Powered by GPT Image 2.5 Flare from OpenAI · Independent, unofficial tool
GPT Image 2.5 Online Generator & Editor
GPT Image 2.5 is OpenAI's newest image model, out since September 8, 2026, in two versions: Flare for fast everyday images and Sunburst for precise edits. This tool runs Flare at medium quality for 2 credits an image, with no subscription. On the same photo and prompt it finished in 18 seconds, where GPT Image 2 took 34.
Images to edit (optional, up to 4)
JPG, PNG or WebP, up to 10 MB each. 3 or 4 photos: +1 credit. Leave empty to make a new image.
Size
about 1 megapixel · 2 credits
BeforeMade with GPT Image 2.5 Flare at medium, this tool's setting, from one photo. 18 seconds, first try.
Your own image will show up here.
Flare or Sunburst?
- Flare is OpenAI's quickest image model, made for everyday pictures at high quality. It runs this tool.
- Sunburst is OpenAI's most capable model, meant for work where editing precision matters most. It isn't on this site yet.
Both versions came out on September 8, 2026, and both take a photo, several photos or just a sentence.
GPT Image 2.5 Flare vs GPT Image 2, same photo and prompt
On October 11, 2026 we gave both models one photo of a house on a hill and the same poster prompt, at medium quality and 3:4, one try each. Both kept the photo on top, drew the house as a cartoon below and spelled every word right. Each picked its own headline: GPT Image 2 wrote BIGGER DREAM, Flare wrote GOOD DAY. Flare was done in 18 seconds, GPT Image 2 in 34. Its page: GPT Image 2.
Running film stills from one prompt, made in ChatGPT
One of our favorite prompts: three stacked frames of someone running through a wide field, close, far and close again, like stills from one moving shot. We wrote and tested it in ChatGPT, where these three were made. It needs no photo. You can run the same prompt here with 2.5 Flare, in English below, plus the lines we added for a wedding teaser.
The prompt
Make three stacked frames of the same subject crossing a wide landscape, each like a still pulled from one running shot. Frame 1: close to the subject, partly hidden by the surroundings, with fabric or hair trailing behind. Frame 2: pull far back so the subject is small in a big field, with a low horizon and open sky giving a sense of speed. Frame 3: close again, with big blurred foreground shapes sweeping past the edge of the frame, as if brushing past the viewer. Fill the scene with many small repeated natural elements that fit the theme: soft out-of-focus blobs in front, a dense but readable rhythm in the middle, a low-contrast distance. The mood is a bright, clean, windy dusk. Colors: a large cool, muted sky or open background; a deeper plant or ground tone in the middle to hold the frame; a few small bright warm dots in front and in the middle; the subject light and airy, lightly tinted by its surroundings. Soft light that never turns gray, slight motion blur at the edges, a little film grain and gentle highlight bloom, like frames from a film, not a sharp posed photo. No large text or logos.
The wedding lines, added at the end
For a wedding teaser: teal-green grass and deep ink-green ground, with coral orange and cream dusk dots. Keep the far shot in the middle frame as the anchor, with blurred foreground curving up from the bottom. Put a large title across the bottom left and small lines of detail along the bottom edge, low in contrast. Portrait 3:4.
What it costs
| Shape | Pixels | Credits per image |
|---|---|---|
| 3:4 | 768 × 1024 | 2 |
| 1:1 | 1024 × 1024 | 2 |
| 2:3 | 1024 × 1536 | 2 |
| 4:3 | 1024 × 768 | 2 |
| 3:2 | 1536 × 1024 | 2 |
- With 3 or 4 photos attached, an image takes one more credit than the table shows.
- Packs cost $4.90 for 20 credits or $9.90 for 70, paid once: up to 10 or 35 GPT Image 2.5 images, about $0.28 each on the bigger pack (fewer with 3 or 4 photos). No subscription, and credits don't expire.
- If an image fails, its credits go straight back to your balance.
- See credit packs
Questions
Is this the official GPT Image 2.5?
No. GetStepSnap is an independent service, not affiliated with or endorsed by OpenAI. We run OpenAI's GPT Image 2.5 Flare model through an API provider and charge credits for the images you make here.
What's the difference between GPT Image 2.5 Flare and Sunburst?
OpenAI presents Flare as its quickest model, made for everyday pictures at high quality, and Sunburst as its most capable one, for work where editing precision matters most. Both came out on September 8, 2026. This tool runs Flare.
How much does a GPT Image 2.5 image cost here?
2 credits, in any shape. Attaching 3 or 4 photos adds one more. Packs cost $4.90 for 20 credits or $9.90 for 70, paid once, so the 70-credit pack makes up to 35 images.
Is GPT Image 2.5 free here?
No. Every image here is paid with credits: 2 for a GPT Image 2.5 image. There's no subscription, and credits don't expire.
Should I use GPT Image 2.5 or GPT Image 2?
Both take the same credits here. On the same photo and prompt, Flare finished in 18 seconds and GPT Image 2 in 34, and both spelled every word on the poster right. Pick 2.5 for speed; GPT Image 2 has its own page if you want that model.
How big are the images?
About 1 megapixel: 768 × 1024 for 3:4, 1024 × 1024 for 1:1 and 1024 × 1536 for 2:3, with 4:3 and 3:2 the same sizes turned on their side. Every image runs at medium quality.
What happens to the photos I upload?
We store them so the model can read them and so your results stay in your account. We don't sell them or use them for anything else. To have them deleted, email support@getstepsnap.com.
What if an image fails or the model refuses it?
Credits spent on a failed or refused image go straight back to your balance. You only pay for images that finish.
More AI photo tools
GetStepSnap is an independent product and is not affiliated with, endorsed by, or sponsored by OpenAI. GPT Image and ChatGPT are OpenAI's names for its products.