Nodes/ComfyUI-FLOAT_Optimized/Apply FLOAT Synthesis (VA)
ComfyUI Node

Apply FLOAT Synthesis (VA)

The final image generation step. It takes the bundled appearance latents (from the Appearance Pipe) and the driven motion sequence (r_d), then uses the loaded Synthesis/Decoder model to render the final animated image sequence frame by frame.

By set-soft·Created about a year ago·Updated 12 months ago· 33
Apply FLOAT Synthesis (VA)
  • appearance_pipe
  • float_synthesis
  • r_d_latents
  • images
  • float_synthesis_out
CategoryFLOAT/Very Advanced

Inputs (3)

NameTypeDefaultDescription
appearance_pipeFLOAT_APPEARANCE_PIPEThe bundled appearance information (s_r latent and feature maps) from the ApplyFloatEncoder node. (Ws→r)
float_synthesisFLOAT_SYNTHESIS_MODELThe loaded FLOAT Synthesis (Decoder) model module.
r_d_latentsTORCH_TENSORThe driven motion latent sequence generated by the FMT sampler. (Wr→D)

Outputs (2)

NameTypeDescription
imagesIMAGE
float_synthesis_outFLOAT_SYNTHESIS_MODEL