Gpt Image 2

Wraps OpenAI's GPT Image 2 model through RunComfy's CLI so you can generate and edit images without managing API keys. The real draw is embedded text rendering: it handles logos, multilingual typography, and multi-element layouts more reliably than Flux or SDXL. Three fixed sizes only (1:1, 2:3, 3:2), and the edit endpoint lets you pass up to 10 reference images with natural language instructions like "keep the person's face, swap the background to studio white." Best for e-commerce mockups, localized ad variants, or any case where what's in the frame matters more than heavy stylization. Prompts that quote exact text and specify one style anchor tend to work best.