VegaNec Interaction Director is a positive-direction creator utility LoRA for
Krea 2. It is designed to make exactly two fictional adults visibly relate
through aligned attention, orientation, spacing, reciprocal action, contact,
and one shared relation, target, object, or task.
This is not a couple-identity model and does not create fixed people. It is a
composition-support tool for scenes where two attractive subjects still look
like independent portraits pasted into the same frame.
The model supports:
- mutual attention between exactly two adults;
- shared attention toward one visible or implied target;
- one-object handoffs with a readable giver and receiver;
- coordinated practical tasks with complementary roles;
- conversational orientation with visible speaker and listener cues;
- consent-forward greeting or supportive contact;
- woman-man, woman-woman, and man-man pairings;
- adult ages spanning young, middle-aged, and mature presentation;
- waist-up, three-quarter, full-body, and environmental framing;
- varied skin tones, body classes, wardrobes, professions, settings, and
lighting.
Trigger: veganec_interaction_director
Recommended strength: 0.8
Useful range: 0.8 release-validated; other strengths unverified
For best results, state exactly two adults and give each one a distinct visible
identity, side, role, and action. Then define one shared relation:
exactly two fictional adults, left adult gives one sealed sample case with both hands, right adult receives the same case with both hands, readable giver-to-receiver direction
For shared attention, name one target and require both eye lines to converge:
both adults visibly attend to the same single evacuation route on one wall map while remaining distinct subjects
For coordinated work, name one object, opposite ownership points, and the
complementary actions. Avoid vague prompts such as “two people interacting,”
which do not define the relation the image must show.
Interaction Director does not claim fixed couple identity, arbitrary
multi-character composition, groups of three or more, general romance or
chemistry enhancement, perfect hands, perfect anatomy, regional prompting, or
ControlNet replacement. It can still inherit normal generation failures in
subject count, identity separation, contact, hands, anatomy, object ownership,
and spatial composition.
The exact epoch-10 adapter at strength 0.8 passed six fixed unseen,
identical-seed comparisons covering mutual attention, shared attention, object
handoff, coordinated carrying, speaker-listener conversation, and a reciprocal
handshake. Fourteen additional owner-approved launch scenes then demonstrated
broader pairings, identities, environments, role directions, shared objects,
tasks, conversations, and contact without a recurring shortcut. The larger
twenty-four-holdout/four-strength statistical study remains unperformed and is
not claimed.
Every gallery subject is a fictional adult aged 21 or older. The model is not
based on actual people.

