TextEncodeMingImageEdit - ComfyUI Built-in Node Documentation - ComfyUI

Text Encode Ming Image Edit encodes a text prompt into conditioning for Ming image editing, optionally mixing in reference images. The prompt and the reference images are tokenized by a CLIP model, and when a VAE is connected the reference images are also encoded into latent frames that are appended to the conditioning.

Inputs

Parameter	Description	Data Type	Required	Range
`clip`	The CLIP model used to tokenize the prompt and the reference images.	CLIP	Yes	-
`vae`	Optional VAE that encodes the reference images into latent frames appended to the conditioning. Without a VAE the images only condition the text encoder through the vision tower.	VAE	No	-
`prompt`	Text prompt to encode. Supports multiline input and dynamic prompts.	STRING	Yes	Multiline text
`images`	Growable slot: optional reference images seen by the text encoder and appended to the latent sequence as clean frames. Connect 1..8 images (`image_1`, `image_2`, …); later images are resized to the first one and the sampled latent should match its size. Only the RGB channels are used.	IMAGE	No	0 to 8

Note: Reference images are read in numeric order of their slot names, and empty slots are ignored. Reference latents are only produced when both vae and at least one image are provided.

Outputs

Output Name	Description	Data Type
`CONDITIONING`	Conditioning containing the encoded prompt, plus the reference latents when a VAE and reference images are provided.	CONDITIONING

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): 675fb3cc0af006e1284fdb2a5ca2c268c540ee92e1ede90b2359f4e3fc8ba5ea