Refract - Image to Prompt

Refract turns any image you see on a webpage into a reusable image-generation prompt. Pick an image straight from the page (or right-click it), and Refract reads its subject, composition, lighting, color, materials, camera treatment, and visual medium — then streams back a well-structured prompt in Chrome's side panel, preserving the source image's aspect ratio. Edit it, copy it, and use it with your favorite image generator. TWO READING MODES • Faithful — reconstruct the exact image: every visible element, ordered by visual weight. • Style match — keep the environment, lighting, mood, and artistic style, but free the subject, so you can create something new inside the same visual world. PERSONALIZATION Save a persistent profile with your own subject photo and a fixed direction. Refract automatically rewrites every prompt around your subject, and lets you compare, edit, and copy the Base and Personalized versions side by side. BRING YOUR OWN AI Refract generates prompts, not images, and has no backend of its own. Connect the vision-capable AI provider you already use — official APIs, gateways, OpenAI-compatible endpoints, or fully local models via LM Studio / Ollama. Refract loads the provider's model list and shows only vision-capable models. PRIVACY • Your API key and settings are stored locally in your browser. • Selected images are sent only to the AI provider you configured — nowhere else. • No account, no tracking, no Refract server. HOW IT WORKS 1. Open Refract and configure a provider (one-time). 2. Click "Select from page" and pick any image, or right-click an image. 3. Watch the prompt stream in live. Stop, retry, or switch modes any time. 4. Copy the prompt into your image generator. Requires Chrome 116+, and an API key for a supported provider or a local model server.