WanImageToVideo - ComfyUI Built-in Node Documentation

Complete documentation for the WanImageToVideo node in ComfyUI. Learn its inputs, outputs, parameters and usage.

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

The WanImageToVideo node prepares conditioning and latent representations for video generation tasks. It creates an empty latent space for video generation and can optionally incorporate starting images and CLIP vision outputs to guide the video generation process. The node modifies both positive and negative conditioning inputs based on the provided image and vision data.

Inputs

Parameter	Data Type	Required	Range	Description
positive	CONDITIONING	Yes	-	Positive conditioning input for guiding the generation
negative	CONDITIONING	Yes	-	Negative conditioning input for guiding the generation
vae	VAE	Yes	-	VAE model for encoding images to latent space
width	INT	Yes	16 to MAX_RESOLUTION	Width of the output video (default: 832, step: 16)
height	INT	Yes	16 to MAX_RESOLUTION	Height of the output video (default: 480, step: 16)
length	INT	Yes	1 to MAX_RESOLUTION	Number of frames in the video (default: 81, step: 4)
batch_size	INT	Yes	1 to 4096	Number of videos to generate in a batch (default: 1)
clip_vision_output	CLIP_VISION_OUTPUT	No	-	Optional CLIP vision output for additional conditioning
start_image	IMAGE	No	-	Optional starting image to initialize the video generation

Note: When start_image is provided, the node encodes the image sequence and applies masking to the conditioning inputs. The clip_vision_output parameter, when provided, adds vision-based conditioning to both positive and negative inputs.

Outputs

Output Name	Data Type	Description
positive	CONDITIONING	Modified positive conditioning with image and vision data incorporated
negative	CONDITIONING	Modified negative conditioning with image and vision data incorporated
latent	LATENT	Empty latent space tensor ready for video generation