Qwen Edit 2509 Multiple angles

提供商dx8152
分类image-to-image
许可证apache-2.0
下载量6.9K
星标28

简介

Qwen Edit 2509 Multiple angles 是一款专注于图像局部编辑与多视角一致性生成的 AI 模型。它解决了传统图生图模型在修改细节时容易导致整体崩坏的痛点，能够根据指令对图像特定区域进行精准调整，并尝试在不同视角下保持主体特征的统一。对于需要快速迭代视觉方案、进行产品多角度展示或精细化修图的开发者和设计师来说，它提供了一个低门槛的工具链，能够有效替代繁琐的手动重绘工作。

核心亮点

  • 精准局部编辑，避免全局图像风格漂移
  • 增强多视角一致性，适合角色或产品展示
  • Apache-2.0 协议，商业集成友好且灵活
  • 上手简单，通过指令快速实现图像微调

使用方法

安装依赖
# 安装 Hugging Face transformers
pip install transformers torch
SDK 使用
# 使用 transformers 加载模型
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("dx8152/Qwen-Edit-2509-Multiple-angles")
tokenizer = AutoTokenizer.from_pretrained("dx8152/Qwen-Edit-2509-Multiple-angles")

Hugging Face 下载

我们推荐使用命令行或者 Hugging Face Hub SDK 来进行模型的下载。

操作指引：在下载前，请先通过如下命令安装 huggingface_hub：

操作指引 
pip install -U huggingface_hub

命令行下载

下载完整模型库

下载完整模型库 
huggingface-cli download dx8152/Qwen-Edit-2509-Multiple-angles

下载单个文件到指定本地文件夹（以下载 config.json 到当前路径下 ./dir 目录为例）

下载单个文件到指定本地文件夹（以下载 config.json 到当前路径下 ./dir 目录为例） 
huggingface-cli download dx8152/Qwen-Edit-2509-Multiple-angles config.json --local-dir ./dir

更多命令行下载选项，可参见官方文档

SDK 下载

SDK 下载 
# 模型下载
from huggingface_hub import snapshot_download
model_dir = snapshot_download('dx8152/Qwen-Edit-2509-Multiple-angles')

Git 下载

请确保 lfs 已经被正确安装

Git 下载 
git lfs install
git clone https://huggingface.co/dx8152/Qwen-Edit-2509-Multiple-angles

如果您希望跳过 lfs 大文件下载，可以使用如下命令

跳过 LFS 
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/dx8152/Qwen-Edit-2509-Multiple-angles

模型文件托管在 Hugging Face Hub，使用 HF CLI / SDK / Git 直接下载，不经过本站。

PyTorch / Transformers 使用

安装 Transformers

安装 Transformers 
pip install -U transformers torch

模型加载和推理

模型加载和推理 
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('dx8152/Qwen-Edit-2509-Multiple-angles')
tokenizer = AutoTokenizer.from_pretrained('dx8152/Qwen-Edit-2509-Multiple-angles')

模型下载

我们推荐使用命令行或者 ModelScope SDK 来进行模型的下载。

操作指引：在下载前，请先通过如下命令安装 ModelScope：

操作指引 
pip install modelscope

命令行下载

下载完整模型库

下载完整模型库 
modelscope download --model dx8152/Qwen-Edit-2509-Multiple-angles

下载单个文件到指定本地文件夹（以下载 README.md 到当前路径下 dir 目录为例）

下载单个文件到指定本地文件夹（以下载 README.md 到当前路径下 dir 目录为例） 
modelscope download --model dx8152/Qwen-Edit-2509-Multiple-angles README.md --local_dir ./dir

更多更丰富的命令行下载选项，可参见具体文档

SDK 下载

SDK 下载 
# 模型下载
from modelscope import snapshot_download
model_dir = snapshot_download('dx8152/Qwen-Edit-2509-Multiple-angles')

Git 下载

请确保 lfs 已经被正确安装

Git 下载 
git lfs install
git clone https://www.modelscope.cn/dx8152/Qwen-Edit-2509-Multiple-angles.git

如果您希望跳过 lfs 大文件下载，可以使用如下命令

跳过 LFS 
GIT_LFS_SKIP_SMUDGE=1 git clone https://www.modelscope.cn/dx8152/Qwen-Edit-2509-Multiple-angles.git

ModelScope 模型页直接下载模型文件；无需将模型文件放在本站服务器。

Notebook 快速开发

下载并安装 ModelScope library

下载并安装 ModelScope library 
pip install "modelscope[audio,cv,nlp,multi-modal,science]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html

模型加载和推理

模型加载和推理 
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

p = pipeline('text-generation', 'dx8152/Qwen-Edit-2509-Multiple-angles')

完整文档

来源: HuggingFace

---
license: apache-2.0
base_model:

  • Qwen/Qwen-Image-Edit-2509

pipeline_tag: image-to-image
tags:
  • lora

library_name: diffusers
---
import os
from huggingface_hub import InferenceClient

client = InferenceClient(
provider="wavespeed",
api_key=os.environ["HF_TOKEN"],
)

with open("cat.png", "rb") as image_file:
input_image = image_file.read()

output is a PIL.Image object

image = client.image_to_image( input_image, prompt="Turn the cat into a tiger.", model="dx8152/Qwen-Edit-2509-Multiple-angles", ) This model is trained (code-free!) on ModelScope. Thanks to ModelScope team for providing the training infra:

https://www.modelscope.cn/aigc/modelTraining

-------

Updated 2025/11/2: Some people mentioned that the model has an unstable consistency issue. I have re-uploaded a version with more training iterations, hoping to fix the consistency problem.

Welcome everyone to use Lora of Qwen-Edit-2509, its performance is very amazing！

There are no trigger words. You can control the camera to move up, down, left, and right, as well as rotate it to the left and right. You can also look down or up. The camera can be changed to a wide-angle or close-up shot.

Online running link： www.runninghub.ai/post/1985311204883243009?inviteCode=rh-v1331

This is a user guide:

YouTube：https://youtu.be/UGdW8W1MqW8

Blibili：https://www.bilibili.com/video/BV1oi1gBBEZV/

“
将镜头向前移动（Move the camera forward.）
将镜头向左移动（Move the camera left.）
将镜头向右移动（Move the camera right.）
将镜头向下移动（Move the camera down.）
将镜头向左旋转45度（Rotate the camera 45 degrees to the left.）
将镜头向右旋转45度（Rotate the camera 45 degrees to the right.）
将镜头转为俯视（Turn the camera to a top-down view.）
将镜头转为广角镜头（Turn the camera to a wide-angle lens.）
将镜头转为特写镜头（Turn the camera to a close-up.）
...
There are many possibilities; you can try them yourself.
”
------

Instructions: Download the lora file to the models/loras folder.

You also need this lora and use them together: https://huggingface.co/lightx2v/Qwen-Image-Lightning/tree/main

For communication/cooperation, you can join the discord group to communicate： https://discord.gg/yVAVa43mWk

If these resources are helpful to you, or if you use them for business purposes, please buy me a coffee. Thank you for supporting original content! PayPal: Daniel8152

!效果图
!效果图2
!效果图3
I saw some people wanted to see a comparison between LoRa and the model's inherent capabilities, so I conducted a test. The trained model is indeed more powerful and intelligent, while the original model already possessed certain abilities. It's very powerful, which is why we love this model, and precisely because of this, we need to explore its potential.
!对比图

查看官方来源