Full Krea2 concept-art pipeline: four custom LoRAs, 5MP native frames, two-step upscale to 7000px
Dacrikka · reddit · 2026-09-21
The author shares the complete AI-native pipeline behind their "MERIDIAN" concept art project built with Krea2:
- The anomaly creature evolved from abstract geometry to animal (shark, crocodile) and insect (spider, butterfly) hybrids as the concept matured.
- Rough sketches in Nano2 lock mood and silhouette first.
- Prompt engineering done in Claude Code with a custom Ultra Prompt agent translating narrative intent into structured, image-ready prompts.
- Generation in Krea2 with four custom LoRAs (LumenWake, KreaMidJ1, Noctilume, Mythgraft), weights rebalanced per shot between 0.2 and 0.8 depending on needed light/texture, composition, atmosphere, or creature detail.
- Every frame generated natively at 5MP, then upscaled to 4000×7000 via a two-step ComfyUI pass: SeedVR2 for detail reconstruction, then an RTX-based upscale for sharpness.
- All in-image typography (labels, warnings, lab documentation) was model-generated; only full-res breakage was rebuilt by hand in Photoshop.
The full project is open on Behance.
More from Multimodal
- Training an img2img LoRA on just 35 pairs to turn photos into Unreal Engine style — NeoHisto · 2026-09-21
- Qwen-Image-2.1: 9 simple ComfyUI workflows, near pixel-perfect editing — nomadoor · 2026-09-21
- Turn any video into volumetric playback with exportable 360° orbits at custom tempo — LinusEkenstam · 2026-09-21
- One image plus one prompt generates an entire cinematic getaway video — umesh_ai · 2026-09-21
- Phone video in, navigable 3D room out: K3-powered splat pipeline runs in the browser — willeastcott · 2026-09-21
- Viral Arabic AI trick: upload a photo and your name to get a magazine-cover portrait — aziz4ai · 2026-09-21