PersonaAnimator: Personalized Video Motion Transfer

• PersonaAnimator is a framework for video-to-video motion personalization that fuses content motion from one video with personalized style from another using 2D skeleton representations.
• It employs disentangled encoders, semantic gating, AdaIN fusion, and a diffusion-based generator to achieve identity-consistent, physics-aware motion transfer.
• Empirical results on the PersonaVid dataset show improved FVD, LPIPS, and SSIM metrics compared to baselines, with significant gains in realism and style fidelity.

to=arxiv_search.query 买天天中彩票 手机天天中彩票{"query":"PersonaAnimator personalized motion transfer unconstrained videos arXiv", "max_results": 5, "sort_by": "relevance"} to=arxiv_search.query ությանը 天天彩票是 code 招商总代{"query":"animepose multi-person 3d pose estimation and animation arXiv (Kumarapu et al., 2020)", "max_results": 3, "sort_by": "relevance"} to=arxiv_search.query 大发彩票网 ￣亚洲 code numerusform{"query":"AniCrafter avatar-background conditioning video diffusion arXiv", "max_results": 3, "sort_by": "relevance"} to=arxiv_search.query 亿贝ీన 彩神争霸可以code เน็ตทรู{"query":"MultiAnimate Pose-Guided Image Animation Made Extensible arXiv", "max_results": 3, "sort_by": "relevance"} to=arxiv_search.query 久久免费热在线精品 公众号天天中彩票 code 早点加盟{"query":"EgoAnimate Generating Human Animations from Egocentric top-down Views arXiv", "max_results": 3, "sort_by": "relevance"} PersonaAnimator denotes a framework for Video-to-Video Motion Personalization (V2V-MP) that learns personalized motion patterns directly from unconstrained videos, enabling synthesis of a new video that follows the content motion from one source while expressing the personalized style from another (Qian et al., 27 Aug 2025). In adjacent arXiv-facing technical reports, the same name is also used as a target system label for broader animation pipelines, including multi-person 3D pose animation, egocentric avatar reconstruction, dynamic-background animation, and multi-character diffusion animation (Kumarapu et al., 2020, Türkoglu et al., 12 Jul 2025, Niu et al., 26 May 2025, Hu et al., 25 Feb 2026). The most specific and canonical use of the term is therefore the 2025 V2V-MP framework, but its surrounding ecosystem situates it within a wider transition from rigid pose transfer toward style-aware, identity-consistent, and physically constrained animation.

1. Conceptual scope and problem setting

PersonaAnimator formalizes Video-to-Video Motion Personalization as a task with two inputs: a content video $V_c$, showing a person performing some motion $m$, and a style video $V_s$, showing a different person performing some motion $n$. Skeletons extracted by a pose estimator are written as $P_c^m \in \mathbb{R}^{F \times J \times 2}$ and $P_s^n \in \mathbb{R}^{F \times J \times 2}$, and the generator is defined so that

$G(P_c^m, P_s^n) = \hat P_s^m,$

where $\hat P_s^m$ preserves the structure of $P_c^m$ while imbuing it with the “dynamic essence” of $P_s^n$ (Qian et al., 27 Aug 2025).

The framework is motivated by three limitations attributed to prior work. First, pose-guided transfer is described as simply replicating joint positions and never learning person-specific style such as swing amplitude or posture idiosyncrasies. Second, motion-style transfer methods are said to rely on high-quality mocap or SMPL data, which is expensive to collect and unavailable for many real-world subjects. Third, the absence of explicit skeleton constraints can produce physical implausibility, including bone-length jitter, foot sliding, or unnatural stretching (Qian et al., 27 Aug 2025).

A common source of confusion is that PersonaAnimator is not a single unified architecture across all documents bearing that name. The 2025 paper defines a specific V2V-MP method and benchmark (Qian et al., 27 Aug 2025). By contrast, several technical reports use “PersonaAnimator” as the name of a downstream system assembled from other backbones, such as AnimePose, AniCrafter, EgoAnimate, or MultiAnimate. This suggests that the term has both a paper-specific meaning and a broader system-design meaning in the recent literature.

2. PersonaVid and the data model of personalization

PersonaAnimator is paired with PersonaVid, described as the first video-based personalized motion dataset (Qian et al., 27 Aug 2025). The dataset is organized around two orthogonal taxonomies: 20 motion content categories and 120 motion style categories. Under a “one-person-one-style” paradigm, each individual is treated as a distinct style label, with examples such as “Walk_Trump_05” and “Dance_D10_01” (Qian et al., 27 Aug 2025).

The reported scale is 18,867 videos, with an average of approximately 157 clips per style and an average video length of approximately 90 frames. The split is defined by style class: 96 style classes for training, 12 for validation, and 12 for test, corresponding to approximately 15,100, 1,900, and 1,900 clips, respectively (Qian et al., 27 Aug 2025).

Annotation and preprocessing are explicitly specified. Each clip is manually labeled Content_Style_ID. Skeletons are extracted with DWPose and normalized into $m$0. All videos are resized to $m$1, and augmentation includes random horizontal flip, small rotation of $m$2, and temporal jitter of $m$3 frames (Qian et al., 27 Aug 2025).

