← Back to Skills Marketplace
macwx

Qwen Image Edit

by 清如许 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
40
Downloads
1
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install qwen-image-edit
Description
千问图像生成与编辑助手。支持文生图、图像编辑(增删改元素 / 修改文字 / 风格迁移)、多图融合。擅长复杂中文文字渲染,可生成海报、封面、展板、宣传图等商用素材。触发词:生成图片、画图、做海报、P图、改图、修图、抠图、换背景、加文字、融合图片。
README (SKILL.md)

Qwen Image

千问图像生成与编辑助手。支持文生图、图像编辑(增删改元素 / 修改文字 / 风格迁移)、多图融合。擅长复杂中文文字渲染,可生成海报、封面、展板、宣传图等商用素材。触发词:生成图片、画图、做海报、P图、改图、修图、抠图、换背景、加文字、融合图片。 Generate and edit images using Alibaba Cloud's Qwen Image API (千问图像). Qwen Image AI generation & editing assistant. Supports text-to-image, image editing (add/remove/change elements, modify text, style transfer), and multi-image fusion. Excels at complex Chinese text rendering, produces commercial assets like posters, covers, banners, and promotional graphics. Triggers: generate image, image editing, text-to-image, poster design, make a poster, photo editing, image fusion.

使用阿里云千问图像 API 生成与编辑图像。 千问图像 AI 生成与编辑助手。支持文生图、图像编辑(增删改元素、修改文字、风格迁移)以及多图融合。擅长复杂中文文字渲染,可生成海报、封面、展板、宣传图等商用素材。触发词:生成图片、画图、做海报、P图、改图、修图、抠图、换背景、加文字、融合图片。

Usage

用法

Text-to-Image

文生图

Generate an image (returns URL only): 生成图像(仅返回 URL):

python {baseDir}/scripts/generate_image.py --prompt "一只橘猫坐在窗台上晒太阳"

Generate and save locally: 生成并保存到本地:

python {baseDir}/scripts/generate_image.py --prompt "a beautiful sunset over mountains" --filename sunset.png

With custom size: 自定义尺寸生成:

python {baseDir}/scripts/generate_image.py --prompt "城市夜景" --size "2688*1536" --filename city.png

Generate multiple images: 生成多张图像:

python {baseDir}/scripts/generate_image.py --prompt "水彩风格的花卉" --n 4 --filename flower.png

Image Editing

图像编辑

Single image editing: 单图编辑:

python {baseDir}/scripts/generate_image.py --images photo.png --prompt "在画面左下角加一只猫" --filename edited.png

Multi-image fusion (up to 3 images): 多图融合(最多 3 张图像):

python {baseDir}/scripts/generate_image.py --images bg.png style.png --prompt "将图二的风格应用到图一" --filename fused.png

Edit with a URL input: 通过 URL 输入进行编辑:

python {baseDir}/scripts/generate_image.py --images https://example.com/photo.jpg --prompt "remove the text on the wall" --filename cleaned.png

Use dedicated edit model: 使用专用编辑模型:

python {baseDir}/scripts/generate_image.py --images input.png --prompt "把天空改成星空" -m qwen-image-edit-max --filename starry.png

API Key

API 密钥

Obtain the API key in the following order: 按以下优先级获取 API 密钥:

  1. --api-key command line argument

  2. DASHSCOPE_API_KEY environment variable

  3. Get your API key from: https://dashscope.console.aliyun.com/

  4. --api-key 命令行参数

  5. DASHSCOPE_API_KEY 环境变量

  6. 从以下地址获取 API 密钥:https://dashscope.console.aliyun.com/

Models

模型

Model Mode Description
qwen-image-2.0-pro generate + edit (default) Best text rendering & realism
qwen-image-2.0 generate + edit Faster, balanced quality
qwen-image-edit-max edit Strongest editing: industrial design, geometry, character consistency
qwen-image-edit-plus edit Multi-output & custom resolution
qwen-image-edit edit Basic single-image editing
模型 模式 说明
qwen-image-2.0-pro 生成 + 编辑(默认) 最佳文字渲染与写实效果
qwen-image-2.0 生成 + 编辑 更快速,质量均衡
qwen-image-edit-max 编辑 最强编辑能力:工业设计、几何精度、角色一致性
qwen-image-edit-plus 编辑 多输出与自定义分辨率
qwen-image-edit 编辑 基础单图编辑

Options

参数选项

--images / -i — Input image(s) for editing (1-3 local paths or URLs). Omit for text-to-image. --images / -i — 用于编辑的输入图像(1-3 个本地路径或 URL)。文生图时无需指定。

--size / -s — Output resolution WxH. Omit to auto-match input image ratio. --size / -s — 输出分辨率 宽x高。省略时自动匹配输入图像比例。

  • qwen-image-2.0 series (512x512 ~ 2048x2048):
  • qwen-image-2.0 系列(512x512 ~ 2048x2048):
    • 2048*2048 — 1:1 (default when generating) / 1:1(生成时默认)
    • 2688*1536 — 16:9
    • 1536*2688 — 9:16
    • 2368*1728 — 4:3
    • 1728*2368 — 3:4
  • Edit models (width & height each in [512, 2048]):
  • 编辑模型(宽高各在 [512, 2048] 范围内):
    • 1024*1024, 1536*1024, 1024*1536, 1280*960, 720*1280, etc.

