Avatar Set · Semi-realistic · AI Lab

Black Tailored Military — Avatar Set

Male of the pair — matching matte-black tailoring, gunmetal hardware, blood-red accent.

Stack

Role: Set art direction, paired generation in one pipeline run, targeted hardware detailing & consistency QA
Base model: Krea 2 (open-weights)
LoRA: Self-trained style LoRA stack ↓
Tools: ComfyUI, self-trained style LoRA stack + Krea 2, directed generation, targeted detailing
Year: 2026

Process

Generated as one set

Both characters come from a single pipeline run — a self-trained style LoRA stack on an open-weights base, with one shared prompt skeleton binding the pair: the same matte-black tailored-military language, a single blood-red accent, gunmetal hardware, and matching unit insignia. Selection from a seed batch, then targeted hardware detailing and consistency QA.

Directed hardware pass

The gear was finished in a second pass rather than left to the base generation: the drop-leg pistol holster, the tactical belt with its quick-release buckle and magazine pouches, and the suppressed carbines were each regenerated as crisp, correctly-oriented modern hardware and composited back in place — so the fashion-forward silhouette keeps a plausible, readable kit.

From the AI Lab

Model Card Cinematic Painterly Realism — LoRA Low-saturation painterly realism for splash art and character concepts. 85-image dataset.

Overview

A style LoRA for cinematic, low-saturation painterly realism — the workhorse behind the splash-pipeline posts (Empress of the Lotus Pond, Saint of the Storm, Golden Emperor).

Card


Base model	Krea 2 (open-weights, turbo)
Dataset	85 curated images, pure-content prose captions, no trigger word
Network	LoRA dim 32 / alpha 32 · 16 epochs · final loss 0.067
Validation	Epoch sweep with fixed-seed probes; epoch 16 selected as standard
Operating notes	Dark cinematic backdrops are part of the style contract — flat studio backgrounds suppress it

Notes

Prompting rules discovered in validation (front-loaded subject, positive-only phrasing at CFG 1, backdrop placement) are documented per post in the Process sections.

Sample output

Delivered splash art generated with this LoRA at its standard operating point:

Sample output — splash art

Open in LAB →