Human Image Animation

Benchmarks

Dataset Name	SOTA Method	Metric
RealisDance (val)	Moore-AnimateAnyone	Subject Consistency94.65	27	25d ago
TikTok	StableAnimator	FVD140.62	15	2mo ago
TikTok (test)	Position Shift Adaptive Module	FVD172.16	15	2mo ago
TikTok (sequences 335 to 340)	Implicit Preference Alignment	FID-VID5.9	10	2mo ago
Curated (test)	DynamiCrafter	CPBD0.8333	9	2mo ago
Unseen100	StableAnimator	L1 Loss27,100	9	4mo ago
Unseen	DirectAnimator	FID27.62	8	1mo ago
X-Dance (test)	SteadyDancer	Subject Quality (I2V)96.17	8	2mo ago
TikTok dataset 14 (test)	MagicAnimate	L1 Loss0.0003	8	4mo ago
TED-talks (test)	TPS	AKD2.16	8	4mo ago
User Study 30 selected videos	StableAnimator	M-A95.6	7	4mo ago
Authors' Hand Dynamics Benchmark 100 curated cases	Implicit Preference Alignment	SSIM (Hand)60.6	6	2mo ago
Custom Challenging Benchmark 100 curated cases with complex hand dynamics 1.0 (test)	Implicit Preference Alignment (Ours)	FID-VID6.3	6	2mo ago
HyperMotionX Bench	Hypermotion	PSNR22.03	6	4mo ago
User Study 50 participants 5-point Mean Opinion Score	Position Shift Adaptive Module	VC Score4	5	4mo ago
Our dataset (test)	Position Shift Adaptive Module	Subject Consistency92.54	5	4mo ago
TikTok and internet videos (test)	MTVCrafter	SSIM0.7489	5	4mo ago
In-the-wild cross-driving (test)	X-Dyna	DTFVD5.923	5	4mo ago
Single-character benchmark	EverybodyDance	SSIM0.772	4	4mo ago
Multi-character benchmark	EverybodyDance	SSIM0.671	4	4mo ago
Human Image Animation 512x512, 49 frames	AnimateAnyone	Preprocessing Time (ms)9.6	3	1mo ago
Human Preference Study Evaluation Set	Implicit Preference Alignment	Win Rate (%)91.7	3	2mo ago

Showing 22 of 22 rows