ComfyUI Node
Mpi P-H3 Image Conditioning
P-H3 Image conditioning with all 8 picture slots exposed at once: image_1 is the canvas (the picture edited; the output keeps its size), image_2..8 are references. Empty slots are dropped and the rest renumbered, so ONE graph covers text-to-image and every edit combination instead of growing core's Autogrow input per run. Empty means nothing connected or an Mpi loader with block_if_empty OFF (a 1x1 image) - a real black image passes through. Write the prompt against SLOT numbers - <Picture 3> means image_3 - and the node rewrites them to what core presents, dropping any tag whose slot is empty. Conditioning is built by core's own PH3ImageConditioning (ComfyUI 0.44.0+), so it cannot drift.
Mpi P-H3 Image Conditioning
- clip
- vae
- image_1
- image_2
- image_3
- image_4
- image_5
- image_6
- image_7
- image_8
- positive
- latent
- ref_tags
◄prompt—►
◄width0►
◄height0►
◄edit_max_megapixels0.0►
CategoryMpiNodes/Utils
Inputs (14)
| Name | Type | Default | Description |
|---|---|---|---|
| clip | CLIP | — | |
| vae | VAE | — | |
| prompt | STRING | Cite pictures by SLOT: <Picture 3> is whatever is wired into image_3. The node rewrites tags to the order core presents once empty slots are dropped (see ref_tags); a tag naming an empty slot is removed. With pictures, a short instruction gets "Edit <Picture 1>. " in front (core); one starting "Edit <Picture" or an upsampler brief is used as it is. | |
| width | INT | 00–16384 | 0 uses the canvas width after scaling, or 2912 without pictures. |
| height | INT | 00–16384 | 0 uses the canvas height after scaling, or 2912 without pictures. |
| image_1opt | IMAGE | The CANVAS: the picture being edited; the output keeps its size and aspect. Cited in the prompt as <Picture 1>. Leave unconnected, or feed a loader with block_if_empty OFF, to skip it. | |
| image_2opt | IMAGE | Reference picture. Cited in the prompt as <Picture 2>. Leave unconnected, or feed a loader with block_if_empty OFF, to skip it. | |
| image_3opt | IMAGE | Reference picture. Cited in the prompt as <Picture 3>. Leave unconnected, or feed a loader with block_if_empty OFF, to skip it. | |
| image_4opt | IMAGE | Reference picture. Cited in the prompt as <Picture 4>. Leave unconnected, or feed a loader with block_if_empty OFF, to skip it. | |
| image_5opt | IMAGE | Reference picture. Cited in the prompt as <Picture 5>. Leave unconnected, or feed a loader with block_if_empty OFF, to skip it. | |
| image_6opt | IMAGE | Reference picture. Cited in the prompt as <Picture 6>. Leave unconnected, or feed a loader with block_if_empty OFF, to skip it. | |
| image_7opt | IMAGE | Reference picture. Cited in the prompt as <Picture 7>. Leave unconnected, or feed a loader with block_if_empty OFF, to skip it. | |
| image_8opt | IMAGE | Reference picture. Cited in the prompt as <Picture 8>. Leave unconnected, or feed a loader with block_if_empty OFF, to skip it. | |
| edit_max_megapixelsopt | FLOAT | 0.00–64 | With pictures (an edit), the output never goes above this many megapixels: width/height, or picture 1's own size when they are 0, is scaled down to it, aspect kept. Never scaled up. 0 = no cap. Text-to-image ignores it. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| positive | CONDITIONING | — |
| latent | LATENT | — |
| ref_tags | STRING | — |