PH3ImageConditioning - ComfyUI Built-in Node Documentation

This node turns a prompt into conditioning and an empty latent for a single still image with p-h3-image. Without images it is text to image; with images connected, image_1 is the image being edited and the other slots act as references, which the prompt addresses as <Picture 1>, <Picture 2>, and so on. The sampler needs a model patched by the MiniMax H3 sigma shift node with both shift values at 1.0.

Inputs

Parameter	Description	Data Type	Required	Range
`clip`	The p-h3-image CLIP model used to tokenize the prompt, including any reference images, and to encode the conditioning.	CLIP	Yes	-
`prompt`	What to generate, or how to edit `image_1`. With images connected, a short instruction gets `Edit <Picture 1>.` prepended. An instruction that already starts with `Edit <Picture` or an edit brief produced by a prompt upsampler is used exactly as written.	STRING	Yes	-
`width`	Output width in pixels. 0 uses the width of `image_1` after scaling, or 2912 when no image is connected (default: 0).	INT	Yes	0 to MAX_RESOLUTION
`height`	Output height in pixels. 0 uses the height of `image_1` after scaling, or 2912 when no image is connected (default: 0).	INT	Yes	0 to MAX_RESOLUTION
`vae`	The p-h3-image VAE. Required as soon as an image slot is connected, because the reference latents are encoded with it.	VAE	No	-
`images`	Growable group of up to 9 images, `image_1` through `image_9`. `image_1` is the image being edited, the remaining slots are references.	AUTOGROW	No	0 to 9 images

Outputs

Output Name	Description	Data Type
`positive`	Conditioning for the p-h3-image sampler. When images are connected it also carries the packed reference latents.	CONDITIONING
`latent`	Empty latent for one still frame, including the two audio latents the model expects.	LATENT

Note: Images require the vae input. Running the node with an image connected and no VAE raises an error. Note: The conditioning covers a single frame. For video or audio, use the MiniMax H3 video nodes instead.

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): f1f5e5b2d3ff9bc2f3b81f3d0dd722e725c7930c6f95bf44ad83bc64357d2fb7

​Inputs

​Outputs

Inputs

Outputs