Pose-to-Frame

Animation & Motion

Open Source

• EM

  EverAnimate: Minute-Scale Human Animation via Latent Flow Restoration Discovered by AI

  2026

  An efficient post-training method for long-horizon animated video generation that preserves visual quality and character identity. It introduces Persistent Latent Propagation to maintain context memory across chunks and Restorative Flow Matching to improve within-chunk fidelity, enabling minute-scale human animation.

  flow-matching human-animation long-video pose-guided

  Project · GitHub · arXiv

• LI

  LiveAnimate Discovered by AI

  2026

  A real-time, long-form pose-driven human animation system built on a 14B-parameter video Diffusion Transformer. It utilizes bounded attention caching and Pose-Retrieval Sink Attention to maintain appearance consistency and stable memory usage over extended streaming sequences.

  long-form pose-driven real-time streaming

  GitHub · arXiv

• MU

  MultiAnimate Discovered by AI

  2026

  A pose-guided human image animation framework designed for multi-character scenarios, addressing identity confusion and occlusion issues using an extensible architecture.

  diffusion human-animation multi-character pose-guided

  GitHub · arXiv

• SL

  SoulX-LiveAct Discovered by AI

  2026

  An autoregressive diffusion framework for hour-scale, real-time human animation, utilizing neighbor forcing and ConvKV memory to achieve streaming inference at 20 FPS.

  autoregressive diffusion human-animation real-time

  GitHub · arXiv

• WA

  Wan-Animate-2 Discovered by AI

  2026

  An end-to-end character animation framework that directly consumes driving videos in a redesigned Diffusion Transformer, eliminating intermediate motion extractors to achieve high-fidelity motion generation, strong identity preservation, and text-driven viewpoint control.

  character-animation diffusion-transformer motion-transfer viewpoint-control

  GitHub · arXiv

• AA

  Animate Anyone 2 Discovered by AI

  2025

  Animate Anyone 2 extends character image animation by incorporating environment affordance, capturing environmental interaction signals from source videos to generate realistic character-environment interactions.

  character-animation environment-affordance pose-guided

  arXiv

• AX

  Animate-X++ Discovered by AI

  2025

  An extension of the Animate-X framework that supports dynamic, text-driven backgrounds alongside universal character animation, utilizing a Diffusion Transformer (DiT) architecture.

  character-animation dynamic-background pose-guided

  GitHub · arXiv · Paper

• AR

  AnimateAnywhere: Rouse the Background in Human Image Animation Discovered by AI

  2025

  A human image animation framework built on CogVideoX that generates realistic videos with synchronized motion of both the human and the background. It introduces a Background Motion Learner (BML) to infer background motion from human pose sequences without requiring camera trajectories.

  background-motion cogvideox human-animation pose-guided

  Project · GitHub · arXiv

• DY

  DynamiCtrl Discovered by AI

  2025

  A pose-guided human image animation framework that integrates pose control structures with text guidance in a Multimodal Diffusion Transformer (MM-DiT) to preserve identity and enable controllable backgrounds.

  dit human-animation pose-guided text-guidance

  GitHub · arXiv

• HU

  HumanDiT Discovered by AI

  2025

  A pose-guided Diffusion Transformer (DiT) framework trained on 14,000 hours of video for long-form human motion video generation. It achieves high-fidelity rendering of detailed body parts like hands and faces across variable resolutions and sequence lengths.

  diffusion-transformer long-form pose-guided

  Project · arXiv

• HY

  HyperMotion Discovered by AI

  2025

  A pose-guided human image animation framework based on a large-scale video diffusion Transformer, designed to handle complex, highly dynamic, and non-standard motions.

  complex-motion diffusion-transformer human-animation pose-guided

  GitHub · arXiv

• IT

  InterAnimate: Taming Region-aware Diffusion Model for Realistic Human Interaction Animation Discovered by AI

  2025

  InterAnimate is a region-aware diffusion model designed for realistic human interaction animation, specifically targeting complex hand-face interactions. It introduces a region-aware attention system and an ID Preserver to generate natural and temporally coherent animations from reference images and motion guidance.

  character-animation diffusion-model pose-guided

  arXiv · Paper

• MT

  MTVCraft: Tokenizing 4D Motion for Arbitrary Character Animation Discovered by AI

  2025

  A character animation framework that directly models raw 3D motion sequences (4D motion) instead of relying on 2D-rendered pose images. It tokenizes 4D motion to enable precise, open-world animation of arbitrary characters.

  4d-motion character-animation pose-guided

  GitHub · arXiv

• PO

  PoseAnything Discovered by AI

  2025

  A universal pose-guided video generation framework capable of handling both human and non-human characters from arbitrary skeletal inputs with part-aware temporal coherence.

  pose-guided temporal-coherence universal video-generation

  GitHub · arXiv

• RD

  RealisDance-DiT: Simple yet Strong Baseline towards Controllable Character Animation in the Wild Discovered by AI

  2025

  A structurally simple and robust baseline model for controllable character animation built upon the Wan-2.1 video diffusion transformer (DiT). It replaces the complex Reference Net design with a simpler, more efficient conditioning mechanism to achieve high-quality, temporally consistent human motion generation in diverse real-world scenarios.

