and text-based partial customization.
avatardiffusionreconstructiontext-to-3d -
AniGS: Animatable Gaussian Avatar from a Single Image with Inconsistent Gaussian Reconstruction Discovered by AI
2024
A method for generating animatable 3D human avatars from a single image. It utilizes a video diffusion model to generate multi-view canonical pose images and models multi-view inconsistencies as temporal changes in 4D Gaussian Splatting, enabling high-fidelity reconstruction and real-time animation.
3dgs4dgshuman-avatarsingle-image -
Audio2Photoreal Discovered by AI
2024
An open-source framework and dataset by Meta for synthesizing photorealistic, full-body Codec Avatars with natural body and hand gestures driven directly from conversational audio.
audio-drivenavatarcodec-avatarfull-body -
DreamBarbie: Text to Barbie-Style 3D Avatars Discovered by AI
2024
A text-driven 3D avatar generation framework that produces high-quality, fine-grained, and disentangled 3D avatars in a stylized aesthetic, supporting expressive animation and simulation.
3d-avatarstylized-avatartext-to-3d -
DreamWaltz-G Discovered by AI
2024
A learning framework for animatable 3D avatar generation from text using skeleton-guided score distillation and hybrid 3D Gaussian avatar representation.
3dgsavatardiffusiontext-to-3d -
Duix-Mobile Discovered by AI
2024
An open-source SDK for deploying real-time interactive AI avatars directly on mobile devices and embedded screens. It supports on-device deployment with low latency, enabling developers to integrate LLM, ASR, and TTS services for natural conversations.
avatarinteractiveon-devicesdk -
Duix.Avatar Discovered by AI
2024
An open-source, fully offline AI digital human toolkit that enables appearance and voice cloning from short video inputs. It supports text-to-video and audio-driven lip-synced video generation with high privacy and low latency.
avatarcloningdigital-humanoffline -
ExAvatar Discovered by AI
2024
A method to reconstruct expressive whole-body 3D human avatars (including body, face, and hands) from a short monocular video using SMPL-X and 3D Gaussian Splatting.
3dgsavatarsmpl-xwhole-body -
From Audio to Photoreal Embodiment: Synthesizing Humans in Conversations (Audio2Photoreal) Discovered by AI
2024
A framework for synthesizing photorealistic, full-body 3D avatars (Codec Avatars) that gesture and move naturally in response to conversational audio.
audio-drivenavatarcodec-avatarmotion-synthesis -
Human-3Diffusion Discovered by AI
2024
A 3D consistent diffusion model designed to create realistic, animatable 3D human avatars from a single RGB image by integrating 2D diffusion models with 3D Gaussian Splatting.
3dgsavatardiffusionreconstruction -
IDOL: Instant Photorealistic 3D Human Creation from a Single Image Discovered by AI
2024
A feed-forward transformer-based framework that reconstructs high-fidelity, animatable 3D human avatars from a single image. It predicts 3D Gaussian representations in a uniform space, allowing instant 1K resolution reconstruction and seamless animation.
3d-gaussiananimatableavatarreconstruction -
One Shot, One Talk: Whole-body Talking Avatar from a Single Image Discovered by AI
2024
A pipeline for constructing a whole-body talking avatar from a single image, utilizing pose-guided video generation models and 3D Gaussian Splatting to achieve realistic body gestures and natural expression changes.
3dgsavatarone-shottalking-head -
PuzzleAvatar Discovered by AI
2024
A framework for generating personalized, high-fidelity 3D clothed human avatars from unconstrained personal photo collections (albums) by encoding appearance and identity features into tokens and assembling them into a canonical 3D model.
3d-avataravatar-generationclothed-humanreconstruction -
UltrAvatar Discovered by AI
2024
A 3D avatar diffusion model that generates realistic, animatable 3D avatars with high-quality PBR textures from a single image or text prompt, utilizing a Diffuse Color Extraction model and an Authenticity Guided Texture Diffusion Model.
