InstructPixToPixConditioning - ComfyUI Built-in Node Documentation

The InstructPixToPixConditioning node prepares conditioning data for InstructPix2Pix image editing by combining positive and negative text prompts with image data. It encodes the input image through a VAE into a latent representation and attaches that latent to both the positive and negative conditioning, while also returning a matching empty latent. Image dimensions are automatically cropped to multiples of 8 pixels so the VAE can process them.

Inputs

Parameter	Description	Data Type	Required	Range
`positive`	Positive conditioning data containing text prompts and settings for desired image characteristics	CONDITIONING	Yes	-
`negative`	Negative conditioning data containing text prompts and settings for undesired image characteristics	CONDITIONING	Yes	-
`vae`	VAE model used for encoding input images into latent representations	VAE	Yes	-
`pixels`	Input image to be processed and encoded into latent space	IMAGE	Yes	-

Note: The input image dimensions are automatically adjusted by center-cropping to multiples of 8 pixels in both width and height to ensure compatibility with the VAE encoding process.

Outputs

Output Name	Description	Data Type
`positive`	Positive conditioning data with the encoded image latent attached as `concat_latent_image`	CONDITIONING
`negative`	Negative conditioning data with the encoded image latent attached as `concat_latent_image`	CONDITIONING
`latent`	Latent tensor of zeros with the same dimensions as the encoded image	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): e9a5a05cdeafe9337ca2033111f1ad4f7314fa33d71a4764f62919857efc79f4

​Inputs

​Outputs

Inputs

Outputs