Update the ComfyUI-WanVideoWrapper node to the latest
SteadyDancer is a fine-tuned 14B parameter variant of Wan 2.1 I2V, optimized for flicker-free, identity-preserving image-to-video generation.
Perfect for creating coherent dance animations from static human photos, keeps faces and poses locked across frames with SOTA temporal consistency.
FP16 precision, 480p resolution generates 5-10s clips (81 frames @ 24 FPS). Requires ComfyUI with WanVideoWrapper nodes, Trained on dance datasets for fluid motions;
Example prompt: "steady dancer twirling gracefully, no drift."
Apache 2.0 license
The model is mirrored here for convenience.