Image to Text
Pull text from or about an image — captions, descriptions, Q&A, structured data, color extraction, copyright fingerprints. Today: Florence-2 Large as the captioning specialist, Llama 3.2 Vision and Qwen3-VL for open-source vision Q&A, plus utility models for color-palette extraction and the upcoming IP Copyright Detector.
Florence-2 Large is Microsoft's specialist image-captioning model — pass one image, get a text description back. Pick the caption verbosity (`Caption` for alt-text, `Detailed` for a sentence, `More Detailed` for a paragraph). Fast, cheap, ideal for batch metadata work on product catalogues.
Extract the dominant colour palette from an image using ColorThief. Returns a JSON list of hex and RGB values ordered by prominence.
**COMING SOON.** Flag designs that may infringe a trademark or copyrighted brand — logo similarity, recognisable characters, registered marks — before they reach a customer order.
**COMING SOON.** Upload a design and get a structured report — type (photo / illustration), DPI, palette, possible IP issues, metadata, recommended next steps.
Gemini 3.7 Flash is Google's workhorse multimodal LLM: send a prompt plus up to 11 images and get back copy, structured JSON, captions or analysis. Its sharpest edge over earlier Flash models is how closely it follows a brief — visual references, brand rules and dense documents — with a reasoning dial you set per call, from quick low-effort answers to careful multi-step work.