OpenAI launched ChatGPT Images 2.0 with reliable in-image text rendering, multi-panel generation, and a new gpt-image-2 API with tiered pricing.
OpenAI released ChatGPT Images 2.0 (gpt-image-2), a major upgrade to its image generation model that finally renders legible, accurate text within images. The model includes 'thinking capabilities' — web search, multi-image generation from one prompt, and iterative editing. All ChatGPT and Codex users get access starting Tuesday, with paid users unlocking higher-quality outputs. The gpt-image-2 API is also available with quality- and resolution-based pricing.
The gpt-image-2 API drops the single biggest blocker for programmatic image generation: legible text. Any pipeline that previously required post-processing to overlay text on images — receipts, menus, infographics, certificates, banners — can now be simplified to a single API call. Pricing is tiered by quality and resolution, which means you can tune cost vs. output fidelity per use case.
Hit the gpt-image-2 API this week with a prompt requiring embedded text — like a product label or event flyer — and compare output quality and cost against your current Stable Diffusion + text overlay pipeline to see if you can collapse two steps into one.
Suggested exercise · not independently tested. Access requirements and costs may vary.
Run: curl https://api.openai.com/v1/images/generations -H 'Authorization: Bearer $OPENAI_API_KEY' -H 'Content-Type: application/json' -d '{"model": "gpt-image-2", "prompt": "A professional restaurant menu for a Mexican restaurant with accurate English text, prices in USD, and clean typography", "n": 1, "size": "1024x1024"}'
Download the returned image URL and inspect the text accuracy, font rendering, and layout quality
Check the usage field in the response to calculate cost per image at your target resolution
Tags
Sources
Related