Nodes/ComfyUI-SD-CN-Animation/SD-CN Animation Vid2Vid
ComfyUI Node

SD-CN Animation Vid2Vid

Stylizes input video frames using RAFT optical flow for temporal coherence with two-pass SD sampling.

By pxl-pshr·Created 4 months ago·Updated 4 months ago· 10
SD-CN Animation Vid2Vid
  • model
  • vae
  • positive
  • negative
  • raft_model
  • frames
  • control_net
  • frames
seed0
steps15
cfg5.5
sampler_name
scheduler
processing_strength0.85
fix_frame_strength0.15
blend_alpha1.00
occlusion_mask_blur3.0
occlusion_mask_flow_multiplier5.0
occlusion_mask_difo_multiplier2.0
occlusion_mask_difs_multiplier0.0
occlusion_mask_trailingfalse
cn_strength1.00
CategorySD-CN-Animation

Inputs (21)

NameTypeDefaultDescription
modelMODEL
vaeVAE
positiveCONDITIONING
negativeCONDITIONING
raft_modelRAFT_MODEL
framesIMAGE
seedINT00–18446744073709550000
stepsINT151–200
cfgFLOAT5.50–100
sampler_nameCOMBO44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38
schedulerCOMBO9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3
processing_strengthFLOAT0.850–1Denoising strength for main stylization pass
fix_frame_strengthFLOAT0.150–1Denoising strength for refinement pass (0 to skip)
blend_alphaFLOAT1.000–1Blend between warped styled frame (0) and current input frame (1)
occlusion_mask_blurFLOAT3.00–10Gaussian blur applied to occlusion mask
occlusion_mask_flow_multiplierFLOAT5.00–10Weight for flow-based occlusion detection
occlusion_mask_difo_multiplierFLOAT2.00–10Weight for original frame difference occlusion
occlusion_mask_difs_multiplierFLOAT0.00–10Weight for styled frame difference occlusion
occlusion_mask_trailingBOOLEANfalseBlend current occlusion with previous to reduce ghosting
control_netoptCONTROL_NET
cn_strengthoptFLOAT1.000–10

Outputs (1)

NameTypeDescription
framesIMAGE