TextEncodeZImageOmni - ComfyUI Built-in Node Documentation

Complete documentation for the TextEncodeZImageOmni 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 TextEncodeZImageOmni node is an advanced conditioning node that encodes a text prompt along with optional reference images into a conditioning format suitable for image generation models. It can process up to three images, optionally encoding them with a vision encoder and/or a VAE to produce reference latents, and integrates these visual references with the text prompt using a specific template structure.

Inputs

Parameter	Data Type	Required	Range	Description
clip	CLIP	Yes		The CLIP model used for tokenizing and encoding the text prompt.
image_encoder	CLIPVision	No		An optional vision encoder model. If provided, it will be used to encode the input images, and the resulting embeddings will be added to the conditioning.
prompt	STRING	Yes		The text prompt to be encoded. This field supports multiline input and dynamic prompts.
auto_resize_images	BOOLEAN	No		When enabled (default: True), input images will be automatically resized based on their pixel area before being passed to the VAE for encoding.
vae	VAE	No		An optional VAE model. If provided, it will be used to encode the input images into latent representations, which are added to the conditioning as reference latents.
image1	IMAGE	No		The first optional reference image.
image2	IMAGE	No		The second optional reference image.
image3	IMAGE	No		The third optional reference image.

Note: The node can accept a maximum of three images (image1, image2, image3). The image_encoder and vae inputs are only utilized if at least one image is provided. When auto_resize_images is True and a vae is connected, images are resized to have a total pixel area close to 1024x1024 before encoding.

Outputs

Output Name	Data Type	Description
CONDITIONING	CONDITIONING	The final conditioning output, which contains the encoded text prompt and may include encoded image embeddings and/or reference latents if images were provided.