UltraTex Conditioning - ComfyUI Built-in Node Documentation

Conditions FLUX.2 Klein 9B with the UltraTex LoRA to paint a reference image’s look onto a mesh, as one image per view. Render the views with Render Mesh in world normal mode, from a Create Camera Orbit of 4 plus top and bottom, and bake the result with Bake Texture From Views.

Inputs

Parameter	Description	Data Type	Required	Range
`conditioning`	Text conditioning. UltraTex was trained with an empty prompt.	CONDITIONING	Yes	-
`vae`	VAE used to encode the normals, the views and the reference.	VAE	Yes	-
`normals`	Render Mesh `world normal` images: the 6 canonical views, or the 4 side views.	IMAGE	Yes	-
`masks`	Render Mesh masks of the same views, one per view.	MASK	Yes	-
`reference`	The object as it should look, seen from the front camera’s side.	IMAGE	Yes	-
`reference_mask`	Object mask of the reference, white for the object. Default: fully white.	MASK	No	-
`extra_views`	More references: the object seen from the other views, each tied to its view as the reference is tied to the front. Default: `"off"`.	DYNAMIC_COMBO	Yes	`"off"` `"on"`

Extra Views “on” Inputs

Parameter	Description	Data Type	Required	Range
`right`	The object seen from the right view.	IMAGE	No	-
`back`	The object seen from the back view.	IMAGE	No	-
`left`	The object seen from the left view.	IMAGE	No	-
`top`	The object seen from the top view.	IMAGE	No	-
`bottom`	The object seen from the bottom view.	IMAGE	No	-

Outputs

Output Name	Description	Data Type
`conditioning`	Conditioning carrying the view grid, the reference latents and the token masks UltraTex needs.	CONDITIONING

Note: normals must hold either the 6 canonical views (Create Camera Orbit with count 4 and poles top and bottom) or the 4 side views, and masks must match its view count.

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): 5b05f5151d4c12394ef893db69538adaef86e9cbdbb118a37c20919758520b6f