Apply FreeLong spectral blending to Wan 2.2 for better motion consistency in long videos. Connect between model loader and KSampler. — ComfyUI node from…

Custom nodes by IAMCCS, including a critical fix for LoRA loading in native WANAnimate workflows, especially useful for advanced models like FLUX and WAN 2.1. This bypasses the need for the WanVideoWrapper for LoRA functionality.

IAMCCS-nodes

model

enabled

blend_strength

low_freq_ratio

auto_window

manual_window_frames

blend_curve

blend_start_block

blend_end_block

vram_preset

Apply FreeLong spectral blending to Wan 2.2 for better motion consistency in long videos. Connect between model loader and KSampler.

IAMCCS_MoveAhead

MoveAhead (FreeLong spectral blend) ??

Name	Type	Default	Description
model	MODEL		—
enabled	BOOLEAN	true	Enable / disable MoveAhead. Leave connected and toggle for quick A/B.
blend_strength	FLOAT	0.80–1	How strongly the spectral blend affects the output. 0 = passthrough, 1 = full blend. 0.8 is a safe default.
low_freq_ratio	FLOAT	0.800–1	Fraction of the frequency spectrum taken from the global (full-sequence) stream. Higher → smoother motion, less detail. Lower → sharper but may drift. 0.8 is a good balance for most scenarios.
auto_window	BOOLEAN	true	AUTO window mode. When enabled, MoveAhead auto-detects latent frames and uses a window of ~40% of the detected length.
manual_window_frames	STRING		Manual window size in VIDEO frames (example: 33 for 81-frame videos). Used only if AUTO window is OFF.
blend_curve	COMBO	flat	flat = constant blend_strength across all blocks (original behaviour). sigmoid = strength ramps from near-zero at the first blocks to full blend_strength at later blocks. Early blocks keep global motion anchoring, later blocks progressively add detail blending.
blend_start_block	INT	00–40	Transformer block at which blending starts (0 = first block).
blend_end_block	INT	-1-1–40	Transformer block at which blending ends (-1 = all blocks, respects actual model depth).
vram_preset	COMBO	normal	VRAM preset for the FFT blending implementation. low = smallest chunks (slowest, safest), normal = balanced, high = larger chunks (faster).

MoveAhead (FreeLong spectral blend) ??

Inputs (10)

Outputs (1)