BREAKING
Nanjing U & Tencent release SteadyDancer
0
B
parameters
0
released
I2V replaces the R2V approach
R2V
old
●
Spatial mismatches
●
Identity drift
I2V
SteadyDancer
●
First-frame anchor
●
Preserves identity
How SteadyDancer works
1
Anchor first frame
↓
2
Track pose
↓
3
Generate video
Multiple distribution formats
Standard
14B
●
Hugging Face weights
Quantized
GGUF
●
Lightweight inference
For China
14B
●
ModelScope
ComfyUI support still maturing
AI NEWS BLITZ
Nanjing University and Tencent have unveiled a new human animation model called SteadyDancer.