---
title: QWEN2.1-Opaque_pantyhose
canonical_url: "https://www.modelscope.cn/models/heywego12/QWEN2.1-Opaque_pantyhose"
md_url: "https://www.modelscope.cn/models/heywego12/QWEN2.1-Opaque_pantyhose.md"
repository: heywego12/QWEN2.1-Opaque_pantyhose
chinese_name: "QWEN-2.1-厚裤袜lora"
last_updated: 2026-09-23
license: "Apache License 2.0"
pipeline_tag: text-to-image-synthesis
tasks:
  - text-to-image-synthesis
base_model:
  - Qwen/Qwen-Image-2.1
base_model_relation: adapter
parameters: 503.3M
tensor_type:
  - BF16
library_name:
  - pytorch
  - lora
  - safetensors
supports_inference: txt2img
downloads: 21
stars: 1
tags:
  - LoRA
  - text-to-image
---

# QWEN2.1-Opaque_pantyhose

> QWEN2.1-Opaque_pantyhose - heywego12 在 ModelScope 开源的模型。Qwen-Image 2.1 — Opaque Tights & Pantyhose (8 Colour × Finish Combos)

heywego12/QWEN2.1-Opaque_pantyhose 是 ModelScope 魔搭社区上的 503.3M 参数text-to-image-synthesis模型，采用 Apache License 2.0 许可，基于 Qwen/Qwen-Image-2.1 构建，并支持在线推理（txt2img）。

- **Repository**: heywego12/QWEN2.1-Opaque_pantyhose
- **License**: Apache License 2.0
- **Tasks**: text-to-image-synthesis
- **Parameters**: 503.3M
- **Base model**: Qwen/Qwen-Image-2.1
- **Online inference**: txt2img
- **Tags**: LoRA, text-to-image
- **Downloads**: 21
- **Stars**: 1
- **Last updated**: 2026-09-23

Source: https://www.modelscope.cn/models/heywego12/QWEN2.1-Opaque_pantyhose

---

# QWEN-2.1-厚裤袜lora

## 模型介绍

# Qwen-Image 2.1 — Opaque Tights & Pantyhose (8 Colour × Finish Combos)

**Base model: Qwen-Image 2.1**

A trigger-driven LoRA for **opaque pantyhose**, built for Qwen-Image 2.1. Eight combinations of colour × finish, so the tights can be matched to the light and the outfit instead of defaulting to black.

## Trigger words

`opaque_tights` is the base trigger — add **exactly one** colour/finish pair from the table below.

| Trigger combination | Colour | Finish | Suits |
|---|---|---|---|
| `opaque_tights, black_matte_tights` | black | matte | dim rooms, warm indoor light, dark outfits |
| `opaque_tights, black_glossy_tights` | black | glossy | hard directional light, night scenes |
| `opaque_tights, gray_matte_tights` | grey | matte | bright, low-saturation daylight |
| `opaque_tights, gray_glossy_tights` | grey | glossy | hard side light |
| `opaque_tights, white_matte_tights` | white | matte | bright, airy scenes |
| `opaque_tights, white_glossy_tights` | white | glossy | hard light against a darker set |
| `opaque_tights, flesh_matte_tights` | nude | matte | soft daylight, natural look |
| `opaque_tights, flesh_glossy_tights` | nude | glossy | hard light, sheen-focused shots |

## How to prompt it

**1. Put the triggers at the very front.** It works best when `opaque_tights, <colour>_<finish>_tights` sits in the first comma segment, ahead of the rest of the sentence.

**2. Do not describe the tights in the body text.** Transparency, gloss, highlights, weave and toe-cap wording are all produced by the LoRA — no need to write them. Describe the scene, the pose and the light instead.

**3. Match the finish to the light.** Matte reads under any light. Glossy needs a directional source (hard sun, side light or rim light); in fully diffused light the sheen will not come through.

**4. Feet in frame.** For raised-leg or foot close-ups, state the coverage ("runs unbroken from the waist over both feet as far as the tips of the toes") and add one line of anatomy for that foot (five toes held apart, each reading as its own rounded shape). This visibly steadies toe structure. Toe rendering stays probabilistic — judge results across several seeds, not a single image.

**5. Recommended setup.** Base model Qwen-Image 2.1. Default sampler settings from your own Qwen workflow are fine.

## 中文补充

面向 Qwen-Image 2.1 的**不透明连裤袜**触发词 LoRA，共 **8 档**（4 色 × 2 种光泽），目的是让裤袜跟着光线与穿搭走，而不是默认的稀烂的裤袜丝袜。

- **触发词**：`opaque_tights` 是基础触发词，后面**只加一组**颜色_光泽组合（见上表/或下表）。
- **写法**：触发词放最前（第一个逗号段）。**正文不要再写袜子的透明度、光泽、高光、织纹**——这些全部由 LoRA 输出，正文只写场景、姿势与光线。
- **光泽档要配定向光**：哑光任何光都能拍；油光需要硬质定向光（正射硬光／侧光／轮廓光），全漫射环境下光泽出不来。
- **脚部入镜**：抬腿或脚部近景时，写明"从腰到趾尖连续覆盖"，并给该脚补一句解剖（五趾微张、各自成形）——可明显稳定脚趾结构；脚趾属概率性结果，请用同参数多张判断，不要单张下结论。
- **推荐配置**：底模 Qwen-Image 2.1；采样参数沿用你自己的 Qwen 工作流默认值即可。
以下是不同类型裤袜的触发词
|触发词|颜色| 光泽| 训练时的光线和环境 |
|---|---|---|---|
| `opaque_tights, black_matte_tights` | black | matte | dim rooms, warm indoor light, dark outfits |
| `opaque_tights, black_glossy_tights` | black | glossy | hard directional light, night scenes |
| `opaque_tights, gray_matte_tights` | grey | matte | bright, low-saturation daylight |
| `opaque_tights, gray_glossy_tights` | grey | glossy | hard side light |
| `opaque_tights, white_matte_tights` | white | matte | bright, airy scenes |
| `opaque_tights, white_glossy_tights` | white | glossy | hard light against a darker set |
| `opaque_tights, flesh_matte_tights` | nude | matte | soft daylight, natural look |
| `opaque_tights, flesh_glossy_tights` | nude | glossy | hard light, sheen-focused shots |


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：[Qwen/Qwen-Image-2.1](https://modelscope.cn/models/Qwen/Qwen-Image-2.1)
* 训练代码：[DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)
* 训练数据量：100
* 总训练步数：6000
* 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.qwen_image_21 import QwenImage21Pipeline, ModelConfig
import torch

pipe = QwenImage21Pipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="text_encoder/model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="vae/diffusion_pytorch_model*.safetensors"),
    ],
    processor_config=ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="processor/"),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="heywego12/QWEN2.1-Opaque_pantyhose", origin_file_pattern="QWEN2.1-Opaque_pantyhose_c1-st6000.safetensors"))

prompt = "a cat"
image = pipe(prompt, seed=0, num_inference_steps=50)
image.save("image.png")
```