  character-animation diffusion-transformer pose-guided video-generation

  Project · GitHub · arXiv

• SA

  SkyReels-A1 Discovered by AI

  2025

  An expressive portrait and character animation framework built on video diffusion Transformers (DiTs) that addresses identity distortion and background instability.

  diffusion-transformer portrait-animation pose-guided

  GitHub · arXiv

• ST

  StableAnimator++ Discovered by AI

  2025

  An identity-preserving video diffusion framework for human image animation that introduces learnable pose alignment to overcome pose misalignment and face distortion. It generates high-quality videos conditioned on a reference image and a pose sequence without requiring post-processing.

  diffusion identity-preserving pose-guided

  GitHub · arXiv

• UD

  UniAnimate-DiT Discovered by AI

  2025

  An advanced human image animation framework based on the large-scale video diffusion transformer Wan2.1-14B-I2V, enabling highly consistent and high-fidelity character animation.

  diffusion-transformer human-animation pose-guided

  GitHub · arXiv · Hugging Face · Paper

• AX

  Animate-X Discovered by AI

  2024

  A universal character image animation framework based on latent diffusion models that generalizes well to anthropomorphic, anime, and 3D characters. It introduces an enhanced motion representation to capture complex movement patterns.

  character-animation diffusion pose-guided

  GitHub · arXiv · Paper

• CH

  Champ Discovered by AI

  2024

  Champ is a controllable and consistent human image animation framework that leverages a 3D human parametric model (SMPL) within a latent diffusion framework to enhance shape alignment and motion guidance.

  character-animation pose-guided smpl

  GitHub · arXiv

• DI

  DisPose Discovered by AI

  2024

  A controllable human image animation method that disentangles sparse skeleton pose guidance into motion field guidance and keypoint correspondence, improving generalization across different body shapes.

  diffusion disentanglement pose-guided

  GitHub · arXiv

• FY

  Follow-Your-Pose v2 Discovered by AI

  2024

  A robust pose-controllable character animation framework that integrates multiple conditions into a video latent diffusion model. It is designed to address background stability and body occlusion issues in multi-character scenarios.

  character-animation multi-character pose-guided

  arXiv

• MI

  MimicMotion Discovered by AI

  2024

  MimicMotion is a controllable video generation framework that generates high-quality human motion videos of arbitrary length with confidence-aware pose guidance and regional loss amplification.

  character-animation pose-guided video-generation

  GitHub · arXiv

• MA

  Moore-AnimateAnyone Discovered by AI

  2024

  An open-source implementation of the Animate Anyone pipeline that enables character animation and face reenactment using reference images and driving pose or facial landmark sequences.

  character-animation face-reenactment pose-guided

  GitHub

• MU

  MusePose Discovered by AI

  2024

  MusePose is a diffusion-based, pose-guided virtual human video generation framework designed to generate high-quality dance and movement videos of a character from a single reference image.

  character-animation pose-guided virtual-human

  GitHub

• ST

  StableAnimator Discovered by AI

  2024

  An end-to-end identity-preserving video diffusion framework for human image animation that synthesizes high-quality videos conditioned on a reference image and a sequence of poses without post-processing.

  diffusion identity-preserving pose-guided

  GitHub · arXiv · Paper

• TA

  TCAN: Animating Human Images with Temporally Consistent Pose Guidance using Diffusion Models Discovered by AI

  2024

  A pose-driven human image animation method designed to be robust to erroneous poses and highly consistent over time. It utilizes a pre-trained ControlNet without fine-tuning, combined with a temporal consistency module and LoRA, to achieve stable video synthesis.

  controlnet lora pose-guided temporal-consistency

  GitHub · arXiv · Website

• UN

  UniAnimate Discovered by AI

  2024

  UniAnimate is a framework for efficient and long-term human video generation that integrates a unified video diffusion model with a pose guider and temporal modeling to support consistent character animation.

  character-animation pose-guided video-generation

  GitHub · arXiv

• VI

  VividPose Discovered by AI

  2024

  An end-to-end human image animation framework that leverages the motion-aware prior knowledge of Stable Video Diffusion (SVD) to generate temporally consistent and smooth videos from a single reference image and pose sequence.

  diffusion pose-guided svd

  GitHub · arXiv

• AA

  Animate Anyone Discovered by AI

  2023

  Animate Anyone is a framework for consistent and controllable image-to-video synthesis for character animation, utilizing a ReferenceNet to maintain appearance consistency and a pose guider to control character movement.

  character-animation diffusion pose-guided

  GitHub · arXiv · Website

• DI

  DisCo Discovered by AI

  2023

  A generalized human dance generation framework that disentangles control over the foreground person, background, and pose, enabling high-fidelity human image and video generation in real-world scenarios.

  dance-generation human-animation pose-guided

  Project · GitHub · arXiv

• MA

  MagicAnimate Discovered by AI

  2023

  A framework for temporally consistent human image animation using diffusion models. It leverages a video diffusion model with a novel appearance encoder and temporal attention to animate a reference identity following a target motion sequence.

  diffusion-model human-animation pose-guided

  Project · GitHub · arXiv

• MA

  MagicDance Discovered by AI

  2023

  A diffusion-based model for 2D human pose and facial expression retargeting. It generates new images of a person by controlling poses and facial expressions while accurately maintaining identity from a reference image.

  diffusion-model expression-retargeting pose-guided

  Project · GitHub · arXiv