Image Lens

Analyze webpage images and generate editable prompts for image generation.

Image Lens turns images you find on the web into detailed, editable prompts for AI image generation, and lets you test a prompt right on the page. Three ways to start • Hover over an image and click the Image Lens button in its corner. • Right-click and choose "Analyze this image with Image Lens". This also works for images behind transparent overlays, CSS background images, canvas and video frames. • Press Alt+Shift+S, click the toolbar icon, or choose "Select an area to analyze with Image Lens" from the right-click menu, then drag a box around any part of the page. What you get • A full prompt and a short prompt in English, with an optional Simplified Chinese version. • A breakdown into nine aspects: subject, environment, style, lighting, color, camera, composition, material and rendering. Click any aspect to copy just that part. • Descriptions of what is actually visible: precise colors, lighting, framing and textures. Legible text in the image is copied exactly. • Edit the prompt in the panel and copy it with one click. Test generation • Generate an image from the prompt without leaving the page. The aspect ratio defaults to the source image's shape, or choose 1:1, 3:4, 4:3, 9:16 or 16:9. • Download the results or open them in a full-page viewer. Bring your own API key • Works with Google Gemini (including Nano Banana image models) and any OpenAI-compatible service. Analysis and generation can use different providers. • Image Lens does not include, sell or host API keys. Add your own key in Settings. Privacy • Your settings and API keys are stored only in your browser. • An image is sent to the provider you configured only when you start an analysis. Large images are resized locally first. The page address is not sent to the provider. • When a site blocks direct image access, or when you select an area, Image Lens takes a screenshot of the visible tab locally and sends only the selected part. • The developer runs no server and never receives your API keys, images, prompts or generated images. The interface is available in English and Chinese, with light and dark themes that follow your system. 图透镜 Image Lens 可以把你在网页上看到的图片，反推成详细、可编辑的 AI 生图提示词，并在当前页面直接测试生图效果。 三种打开方式 • 鼠标移到图片上，点击图片左上角的图透镜按钮。 • 在图片上右键，选择「用图透镜分析这里的图片」。被透明遮罩盖住的图片、CSS 背景图、canvas 和视频画面也能识别。 • 按 Alt+Shift+S、点击工具栏图标，或在右键菜单选择「用图透镜框选区域分析」，拖动框选页面上任意区域进行分析。 识别结果 • 完整版和精简版两种提示词，英文为主，可同时生成简体中文版本。 • 按主体、环境、风格、光线、色彩、镜头、构图、材质、渲染九个维度拆解，点击任一维度即可单独复制。 • 只描述画面里看得到的内容：具体的色调、光线、构图和材质；图中可辨认的文字会原样照抄。 • 可以在面板里直接修改提示词，一键复制。 立即测试生图 • 不离开当前页面就能用提示词生成图片。比例默认匹配原图，也可选择 1:1、3:4、4:3、9:16、16:9。 • 生成结果可以下载，也可以在新页面中查看。 使用你自己的 API Key • 支持 Google Gemini（包括 Nano Banana 生图模型）和 OpenAI 兼容接口。识图和生图可以分别使用不同的服务商。 • 插件不内置、不出售、也不托管任何 API Key，需要在设置页填写你自己的 Key。 隐私 • 设置和 API Key 只保存在你的浏览器本地。 • 只有在你主动发起识别时，图片才会发送给你配置的服务商；较大的图片会先在本地缩小。网页地址不会发送给服务商。 • 当网站限制直接读取图片，或你使用框选功能时，插件会在本地截取当前可见页面，只把对应区域发送给服务商。 • 开发者没有服务器，不会收到你的 API Key、图片、提示词或生成结果。 界面支持中文和英文，浅色、深色外观跟随系统自动切换。