Nodes/ComfyUI Smart Helper Nodes/Smart Mobius WAN Loop
ComfyUI Node

Smart Mobius WAN Loop

Mobius seamless loop for WAN 2.2 dual-pass (high/low noise). Three independent mechanisms: (1) Latent shift — cycles frame positions across denoising steps to distribute RoPE bias. (2) Loop blend — corrects end-frame noise by blending toward start-frame predictions. (3) Circular RoPE — compresses temporal positions so the last frame is closer to the first in attention space.

By slvslvslv·Created 2 years ago·Updated 2 months ago· 3
Smart Mobius WAN Loop
  • model_high
  • model_low
  • positive
  • negative
  • vae
  • start_image
  • end_image
  • clip_vision_start_image
  • clip_vision_end_image
  • model_high
  • model_low
  • positive
  • negative
  • latent
width832
height480
length81
batch_size1
motion_amplitude1.00
enable_latent_shifttrue
shift_skip0
enable_loop_blendtrue
blend_frames5
blend_strength0.50
enable_circular_ropetrue
roll_concat_conditioningfalse
circular_rope_strength0.50
CategorySmartHelperNodes/Wan

Inputs (22)

NameTypeDefaultDescription
model_highMODEL
model_lowMODEL
positiveCONDITIONING
negativeCONDITIONING
vaeVAE
widthINT83216–8192
heightINT48016–8192
lengthINT811–8192
batch_sizeINT11–4096
start_imageoptIMAGE
end_imageoptIMAGE
clip_vision_start_imageoptCLIP_VISION_OUTPUT
clip_vision_end_imageoptCLIP_VISION_OUTPUT
motion_amplitudeoptFLOAT1.001–21.0 = official FLF, 2.0 = max structural repulsion boost
enable_latent_shiftoptBOOLEANtrueMobius-style circular latent rolling each denoising step
shift_skipoptINT00–100Roll stride per step. 0 = auto (coprime with latent frame count)
enable_loop_blendoptBOOLEANtrueBlend end-frame predictions toward start-frame predictions after CFG
blend_framesoptINT51–20Number of end frames to correct via boundary blending
blend_strengthoptFLOAT0.500–10 = no correction, 1 = fully replace end with start predictions
enable_circular_ropeoptBOOLEANtrueCompress temporal RoPE so last frame is closer to first
roll_concat_conditioningoptBOOLEANfalseExperimental. Roll concat_mask/concat_latent_image together with the latent cycle.
circular_rope_strengthoptFLOAT0.500–10 = no change, 1 = maximum temporal compression

Outputs (5)

NameTypeDescription
model_highMODEL
model_lowMODEL
positiveCONDITIONING
negativeCONDITIONING
latentLATENT