Sign In

Kreamagine (LoRA)

Updated: Aug 21, 2026

base model

Download

1 variant available

bf16 SafeTensor

Kreamagine_LoRA.safetensors

BF16, good balance • 218 MB

Verified:

Type
LoRA
Stats

51

Reviews
Published

Aug 24, 2026

Base Model

Other

Hash
AutoV2
D0B462B5F1
default creator card background decoration
Followers - 627

627

Likes - 2208

2.2K

Test__00080_.png

Kreamagine: The Krea-2-Turbo Style Finetune

"More Aggressive Lighting. Higher Detail Density. Same Turbo Speed."

Kreamagine is a precision-tuned adaptation of Krea-2-Turbo, designed to push the boundaries of what this DiT architecture can do in terms of lighting contrast and micro-detail. Trained on a curated dataset of Grok Imagine v0.9/1.0 aesthetics, this LoRA shifts the model away from the "safe," even illumination of the base model and into a more cinematic, high-contrast space.

Key Features

  • Aggressive Light Shaping: Kreamagine trades the base model's even lighting for hard key lights, deep contrast ratios, and dramatic falloff. It’s not afraid to let regions go completely black or push highlights.

  • Enhanced Detail Density: Significant improvements in micro-detail for skin, fabric, and foliage. Structures hold their integrity better at high resolutions (1024–1536).

  • No Trigger Word: The aesthetic is baked directly into the weights. You don't need a specific tag to activate it; simply prompt for the mood, cinematography, and lighting style you want, and the model will react accordingly.

  • Optimized Performance: Built for speed. It retains the Krea-2-Turbo 8-step generation flow, ensuring you get high-fidelity results without sacrificing the snappy inference speeds you expect from the Turbo base.

Technical Specs & Usage

  • Steps: 8 (Do not over-step)

  • CFG: 1.0 (Strictly 1.0—Turbo models are distilled and do not support higher CFG; raising it will degrade output)

  • Sampler: Euler

  • Scheduler: Simple

  • Resolution: 1024–1536 (Optimal)

Prompting Advice

Because Kreamagine responds strongly to lighting and camera terminology, treat your prompts like a photography brief. If you want a standard look, keep your prompts descriptive and plain. If you want to see the LoRA shine, use strong lighting language (e.g., "chiaroscuro," "sharp rim lighting," "deep shadows") to leverage the new weight mapping.

Note: This model is governed by the Krea 2 Community License Agreement. Commercial use is restricted.