what is pagewright fr?

Pagewright is a tool that builds the long, scrolling product detail images used on shopping sites like Taobao, Tmall, Amazon, and Shopify. You give it either a product page URL or your own photos plus a written description, and it produces a finished detail page image, optionally shown in both English and Chinese side by side. Its core idea is that a language model should never be asked to paint the actual picture. Instead of using an image generating AI to draw the page directly, which tends to garble prices, mangle text, and invent logos that do not match the real brand, Pagewright uses a language model only for the parts language models are good at: reading a source page, organizing facts into a structured record, writing description text, and translating it. That structured record then fills a template written in HTML, using your real product photos and real logos rather than anything an image model invented. A headless browser takes a screenshot of that finished HTML page, which is then automatically cropped and sliced into the long panel images you upload to a store listing. Because the whole page is built from real text and real image files rather than pixels painted by a model, changing one price or spec later just means editing that one field and re rendering, with no need to regenerate a whole new image. Fetching product information from a store page is handled by a layered system with no site specific code: it first tries a plain, lightweight request, and automatically falls back to a full headless browser if the page needs JavaScript or blocks simple requests, as some sites protected by services like Akamai do. There are two ways to use it. A desktop application, available for macOS and Windows, needs no coding knowledge: you upload photos, type or paste a description, choose a language, and click generate. A command line and library version is aimed at developers and can be scripted, and includes a bundled example you can render without needing an API key of your own. Every generation is automatically saved locally with its full image, individual panels, structured data, and a thumbnail, so past versions can be reviewed or re rendered later. The project is released under the MIT license.