Krea2 Surrealism Fantasy Style LoRA Released with 248 Images
kunge666 · reddit · 2026-08-22
The author released a surrealism fantasy style LoRA trained on Krea2. It was trained for 7 hours using 248 carefully selected images over 6000 iterations. The trigger word is kunge-fantasy with suggested weights of 0.8-1. The model features high detail and generalization, recommended for use with CFG 1, 8 steps, and the eulerancestral sampler.
More from Multimodal
- New Krea 2 Turbo Distillation LoRA: 4-step generation with 13% reduced error gap — TimeTruth2490 · 2026-08-22
- Meitu's CFT Framework Solves Instability in AI Portrait Relighting — 机器之心 · 2026-08-22
- GPT-4o Image Gen Test: Portrait Details with Disposable Camera Style — SimplyAnnisa · 2026-08-22
- Creator showcases AI-generated video and sound work using Luma AI — TinfoilTricorn · 2026-08-22
- Vision for Universal High-Res 3D Model Library as Future of VR — ZeroStateReflex · 2026-08-22
- CMU Lift4D: Single-view 4D reconstruction in-the-wild — rsasaki0109 · 2026-08-22