Explore my workflow creation space! Feel free to click the link below and experience the future of AIGC generation directly.<a target="_blank" rel="ugc" href="https://www.runninghub.ai/?inviteCode=rh-v1559">https://www.runninghub.ai/?inviteCode=rh-v1559</a>Fan benefits,register to get 1000 points,daily login 100 points,play 4090!Experience the super power of 48G.Description: Elevate your portraits with an elite cinematic, analog aesthetic. Krea2_GrainGaze-portrait specializes in rendering hyper-realistic tactile skin details, rich organic micro-grain textures, and poetic diffused lighting. Perfect for capturing soft, soulful glances, atmospheric cafe/studio setups, and high-end editorial realism without any digital glare.Recommended Settings (Krea 2 Turbo):<ul><li>Steps: 8 steps</li><li>CFG Scale: 1</li><li>LoRA Weight: 0.8 - 1.0</li><li>Aspect Ratio: 3:4 or 1:1 (Works flawlessly with white borders for an editorial magazine look)</li></ul>

Krea2_GrainGaze-portrait

Krea2+Turbo.json

Krea2_GrainGaze-portrait.safetensors

v1.0


# Krea2_微粒幻觉-人像

## 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：100
* 总训练步数：10000
* 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_GrainGaze-portrait", origin_file_pattern="Krea2_GrainGaze-portrait_c1-st10000.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: 737de687ff104d5a70d8d2b02301e943c5c6c69542795371f823d3455a1f751a