GPT Image 2.5 Flare

What it does

Flare covers two workflows: text to image, and image editing with specific instructions. It’s the quick option in the GPT Image 2.5 family—ideal when you need many solid frames in a day.

Upload one reference to restyle it, or several to blend subjects, styles, and details into a single result. Reference images are handled at high fidelity automatically.

Quality settings

Choose low, medium, high, xhigh, max, or auto. Lower tiers are faster and use fewer credits; xhigh and max add detail for finals. Draft on low, then step up when the frame is worth polishing.

Key capabilities

• Photorealism — natural light, believable materials, rich texture
• Text in image — dense copy, small type, UI mockups and layouts
• Precise edits — change one part while keeping face, pose, and lighting
• Style control — keep a look consistent, or transfer style with a short prompt

Use cases

• Generation — marketing visuals, logos, mockups, scenes, comics
• Editing — style transfer, try-on, product shots, text fixes, cleanup, compositing
• Characters — multi-page sets where the same figure stays recognizable

How to get good results

• Be specific — “soft coastal daylight” beats “make it better”
• Use photo language — lens, light, and framing (e.g. 50mm, shallow depth of field)
• Lock what must stay — “change only lighting; keep face, pose, and clothing”
• Put on-image text in “quotes” and describe the typography

Controls on Imagine Weave

• prompt — what to generate or how to edit
• input images — one or more references
• aspect ratio — 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16, or auto
• quality — low · medium · high · xhigh · max · auto
• count — up to 10 images per run
• format — webp, png, or jpeg
• background — auto, transparent, or opaque
• moderation — auto (default) or low