3d-avataravatar-generationdiffusionpbr-textures -
X-Oscar: A Progressive Framework for High-quality Text-guided 3D Animatable Avatar Generation Discovered by AI
2024
A progressive framework for generating high-quality animatable 3D avatars from text prompts. It follows a sequential "Geometry -> Texture -> Animation" paradigm to simplify optimization and avoid oversaturation, producing high-fidelity avatars with realistic textures and fluid animations.
3d-generationanimatableavatartext-to-3d -
X-Portrait Discovered by AI
2024
An innovative conditional diffusion model designed for expressive and temporally coherent portrait animation using hierarchical motion attention.
avatardiffusionportrait-animation -
3DGS-Avatar: Animatable Avatars via Deformable 3D Gaussian Splatting Discovered by AI
2023
An approach that creates animatable human avatars from monocular videos using deformable 3D Gaussian Splatting. It achieves high-quality novel-view and novel-pose image synthesis with fast training and real-time rendering.
3d-avataranimatablegaussian-splattingreal-time -
Animatable Gaussians: Learning Pose-dependent Gaussian Maps for High-fidelity Human Avatar Modeling Discovered by AI
2023
Animatable Gaussians models high-fidelity full-body human avatars from sparse RGB videos by learning pose-dependent 2D Gaussian maps. It combines 2D CNNs with 3D Gaussian Splatting to capture dynamic garment deformations and fine-grained geometric details for realistic rendering.
3d-avataranimatable-avatargaussian-splattinghuman-modeling -
DreamAvatar: Text-and-Shape Guided 3D Human Avatar Generation via Diffusion Models Discovered by AI
2023
A text-and-shape guided framework for generating high-quality 3D human avatars with controllable poses. It utilizes a trainable NeRF for predicting density and color, and leverages pretrained 2D text-to-image diffusion models to guide the generation process.
3d-avatardiffusionnerftext-to-3d -
GaussianAvatars Discovered by AI
2023
A method for creating photorealistic head avatars that are fully controllable in terms of expression, pose, and viewpoint by rigging 3D Gaussian splats to a parametric morphable face model.
3dmmfacial-expressiongaussian-splattinghead-avatar -
2022
Zero-shot text-driven generation and animation of 3D avatars by guiding a neural human model with CLIP, an early framework for describable digital humans.
avatartext-driven
Proprietary / Industry Tools¶
-
MetaHuman Discovered by AI
2024
MetaHuman is Epic Games' framework and creation tool for generating fully rigged, photorealistic digital human avatars within Unreal Engine. It enables real-time editing of facial features, body types, clothing, and grooms, natively outputting game-ready 3D character rigs.
digital-humanmetahumanrigged-avatarunreal-engine -
Avaturn Discovered by AI
2023
An AI-powered 3D avatar creator that generates realistic, fully rigged, and customizable game-ready avatars from a single selfie.
3davatargame-readyrigged -
Didimo Popul8 Discovered by AI
2023
A generative AI character creation platform designed for game developers to instantly generate, edit, and scale high-quality, game-ready 3D avatars and NPCs.
3d-avataravatargame-readynpc -
Popul8 Discovered by AI
2023
Didimo's AI-powered character creation platform designed for game studios to generate, edit, and manage hundreds of unique, game-ready 3D avatars and NPCs at scale.
avatarcharacter-generationgame-devnpc -
Meshcapade Me Discovered by AI
2021
An AI-powered platform built on the SMPL body model that generates highly accurate, industry-standard 3D human avatars from photos, videos, or body measurements.
3d-avatarbody-measurementdigital-humansmpl -
HeyGen Discovered by AI
2020
A leading commercial AI video generation platform that uses highly realistic talking avatars and voice cloning to create professional videos from text scripts.
avatarcommercialtalking-headvideo-generation -
Meshcapade Discovered by AI
2020
An AI-powered platform that generates realistic 3D human avatars and motion capture data from photos, videos, or 3D scans using the SMPL body model.
3d-avataravatarsmpl -
Epic's tool for creating fully rigged, photorealistic digital humans for Unreal Engine, an industry standard for game character and cinematic avatars.
commercialriggedunreal
