It removes the need to manually recreate or heavily retouch product photos by allowing fast generation or targeted edits directly from prompts and reference images, saving time and reducing reliance on graphic designers.
GPT Image 2.5 is an OpenAI‑based image model that creates new pictures from text prompts and modifies existing images using reference uploads. The service offers two specialized models: Flare, which prioritizes generation speed for draft‑level output, and Sunburst, which allocates more compute to produce higher‑detail results and finer edit control. Users select a mode (Text‑to‑Image or Image‑to‑Image), provide a prompt describing subject, composition, style and any constraints, optionally attach up to 16 reference files (PNG, JPG, WebP ≤10 MB), and set parameters such as quality tier, aspect ratio, background type and output format. The backend calculates credit cost based on the chosen quality level and deducts it before rendering. Generated images can be delivered as PNG with optional transparency, and the system preserves key elements from references to maintain subject fidelity during edits.
Have feedback for the maker?
Sign in to leave a review, report a bug, or suggest a feature.
Leo poppy
