---
title: High-End_Portrait_Retouching_klein_v21
canonical_url: "https://www.modelscope.cn/models/atent2L/High-End_Portrait_Retouching_klein_v21"
md_url: "https://www.modelscope.cn/models/atent2L/High-End_Portrait_Retouching_klein_v21.md"
repository: atent2L/High-End_Portrait_Retouching_klein_v21
chinese_name: "电商_模特皮肤真实质感还原_klein_v21"
last_updated: 2026-07-30
license: "Apache License 2.0"
pipeline_tag: text-to-image-synthesis
tasks:
  - text-to-image-synthesis
base_model:
  - black-forest-labs/FLUX.2-klein-base-9B
base_model_relation: adapter
parameters: 870.3M
tensor_type:
  - BF16
library_name:
  - pytorch
  - lora
  - safetensors
supports_inference: img2img
downloads: 42
stars: 0
tags:
  - LoRA
  - img2img
---

# High-End_Portrait_Retouching_klein_v21

> High-End_Portrait_Retouching_klein_v21 - atent2L 在 ModelScope 开源的模型。【V21 版本更新说明】本模型旨在真实还原人像皮肤质感，打造非 AI 感的天然质性表现。✨ V21 核心升级：儿童肌肤专属优化：针对儿童及青少年皮肤的细腻度、透光感与自然泛红进行了专项优化，保留独特的幼态皮肤质感。头发发丝去“AI感”：大幅改善了传统 AI 容易产生的“胶质发丝”与“塑料光泽”，使头发细节更自然、柔顺且具真实呼吸感。💡…

atent2L/High-End_Portrait_Retouching_klein_v21 是 ModelScope 魔搭社区上的 870.3M 参数text-to-image-synthesis模型，采用 Apache License 2.0 许可，基于 black-forest-labs/FLUX.2-klein-base-9B 构建，并支持在线推理（img2img）。

- **Repository**: atent2L/High-End_Portrait_Retouching_klein_v21
- **License**: Apache License 2.0
- **Tasks**: text-to-image-synthesis
- **Parameters**: 870.3M
- **Base model**: black-forest-labs/FLUX.2-klein-base-9B
- **Online inference**: img2img
- **Tags**: LoRA, img2img
- **Downloads**: 42
- **Stars**: 0
- **Last updated**: 2026-07-30

Source: https://www.modelscope.cn/models/atent2L/High-End_Portrait_Retouching_klein_v21

---

# 电商_模特皮肤真实质感还原_klein_v21

## 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：[black-forest-labs/FLUX.2-klein-base-9B](https://modelscope.cn/models/black-forest-labs/FLUX.2-klein-base-9B)
* 训练代码：[DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)
* 训练数据量：100
* 总训练步数：5000
* 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.flux2_image import Flux2ImagePipeline, ModelConfig
import torch

pipe = Flux2ImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-base-9B", origin_file_pattern="text_encoder/*.safetensors"),
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-base-9B", origin_file_pattern="transformer/*.safetensors"),
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-base-9B", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
    ],
    tokenizer_config=ModelConfig(model_id="black-forest-labs/FLUX.2-klein-base-9B", origin_file_pattern="tokenizer/"),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="atent2L/High-End_Portrait_Retouching_klein_v21", origin_file_pattern="High-End_Portrait_Retouching_klein_v21_c1-st5000.safetensors"))
prompt = "a cat"
image = pipe(prompt, num_inference_steps=50, cfg_scale=4)
image.save("image.jpg")

prompt = "Convert the image style to anime style."
image = pipe(prompt, edit_image=[image], num_inference_steps=50, cfg_scale=4)
image.save("image_edited.jpg")
```