TextEncodeMOSSImageEdit - ComfyUI Built-in Node Documentation

Encodes an edit instruction together with reference images into conditioning for a MOSS-Image edit LoRA. Connect the output as the positive conditioning and run a second instance with the same references for the negative one. The references are used at the size they arrive, so scale them before connecting. Without a VAE, the images only go to the text encoder. This node is marked as experimental in ComfyUI.

Inputs

Parameter	Description	Data Type	Required	Range
`clip`	The CLIP model that tokenizes the instruction and encodes it.	CLIP	Yes	-
`prompt`	The instruction describing the edit. Supports multiline text and dynamic prompts.	STRING	Yes	Any text
`vae`	Encodes the images as reference latents for the edit LoRA. Optional: without it, the images only go to the text encoder.	VAE	No	-
`edit_format`	The layout the edit LoRA was trained with. It selects the reference latents method and whether the text encoder also sees the references.	COMBO	Yes	`"ai-toolkit"` `"ai-toolkit (kv_cache off)"` `"Moss-Image-Edit"`
`images`	Growable slot: connect one or more reference images (`image_1`, `image_2`, …). Each image is used at the size it arrives. When a VAE is connected, alpha is dropped and each image is VAE-encoded into a reference latent. Without a VAE, the images only go to the text encoder.	IMAGE	Yes	1 or more images
`image_strength`	Scales how much the model attends to the images the text encoder sees (ai-toolkit formats, or no VAE): 0 ignores them, 1 is unchanged, above 1 emphasizes them. Advanced. Default: 1.0.	FLOAT	Yes	0.0 to 100.0 (step 0.05)

Note: With "ai-toolkit" and "ai-toolkit (kv_cache off)", or when no VAE is connected and image_strength is above 0, the text encoder also sees every reference, downscaled to at most 384x384 pixels and preceded by a Picture N: marker in the prompt, so the instruction can refer to them by number. With "Moss-Image-Edit" and a VAE connected, the references reach the model only as reference latents and the text encoder sees the instruction alone. "ai-toolkit (kv_cache off)" additionally keeps the references attending to the image being generated, so they are recomputed on every sampling step.

Outputs

Output Name	Description	Data Type
`conditioning`	The encoded instruction with the reference latents and the matching reference latents method attached.	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): 673f098df6a8e466a5c8060009f446eac8f7556311a870de626c5f4bfc3ab196

​Inputs

​Outputs

Inputs

Outputs