The dataset design is significant because it redefines “style” as an individual-specific motion signature rather than a coarse action label. A plausible implication is that personalization is operationalized at the level of recurrent kinematic traits rather than static appearance or categorical motion classes.

3. Architecture and representation learning

The PersonaAnimator pipeline separates content, style, semantics, and appearance before recombining them in a diffusion-based video generator (Qian et al., 27 Aug 2025). The content video $m$4 is passed through DWPose to obtain $m$5, which is then encoded by a ContentEncoder into $m$6. The style video $m$7 is likewise processed by DWPose and a StyleEncoder to obtain $m$8. A SemanticTokenizer converts the content label into $m$9, which is used as a weighting signal for the style pathway.

The semantic-guided weighting is written as

$V_s$0

This is followed by an AdaIN fusion stage,

$V_s$1

and then by an SA-PMT encoder with MHA and FFN, producing $V_s$2, described as “personalized motion features” (Qian et al., 27 Aug 2025).

Appearance is conditioned separately from motion. A reference frame $V_s$3 from $V_s$4 is encoded by CLIP and a VAE into $V_s$5 and $V_s$6. These features, together with $V_s$7 and diffusion noise $V_s$8, are injected into a video-diffusion U-Net equipped with spatial, motion, and temporal attention. The decoded output is the synthesized video $V_s$9 (Qian et al., 27 Aug 2025).

The encoders for skeletons are specified as 4-layer MLPs with hidden size 256 plus positional encoding. The SA-PMT module uses AdaIN with learned scale/shift MLP (hidden size 128) and a Transformer block with 4-head MHA where $n$0, plus an FFN with 512 hidden. The diffusion network is a U-Net with 1,000 diffusion steps and a linear $n$1 schedule. Appearance conditioning uses CLIP ViT-B/16 and a VAE latent of resolution $n$2 (Qian et al., 27 Aug 2025).

Architecturally, the framework is organized around disentanglement followed by controlled recombination: the content path captures pure joint trajectories, the style path captures style dynamics, semantic gating suppresses style elements unrelated to the content category, and AdaIN aligns the first two moments of content features to style statistics. This suggests that PersonaAnimator treats style transfer as a constrained reparameterization of motion dynamics rather than direct skeleton copying.

4. Objective functions and physics-aware regularization

PersonaAnimator’s training objective combines reconstruction, style consistency, and explicit physical regularization (Qian et al., 27 Aug 2025). The losses are grouped into three components.

The first component is motion reconstruction and appearance alignment, denoted $n$3, using MSE. The second is style consistency, denoted $n$4, adopted from Aberman et al. 2020 to ensure that the generated motion $n$5 retains the style embedding of the style video. The third is Physics-aware Motion Style Regularization, denoted

$n$6

The Dynamic Bone Stability term operates on each bone $n$7, with bone length

$n$8

and second-order temporal change

$n$9

The penalty averages $P_c^m \in \mathbb{R}^{F \times J \times 2}$0 over frames and bones (Qian et al., 27 Aug 2025).

The Body Connectivity term uses a skeleton adjacency matrix $P_c^m \in \mathbb{R}^{F \times J \times 2}$1, where $P_c^m \in \mathbb{R}^{F \times J \times 2}$2 if joints $P_c^m \in \mathbb{R}^{F \times J \times 2}$3 should be close, and pairwise distances $P_c^m \in \mathbb{R}^{F \times J \times 2}$4. It combines a positive term over adjacent joints and a negative term over non-adjacent joints with minimum-distance threshold $P_c^m \in \mathbb{R}^{F \times J \times 2}$5, yielding $P_c^m \in \mathbb{R}^{F \times J \times 2}$6 (Qian et al., 27 Aug 2025).

The total objective is

$P_c^m \in \mathbb{R}^{F \times J \times 2}$7

This regularization is central to the method’s identity. The paper’s framing is not merely that style should be transferred, but that style should remain compatible with skeletal plausibility. In that sense, PersonaAnimator differs from motion stylization methods that prioritize expressivity without an explicit mechanism for suppressing jitter or stretching.

5. Empirical results, ablations, and limitations

Evaluation on the PersonaVid test split compares PersonaAnimator against three baselines: StableAnimator+DIFF, AnimateAnyone+DIFF, and Animate-X+DIFF (Qian et al., 27 Aug 2025). The reported metrics include FID, FVD, SSIM, PSNR, LPIPS, style-feature distance, and a user study with 30 participants ranking realism, style faithfulness, and identity consistency.

Metric	Baseline best	PersonaAnimator
FVD $P_c^m \in \mathbb{R}^{F \times J \times 2}$8	4200	3180
LPIPS $P_c^m \in \mathbb{R}^{F \times J \times 2}$9	0.184	0.138
Style dist $P_s^n \in \mathbb{R}^{F \times J \times 2}$0	1.45	1.02
SSIM $P_s^n \in \mathbb{R}^{F \times J \times 2}$1	0.72	0.81

The paper reports that PersonaAnimator is chosen 78% of the time in the user study versus the best baseline (Qian et al., 27 Aug 2025). Qualitatively, it is described as preserving content structure while capturing subtle style cues such as sweep angle of arms and gait sway, and as avoiding bone-length jitter with smooth foot contacts.

