VOIDWarpedNoise - ComfyUI Built-in Node Documentation

Generates temporally-correlated noise for the second pass of the VOID video refinement process.

Generates temporally-correlated noise for the second pass of the VOID video refinement process. It takes the Pass 1 output video and warps Gaussian noise along optical flow vectors, so the noise moves consistently with the video content. The resulting warped noise is used as the starting latent for Pass 2, which improves temporal consistency in the final output.

Inputs

Parameter	Description	Data Type	Required	Range
optical_flow	Optical flow model from OpticalFlowLoader (RAFT-large).	OPTICAL_FLOW	Yes	-
video	Pass 1 output video frames [T, H, W, 3].	IMAGE	Yes	-
width	Target width in pixels (default: 672). The input video is scaled to this width before the noise is generated, and the latent width is derived as width ÷ 8.	INT	Yes	16 to MAX_RESOLUTION (step 8)
height	Target height in pixels (default: 384). The input video is scaled to this height before the noise is generated, and the latent height is derived as height ÷ 8.	INT	Yes	16 to MAX_RESOLUTION (step 8)
length	Number of pixel frames. Rounded down to make latent_t even (patch_size_t=2 requirement), e.g. 49 to 45 (default: 45).	INT	Yes	1 to MAX_RESOLUTION (step 1)
batch_size	Number of identical warped noise sequences to produce (default: 1). The generated noise is repeated this many times along the batch dimension.	INT	Yes	1 to 64

Note on the length parameter: The length value is automatically rounded down to the nearest value that produces an even latent_t dimension, as required by the CogVideoX-Fun-V1.5 model’s patch_size_t=2 constraint (for example, 49 becomes 45). The node logs a warning when this rounding occurs. Frames beyond the adjusted length are ignored, and the noise is resampled to the resulting latent frame count.

Note on width and height: These values are used both to resize the incoming video frames (bilinear, center crop) and to determine the final latent resolution (divided by 8). If the generated noise does not match the requested latent size, it is resized to fit.

Outputs

Output Name	Description	Data Type
warped_noise	A 5D tensor (B, C, T, H, W) containing optical-flow warped Gaussian noise, ready for use as the initial latent in VOID Pass 2.	LATENT

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

---

Source fingerprint (SHA-256): f46b0a73b09a5d2d0bc25676f9571563c6bb8bad8d835e7564ac092c72136107