Character To Image

Every node in this pack exists to answer one question - but CP_CharacterToImage exists to answer the dumbest, most useful one: "wait, where did the image go?" A CHARACTER object bundles the sprite, palette, embedding, and metadata into one opaque type. It doesn't look like an image, and you can't drag it into an IP-Adapter. This node just unpacks the stored reference image back into a plain IMAGE tensor.

How it works

Nothing clever here. The CHARACTER holds its reference image as a float32 array; this node wraps it back into a ComfyUI image tensor. That's the entire mechanism - which is exactly why you'll quietly be grateful it exists when you need the pixels for something that only speaks IMAGE.

The one input and output

• character (required) - any CHARACTER from CP_CharacterEncode or CP_LoadCharacter.
• image - the reference image, out.

The classic use is wiring it into an IP-Adapter node's reference input, or just previewing what's actually stored in a character object you loaded from disk - useful when you saved a .char last month and can't remember which hero it was.

Installing

Same single pack as the rest of the CharacterPose nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/Spit8/_ComfyUI_CharacterPose
pip install -r _ComfyUI_CharacterPose/requirements.txt

Restart ComfyUI; it's under CharacterPose/Character. No dependencies beyond the pack's core ones.

Gotchas

• The image comes back at whatever resolution it was encoded at. If you need it at a specific size, run it through an upscale/scale node after.
• If you pass a character with an empty reference (shouldn't happen via the normal encode path, but possible if you hand-built a CHARACTER), you get a blank image. Encode through CP_CharacterEncode and you're fine.

That's it - a three-line node doing a three-line job. Boring, dependable, and it'll save you from re-encoding a character just to look at it.