It eliminates the need for separate graphic design software by letting developers generate or modify images directly from prompts and reference files in a single web interface, while keeping costs predictable through credit packs.
GPT Image 2.5 is an online AI service that creates new images from text prompts and edits existing images using reference pictures. It runs two OpenAI‑based models: Flare, optimized for fast generation, and Sunburst, which prioritizes higher visual detail and precise edits. Users pick a model, select Text‑to‑Image or Image‑to‑Image mode, upload up to sixteen PNG, JPG or WebP references (max 10 MB each), and configure quality (Auto, Low, Medium, High, XHigh, Max), aspect ratio, background (transparent or opaque) and output format (PNG). The system consumes a credit amount tied to the chosen quality level and returns a PNG that can include transparency. The workflow consists of model selection, prompt composition describing subject, composition, style, lighting and constraints, optional reference assignment, settings adjustment, and generation with a preview of credit cost before the image is produced. Credits are sold in one‑time packs that never expire, with a small free allocation on sign‑up.
Have feedback for the maker?
Sign in to leave a review, report a bug, or suggest a feature.
Leo poppy