The ablation study isolates three major components. Removing SA-PMT and replacing it with DIFF causes style retention to drop, with LPIPS increasing by +0.07 and user choice falling to 42%. Removing PMSR causes marked bone jitter and worsens FVD by +650. Removing semantic gating introduces style noise, with examples such as facial gestures where none exist (Qian et al., 27 Aug 2025).

The stated limitations are equally explicit. The current method operates on 2D skeletons, which creates depth ambiguity and occasional self-occlusion artifacts. Inference for a 5-second clip takes approximately 30 seconds on a single A800, so the method is not real-time. The reported setting is also single-person only (Qian et al., 27 Aug 2025). These limitations delimit the scope of the original framework: it is a style-personalized motion transfer system rather than a full 3D or multi-actor animation platform.

6. Relation to adjacent PersonaAnimator-style systems

Outside the V2V-MP formulation, “PersonaAnimator” appears as a design target in several technically distinct animation pipelines. These works do not redefine the 2025 method; rather, they show how the label is used for neighboring problems.

A multi-person 3D pose and tracking interpretation is built from AnimePose. In that pipeline, an RGB video sequence is processed through Hybrid Task Cascade (HTC) with HRNet, AlphaPose, Multi-Scale Local Planar Guidance depth estimation, Martinez et al.-style 2D-to-3D lifting, STAF for frame-to-frame association, a 3D-IOU metric for occlusion handling, and Scene-LSTM for missing trajectories, with tracked skeletons streamed into Unity3D for real-time visualization (Kumarapu et al., 2020). The same report cites Kumarapu & Mukherjee, IIIT Sri City, 2020 as the source paper and reports 82.1 average 3DPCK$P_s^n \in \mathbb{R}^{F \times J \times 2}$2 on MuPoTS-3D and 60.1 MOTA on PoseTrack 2018 validation, outperforming prior tracking methods by 11.7% on MOTA (Kumarapu et al., 2020).

An egocentric avatar reconstruction interpretation is provided by EgoAnimate. That system uses a single head-mounted camera recording a roughly $P_s^n \in \mathbb{R}^{F \times J \times 2}$3 top-down view, then applies a Stable Diffusion v1.5 backbone with ControlNet and a SMPL segmentation scaffold to convert a single top-down image into a frontal T-pose representation. The result is then passed either to MagicMan + ExAvatar for a Gaussian-splat avatar or to UniAnimate for direct video synthesis (Türkoglu et al., 12 Jul 2025). The model is described as the first study using a generative backbone to reconstruct animatable avatars from egocentric inputs, and its held-out evaluation reports PSNR $P_s^n \in \mathbb{R}^{F \times J \times 2}$4, SSIM $P_s^n \in \mathbb{R}^{F \times J \times 2}$5, and LPIPS $P_s^n \in \mathbb{R}^{F \times J \times 2}$6 for full-body frontal synthesis (Türkoglu et al., 12 Jul 2025).

A dynamic-background diffusion interpretation appears in AniCrafter. There, human-centric animation is reframed as refining a low-frequency avatar-background composite, using a latent I2V diffusion backbone conditioned by a rendered avatar video, a reference image, a soft mask, and 3D-conv features from avatar and SMPL-X streams (Niu et al., 26 May 2025). The reported dynamic-background test on 100 videos gives SSIM $P_s^n \in \mathbb{R}^{F \times J \times 2}$7, PSNR $P_s^n \in \mathbb{R}^{F \times J \times 2}$8, LPIPS $P_s^n \in \mathbb{R}^{F \times J \times 2}$9, FID $G(P_c^m, P_s^n) = \hat P_s^m,$0, and FVD $G(P_c^m, P_s^n) = \hat P_s^m,$1 (Niu et al., 26 May 2025).

A multi-character diffusion interpretation is given by MultiAnimate, where PersonaAnimator is described as a conditional latent diffusion system with a DiT backbone, an Identifier Assigner that converts masks $G(P_c^m, P_s^n) = \hat P_s^m,$2 into a one-hot label tensor, and an Identifier Adapter based on a 3D-conv weight bank that produces time-aware identifier embeddings (Hu et al., 25 Feb 2026). The framework is trained with randomized label assignment over a bank of size $G(P_c^m, P_s^n) = \hat P_s^m,$3 so that a model trained on only two-character data can generalize to more characters than seen during training (Hu et al., 25 Feb 2026).

Taken together, these adjacent uses show that PersonaAnimator has become a convenient name for systems that combine identity conditioning, motion control, and animation rendering, but they differ fundamentally in input modality, representation space, and output type. The 2025 V2V-MP method operates on 2D skeletons extracted from ordinary videos (Qian et al., 27 Aug 2025); the surrounding ecosystem extends the name toward 3D pose pipelines, egocentric reconstruction, open-domain video diffusion, and multi-character synthesis.