--n — Number of output images (1-6, default: 1). Not supported on qwen-image-edit. --n — 输出图像数量(1-6,默认:1)。qwen-image-edit 模型不支持。

--negative-prompt / -n — Elements to avoid (default: common AI artifacts). --negative-prompt / -n — 需要避免的元素(默认:常见 AI 伪影)。

--no-prompt-extend — Disable automatic prompt enhancement (enabled by default). --no-prompt-extend — 禁用自动提示词增强(默认启用)。

--seed — Random seed [0, 2147483647] for reproducible results. --seed — 随机种子 [0, 2147483647],用于生成可复现的结果。

--watermark — Add "Qwen-Image" watermark to the output. --watermark — 在输出图像上添加"Qwen-Image"水印。

--regionbeijing (default) or singapore. Singapore requires --workspace-id. --regionbeijing(默认)或 singapore。新加坡区域需要 --workspace-id

--no-verify-ssl — Disable SSL verification (use behind corporate proxy). --no-verify-ssl — 禁用 SSL 验证(在企业代理网络后使用)。

Workflow

工作流程

  1. If the user provides images, pass them via --images (local paths or URLs).

  2. Execute generate_image.py with the user's prompt and options.

  3. Parse the output for lines starting with MEDIA_URL:.

  4. Display each image using markdown: ![Generated Image](URL).

  5. Do NOT download or save the image unless the user specifically requests it (use --filename).

  6. 如果用户提供了图像,通过 --images 参数传入(本地路径或 URL)。

  7. 使用用户的提示词和选项执行 generate_image.py

  8. 解析输出中以 MEDIA_URL: 开头的行。

  9. 使用 markdown 格式展示每张图像:![生成图像](URL)

  10. 除非用户明确要求,否则不要下载或保存图像(使用 --filename 指定保存)。

Notes

注意事项

  • Supports both Chinese and English prompts.

  • Image editing accepts local files (auto Base64-encoded, max 10MB) or URLs (http/https/oss).

  • Output images are PNG format. Image URLs expire in 24 hours.

  • qwen-image-edit model does not support --size or --no-prompt-extend.

  • Default negative prompt helps avoid common AI artifacts (distorted limbs, low quality, etc.).

  • Beijing and Singapore regions have independent API keys and endpoints — do not mix.

  • 支持中英文提示词。

  • 图像编辑支持本地文件(自动 Base64 编码,最大 10MB)或 URL(http/https/oss)。

  • 输出图像为 PNG 格式。图像 URL 在 24 小时后过期。

  • qwen-image-edit 模型不支持 --size--no-prompt-extend 参数。

  • 默认的反向提示词有助于避免常见 AI 伪影(肢体扭曲、低质量等)。

  • 北京和新加坡区域拥有独立的 API 密钥和端点——请勿混用。

Usage Guidance
Install only if you are comfortable using Alibaba Cloud DashScope for image generation/editing. Treat prompts, local images, and image URLs you provide as data sent to that service, and do not use private, confidential, or regulated images unless that matches your own privacy and compliance expectations.
Capability Tags
requires-sensitive-credentials
Capability Assessment
Purpose & Capability
The script behavior matches the stated purpose: it generates or edits images through Alibaba Cloud DashScope/Qwen, optionally using user-supplied local images or URLs.
Instruction Scope
Activation wording is broad for image-generation and photo-editing requests, but the skill repeatedly identifies the Qwen/Alibaba API purpose and does not include hidden or unrelated instructions.
Install Mechanism
The artifact contains a markdown skill file, a small metadata file, and one Python helper script with a single declared requests dependency; no installer, background service, or startup hook is present.
Credentials
Network access, a DashScope API key, prompt submission, and optional local image upload are proportionate for cloud image generation, but users should avoid sending sensitive images unless they intend to use Alibaba Cloud for that content.
Persistence & Privilege
No persistence or privilege escalation is shown. The script only writes output image files when the user provides a filename and otherwise prints returned media URLs.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install qwen-image-edit
  3. After installation, invoke the skill by name or use /qwen-image-edit
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of qwen-image-edit, a versatile image generation and editing assistant. - Supports text-to-image, image editing (add/remove/change elements, text modification, style transfer), and multi-image fusion. - Excels at complex Chinese text rendering; suitable for posters, covers, banners, and other commercial graphics. - Offers CLI usage examples, model selection, and a wide range of output and prompt customization options. - Compatible with both Chinese and English prompts; supports local file and URL image inputs.
Metadata
Slug qwen-image-edit
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Qwen Image Edit?

千问图像生成与编辑助手。支持文生图、图像编辑(增删改元素 / 修改文字 / 风格迁移)、多图融合。擅长复杂中文文字渲染,可生成海报、封面、展板、宣传图等商用素材。触发词:生成图片、画图、做海报、P图、改图、修图、抠图、换背景、加文字、融合图片。 It is an AI Agent Skill for Claude Code / OpenClaw, with 40 downloads so far.

How do I install Qwen Image Edit?

Run "/install qwen-image-edit" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Qwen Image Edit free?

Yes, Qwen Image Edit is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Qwen Image Edit support?

Qwen Image Edit is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Qwen Image Edit?

It is built and maintained by 清如许 (@macwx); the current version is v1.0.0.

💬 Comments