---
title: Qwen-Image-2.1-Fun-Acc-LoRAs
canonical_url: "https://www.modelscope.cn/models/PAI/Qwen-Image-2.1-Fun-Acc-LoRAs"
md_url: "https://www.modelscope.cn/models/PAI/Qwen-Image-2.1-Fun-Acc-LoRAs.md"
repository: PAI/Qwen-Image-2.1-Fun-Acc-LoRAs
last_updated: 2026-09-24
license: other
pipeline_tag: text-to-image-synthesis
tasks:
  - text-to-image-synthesis
base_model:
  - Qwen/Qwen-Image-2.1
base_model_relation: adapter
library_name:
  - lora
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 293
stars: 1
tags:
  - text-to-image
  - image-editing
---

# Qwen-Image-2.1-Fun-Acc-LoRAs

> Qwen-Image-2.1-Fun-Acc-LoRAs - PAI 在 ModelScope 开源的模型。Qwen-Image-2.1-Fun-Acc-LoRAs

PAI/Qwen-Image-2.1-Fun-Acc-LoRAs 是 ModelScope 魔搭社区上的text-to-image-synthesis模型，采用 other 许可，基于 Qwen/Qwen-Image-2.1 构建。

- **Repository**: PAI/Qwen-Image-2.1-Fun-Acc-LoRAs
- **License**: other
- **Tasks**: text-to-image-synthesis
- **Base model**: Qwen/Qwen-Image-2.1
- **Tags**: text-to-image, image-editing
- **Downloads**: 293
- **Stars**: 1
- **Last updated**: 2026-09-24

Source: https://www.modelscope.cn/models/PAI/Qwen-Image-2.1-Fun-Acc-LoRAs

---

# Qwen-Image-2.1-Fun-Acc-LoRAs

## Introduction

Built with Qwen.

We apply Parallel Decoding Distillation (PDD) <sup>[1](#ref1)</sup>  to
[Qwen-Image-2.1](https://huggingface.co/Qwen/Qwen-Image-2.1), enabling efficient text-to-image and instruction-based image editing at **4 NFE**.

For more details, please refer to our [GitHub repo](https://github.com/aigc-apps/VideoX-Fun).

| Name | Base model | Description |
|---|---|---|
| [Qwen-Image-2.1-Fun-Acc-4Step.safetensors](models/Qwen-Image-2.1-Fun-Acc-4Step.safetensors) | Qwen-Image-2.1 | 4 NFE. Official 4 Step Acc LoRA (`rank=64` and `network_alpha=64` in BF16) for Qwen-Image-2.1.|

## Demo

## Text-to-image

<table border="0" style="width: 100%; table-layout: fixed; text-align: center; margin-top: 20px;">
    <thead>
        <tr>
            <th style="text-align: center;" width="33.33%">Teacher · 40 NFE</th>
            <th style="text-align: center;" width="33.33%">Ours / PDD LoRA · 4 NFE</th>
            <th style="text-align: center;" width="33.33%">Viggle v0.1 full · 4 NFE</th>
        </tr>
    </thead>
    <tbody>
        <tr>
            <td width="33.33%"><a href="results/teacher/pdd_val_04_seed42.png"><img src="results/teacher/pdd_val_04_seed42.png" alt="pdd_val_04" width="100%"></a></td>
            <td width="33.33%"><a href="results/ours/pdd_val_04_seed42.png"><img src="results/ours/pdd_val_04_seed42.png" alt="pdd_val_04" width="100%"></a></td>
            <td width="33.33%"><a href="results/viggle_v01/pdd_val_04_seed42.png"><img src="results/viggle_v01/pdd_val_04_seed42.png" alt="pdd_val_04" width="100%"></a></td>
        </tr>
        <tr>
            <td width="33.33%"><a href="results/teacher/pdd_val_05_seed45.png"><img src="results/teacher/pdd_val_05_seed45.png" alt="pdd_val_05" width="100%"></a></td>
            <td width="33.33%"><a href="results/ours/pdd_val_05_seed45.png"><img src="results/ours/pdd_val_05_seed45.png" alt="pdd_val_05" width="100%"></a></td>
            <td width="33.33%"><a href="results/viggle_v01/pdd_val_05_seed45.png"><img src="results/viggle_v01/pdd_val_05_seed45.png" alt="pdd_val_05" width="100%"></a></td>
        </tr>
        <tr>
            <td width="33.33%"><a href="results/teacher/pdd_val_07_seed43.png"><img src="results/teacher/pdd_val_07_seed43.png" alt="pdd_val_07" width="100%"></a></td>
            <td width="33.33%"><a href="results/ours/pdd_val_07_seed43.png"><img src="results/ours/pdd_val_07_seed43.png" alt="pdd_val_07" width="100%"></a></td>
            <td width="33.33%"><a href="results/viggle_v01/pdd_val_07_seed43.png"><img src="results/viggle_v01/pdd_val_07_seed43.png" alt="pdd_val_07" width="100%"></a></td>
        </tr>
    </tbody>
</table>

> [!NOTE]
> The prompts above are from the PDD paper.

