Create with
GPT Image 2.5.

What GPT Image 2.5 does here

ShimmerPic runs both models OpenAI released on September 8, 2026 — Flare, the fast default, and Sunburst, the precision model — in the browser, with no API key. Write a prompt or open one of 531 community prompts, pick a ratio from 1:1 to 9:16, choose 1K to 4K, add up to three reference photos, and generate. Every result is private until you download it.

How it works

1. Start from something that works. The library has 531 community prompts sorted by what they make — product shots, posters, characters, brand boards — each shown with its actual output. Open one and the text is filled in for you.
2. Change what matters. Swap the product, the headline, the setting. Add a reference photo if the brief depends on a real object or person you have permission to use. The price updates as you change settings, before you generate.
3. Generate, then refine. Try Flare first while you are exploring, switch to Sunburst for the version you keep, and edit any result by opening it as a reference and describing the change. Everything stays in your creations, ready to download.

What people make with it

The model is general; the prompts are not. These are the four jobs the library is deepest in, each with its own guide and starting prompts:

• Product photography — upload one photo of the product and describe the surface, the light and the camera. The product stays recognisable; the studio is rebuilt around it. Packshots, lifestyle scenes, hero crops.
• UGC-style visuals — the product in a hand, on a desk, in a car cupholder, shot like a phone photo rather than a studio. Ten variations for an ad test in an afternoon.
• Posters and flyers — quote the headline and the small print, say where each sits, and the model renders the words in the layout you describe. Event one-sheets, sale flyers, typographic prints at print resolution.
• Mockups — your logo or label as a reference, placed on a T-shirt, a box, a tote or a wall in the angle and light you ask for, without a template library.

Sizes, settings and limits

Six aspect ratios: 1:1, 4:5, 3:2, 2:3, 16:9 and 9:16. Three resolution tiers, 1K, 2K and 4K, in Standard or High quality. Up to three reference photos per image, each up to 10 MB and 4,096 px on the long side. Your prompt is submitted in full without silently cutting it. Every result is saved privately to your creations, in full resolution, and can be downloaded or opened again as a reference for the next edit.

GPT Image 2.5 Flare vs Sunburst

Same credits, different trade-off. Flare is faster; Sunburst adds detail and holds complex layouts together. We ran the same community prompts through both, one run each, and put the results side by side — see the comparison.

What an image costs

Identical for both models. Each reference photo adds 3 credits. Credits come in one-time packs from $2.99 or monthly and annual plans; see pricing. A generation that fails technically is refunded; one turned down by a content check keeps its credits. There is no unlimited tier.

Common questions

Where the prompts come from

The library is curated from the open awesome-gpt-image-2 collection and shows each prompt next to the image its author shared. Every card credits the author and links to the original post, and every page has a report link; if you are the author of a prompt or an image and want it removed or credited differently, use it and we act on it. The images are community examples, not a guarantee that your run will match them.

Model facts from OpenAI, "Introducing ChatGPT Images 2.5". ShimmerPic is an independent creative tool, not affiliated with OpenAI.