license: "other"
license_name: "krea-2-community-license"
license_link: "https://huggingface.co/Noctaluna/Krea2-Noct-Uncensored/blob/main/LICENSE.pdf"
base_model: "krea/Krea-2-Turbo"
base_model_relation: "merge"
library_name: "diffusion-single-file"
pipeline_tag: "text-to-image"
tags:
- "text-to-image"
- "krea-2"
- "comfyui"
- "int8"
- "photorealistic"
- "uncensored"
- "nsfw"
- "not-for-all-audiences"
widget: - output:
url: "images/01.jpg" - output:
url: "images/02.jpg" - output:
url: "images/03.jpg" - output:
url: "images/04.jpg" - output:
url: "images/05.jpg"
Krea Noct: Krea 2 Uncensored Realism
A photoreal Krea 2 merge for NSFW work, built on the official Krea 2 Turbo. Explicit prompts work with no bypass LoRA, in 8 steps.
The full gallery with prompts, the int4, fp8 and bf16 files and the reviews are on Civitai: Krea Noct.
What it does well
- Full-body and seated nudes with natural bodies
- Phone-snapshot and flash looks, daylight interiors, outdoor scenes
- Anime, illustration and 3D too, with no LoRA
- Plain-caption prompts: who, what she does, where, the light
File
| file | size | use |
|---|---|---|
KreaNoct_V3_int8_convrot.safetensors |
12.6 GB | 12 GB cards and up, ComfyUI 0.37 or newer, core nodes only |
It goes in models/diffusion_models and loads with the Load Diffusion Model node. On 8 GB cards, take the int4 mixed file from the Civitai page (about 1.4x faster).
Settings
8 steps, cfg 1, euler, simple, 1024x1536, no upscale needed and no negative prompt.
- Text encoder:
qwen3vl_4b_fp8_scaled.safetensorsinmodels/text_encoders - VAE:
qwen_image_vae.safetensorsinmodels/vae
8 steps is the sweet spot; go to 12 or 16 if hands or small text bother you. Describe the face in one clause and it stays consistent across seeds. Full-body framing and an off-camera gaze give the most natural results.
How to prompt
Write the prompt like a caption for a photo that already exists, in full sentences: the kind of shot first, then the person, the pose, the setting and the light. Name what should be visible (a stack of books on the floor, morning light from the left) rather than moods or quality words. Tag lists and words like "masterpiece" or "8k" work worse.
Template: An amateur photograph of <who, hair, build>, <pose and where>, <where she looks>. <Light source and direction>, <two or three props>. Full body from <angle>, phone camera look, natural skin.
Example (seed 97032, the settings above): A phone photograph of a nude woman in her twenties with strawberry-blonde hair in a loose braid and an athletic build, lying on her stomach on a patchwork quilt on the plank floor of a wooden treehouse, her chin on her folded hands, her bare feet raised and crossed, looking out of a small square window at green leaves. Dappled sunlight through the window across her back and legs, a wicker basket and a string of unlit fairy lights along the wall. Full body from the side at floor level, natural skin texture.
License
Krea Noct is a derivative of Krea 2 and follows the Krea 2 Community License (LICENSE.pdf). The model itself may not be sold. Don't use it on real people; every character must be an adult.
中文：Krea 2 写实无审查融合模型，无需绕过 LoRA，也能画动漫与插画。8 步、cfg 1、euler/simple、1024×1536。int8 convrot 需 12G 显存；8G 显存请在 Civitai 页面下载 int4 版本。