Description
Build an AI App That Generates Branded Social Media Images from Any Website Project overview I'm looking for an experienced developer to build an application that turns any business website into on-brand, publication-ready social media posts (single images, carousels, quote cards) — generated automatically, every day. You enter a website URL; the app extracts the brand's visual identity from it and produces designed graphics + captions that look like the brand's own designer made them. What the app must do Crawl the website — pull text + images from multiple pages (homepage, product/case-study pages) using a custom crawler or the Firecrawl API. Extract a brand kit — brand colors (hex, ranked by CSS usage), fonts, and logo parsed from the site's HTML/CSS, plus a full-page screenshot (e.g., via the mShots API) as a visual style reference. Label site images with AI vision (e.g., GPT-4o-mini) so the system knows which assets are product mockups, photos, or logos. Synthesize a per-site design system with an LLM — color roles, typography, imagery style, decoration motifs, layout preferences. Every website must get its own unique look, never a shared template style. Generate images with Google Nano Banana Pro (gemini-3-pro-image-preview, 2K output), using: reference images passed with roles — the website screenshot as style reference, product shots as subject references, logo separate JSON design-token prompts (exact hex palette, mood, negative-space rules) carousel consistency — sl