You can click on the link below to try it out directly. If the effect is good, you can deploy it locally<a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2082077962611425281?inviteCode=rh-v1559">https://www.runninghub.ai/post/2082077962611425281?inviteCode=rh-v1559</a>Fan benefits,register to get 1000 points,daily login 100 points,play 4090!Experience the super power of 48G.## Model OverviewWelcome to Krea2_DreamCos_Realism!A premium photorealistic LoRA engineered for High-Fidelity Cosplay Photography, dreamy glowing light atmospheres, realistic skin textures, and Bringing Anime/Game characters into real life.This model excels at capturing dreamy high-key lighting, soft bloom effects, and ultra-realistic skin details (SSS). It perfectly balances the aesthetic charm of 2D characters with the physical authenticity of real-world photography—rendering detailed wigs, intricate costumes, and ethereal studio lighting. Perfect for generating live-action anime/game cosplays, realistic character portraits, and cinematic fashion snapshots.## ⚙️ Recommended Parameters (Optimized for Krea2 Turbo)To capture the subtle luminous glow and lifelike skin textures, please apply the following configurations:* Base Architecture: Krea2* **Inference Pipeline:** Krea2 Turbo Version* Sampling Steps: 8 steps (Strictly optimized for Turbo speed)* CFG Scale: 1.0 (Crucial for Turbo models)* LoRA Weight: 0.8 - 1.0 (0.9 is highly recommended for balanced photorealism)* Core Trigger Tags: <code>dreamcos style</code>, <code>cosplay photo</code>, <code>dreamy lighting</code>, <code>soft glow</code>, <code>photorealistic skin texture</code>, <code>masterpiece portrait</code>, <code>cinematic lighting</code>## 🎨 Prompting Guide* To boost dreamy lighting and bloom effects: <code>dreamy atmosphere</code>, <code>soft rim light</code>, <code>ethereal glow</code>, <code>lens flare</code>, <code>subsurface scattering</code>* To enhance photorealistic cosplay textures: <code>detailed cosplay costume</code>, <code>realistic wig texture</code>, <code>fine skin pores</code>, <code>raw photo</code>, <code>8k portrait photography</code>

Krea2_DreamCoser_Realism

Krea2_GlowCosplay_WesternCoser.safetensors

Western

v1.0


# Krea2_幻光欧韵Coser

## 模型介绍

本模型依托魔搭社区（ModelScope）AIGC专区[模型训练](https://modelscope.cn/aigc/modelTraining)环境与算力完成训练。

* 模型类型：LoRA
* 基础模型：[krea/Krea-2-Turbo](https://modelscope.cn/models/krea/Krea-2-Turbo)
* 训练代码：[DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)
* 训练数据量：95
* 总训练步数：4000
* 开源协议：Apache-2.0

## 推理代码

安装 [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)：

```bash
pip install diffsynth
```

开始推理：

```python
from diffsynth.pipelines.krea2 import Krea2Pipeline, ModelConfig
import torch

pipe = Krea2Pipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="krea/Krea-2-Turbo", origin_file_pattern="turbo.safetensors"),
        ModelConfig(model_id="Qwen/Qwen3-VL-4B-Instruct", origin_file_pattern="*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
    ],
    tokenizer_config=ModelConfig(model_id="Qwen/Qwen3-VL-4B-Instruct", origin_file_pattern=""),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="zhouwenbin1994/Krea2_GlowCosplay_WesternCoser", origin_file_pattern="Krea2_GlowCosplay_WesternCoser_c2-st4000.safetensors"))
prompt = "a cat"
image = pipe(
    prompt, seed=0,
    num_inference_steps=8, cfg_scale=1, mu=1.15,
)
image.save("image.jpg")
```



Files with SHA256: 2f122221e7ba83d9bdc0012b08277ea8dd47e7d45a3f50a0f0877ac80fdef226