## Image editing

<table border="0" style="width: 100%; table-layout: fixed; text-align: center; margin-top: 20px;">
    <thead>
        <tr>
            <th style="text-align: center;" width="33.33%">Teacher · 40 NFE</th>
            <th style="text-align: center;" width="33.33%">Ours / PDD LoRA · 4 NFE</th>
            <th style="text-align: center;" width="33.33%">Viggle v0.1 full · 4 NFE</th>
        </tr>
    </thead>
    <tbody>
        <tr>
            <td width="33.33%"><a href="results/teacher/qwen_edit_1412128_seed42.png"><img src="results/teacher/qwen_edit_1412128_seed42.png" alt="qwen_edit_1412128" width="100%"></a></td>
            <td width="33.33%"><a href="results/ours/qwen_edit_1412128_seed42.png"><img src="results/ours/qwen_edit_1412128_seed42.png" alt="qwen_edit_1412128" width="100%"></a></td>
            <td width="33.33%"><a href="results/viggle_v01/qwen_edit_1412128_seed42.png"><img src="results/viggle_v01/qwen_edit_1412128_seed42.png" alt="qwen_edit_1412128" width="100%"></a></td>
        </tr>
        <tr>
            <td width="33.33%"><a href="results/teacher/viggle_edit_2ref_cat_seed43.png"><img src="results/teacher/viggle_edit_2ref_cat_seed43.png" alt="viggle_edit_2ref_cat" width="100%"></a></td>
            <td width="33.33%"><a href="results/ours/viggle_edit_2ref_cat_seed43.png"><img src="results/ours/viggle_edit_2ref_cat_seed43.png" alt="viggle_edit_2ref_cat" width="100%"></a></td>
            <td width="33.33%"><a href="results/viggle_v01/viggle_edit_2ref_cat_seed43.png"><img src="results/viggle_v01/viggle_edit_2ref_cat_seed43.png" alt="viggle_edit_2ref_cat" width="100%"></a></td>
        </tr>
        <tr>
            <td width="33.33%"><a href="results/teacher/viggle_edit_3ref_klein_seed44.png"><img src="results/teacher/viggle_edit_3ref_klein_seed44.png" alt="viggle_edit_3ref_klein" width="100%"></a></td>
            <td width="33.33%"><a href="results/ours/viggle_edit_3ref_klein_seed44.png"><img src="results/ours/viggle_edit_3ref_klein_seed44.png" alt="viggle_edit_3ref_klein" width="100%"></a></td>
            <td width="33.33%"><a href="results/viggle_v01/viggle_edit_3ref_klein_seed44.png"><img src="results/viggle_v01/viggle_edit_3ref_klein_seed44.png" alt="viggle_edit_3ref_klein" width="100%"></a></td>
        </tr>
    </tbody>
</table>

> [!NOTE]
> The editing prompts and reference images are from the [Qwen-Image-2.1](https://github.com/QwenLM/Qwen-Image-2.1/tree/main/prompt_rewrite/data) (flag) and [Viggle's demo examples](https://huggingface.co/spaces/Viggle/Qwen-Image-2.1-viggle-turbo/tree/main/examples) (woman and cat; person, cat and bird).

## Quick Start

### Diffusers only

Keep `qwenimage21_pdd.py` and `lora_utils_pdd.py` beside the scripts:

```bash
python predict_t2i.py
python predict_t2i_edit.py
```

### VideoX-Fun

Use a [VideoX-Fun](https://github.com/aigc-apps/VideoX-Fun) checkout exposing
`QwenImage21Pipeline` from `videox_fun.pipeline`. These entries use that pipeline and
the same bundled `qwenimage21_pdd.py` / `lora_utils_pdd.py` as the standalone entries;

```bash
python predict_t2i_videox_fun.py
python predict_t2i_edit_videox_fun.py
```

## Limitations

- Dense, small text can degrade noticeably compared with the teacher, with distorted character strokes and reduced legibility.
- Some image-editing outputs can appear slightly blurrier and darker than the teacher's results, with reduced fine-detail clarity.

## Reference
<ol>
  <li id="ref1">Neta Shaul, et al. "Parallel Decoding Distillation for Fast Image and Video Generation.". arXiv preprint arXiv:2607.26004 (2026).</li>
</ol>