Summary

Adds support for passing reference images to Ideogram4 DiT.

This is not officially supported, the model was likely not trained for this use case, but it can work for very basic editing tasks.

Related Issue / Discussion

#1679

(inspired by)
Comfy-Org/ComfyUI#14567

Additional Information

prompt:
{"high_level_description":"Johnny Silverhand from Cyberpunk 2077 stands outdoors, adjusting his plastic sunglasses with his prosthetic left hand under a bright blue sky.","compositional_deconstruction":{"background":"A bright, slightly hazy blue sky with thin clouds and a blurred, tall metallic power line tower in the distant background. Soft, warm sunlight streams in from the left side of the frame.","elements":[{"type":"obj","bbox":[41,392,481,610],"desc":"The head of Johnny Silverhand, a male with long, pink, curly hair parted down the middle, a trimmed pink beard, and pink mustache."},{"type":"obj","bbox":[227,458,356,583],"desc":"Pastel blue aviator sunglasses with a thick plastic frame, worn on his face."},{"type":"obj","bbox":[257,581,947,756],"desc":"A highly detailed, wooden prosthetic left arm and hand with light wood grain and visible ball joints, raised to touch his sunglasses."},{"type":"obj","bbox":[418,315,966,610],"desc":"A latex, bright pink tactical vest with thick shoulder straps and reflective, mirror panels, worn over his torso."},{"type":"obj","bbox":[439,181,775,357],"desc":"The bare, muscular right arm of the man, visible from the shoulder down to the mid forearm."}]}}

ref:

output:

Checklist

• I have read and confirmed this PR follows the contribution guidelines.