Multi-Positive Sampler Node with Controlled Interaction (Selective Cross Conditioning)

### Feature Idea

chat gpt said this should be doable, but i haven't gotten it a custom text encode to work so....

I’d love a sampler node that accepts multiple positive conditionings (e.g. different characters or styles) without mixing them too early — something like a “multi-positive KSampler.”

The idea is to keep conditionings separate for some steps (to form distinct subjects or regions), and then allow them to interact in later steps (to unify lighting or composition).

This could let users generate multiple distinct characters or art styles in the same image — for example, one realistic and one cartoon — while keeping them coherent, like a “Space Jam” aesthetic.

Ideal features:

Multiple positive Cond inputs (A, B, etc.)

Optional masks or regions per input

Control over which diffusion steps are “separate” vs. “shared”

Optional “interaction strength” or “overlap” slider

Possible base: modify the existing KSampler or Regional Conditioning Sampler.

Would be a huge help for compositional and mixed-style generations!

### Existing Solutions

i know there are regional prompt options, but. imma just say it, im making porn.
characters don't like to keep their tags separate.
i dont really know much about anything so good luck if this gives you any kind of ideas.

### Other

_No response_