Generate and edit images with GPT Image 2.5 Flare from a prompt and up to 14 references, in HD, 2K or 4K, then choose Flare or Flare High for the finish the job needs.
Try an example
Breakfast tables, product shelves, illustrations, story backgrounds and character art. Hover or focus an image to read the prompt behind it; open it to reuse the prompt in the creator above.
Overhead photograph of a breakfast table: a bowl of yogurt with berries, a glass of orange juice and a croissant on a linen napkin, soft morning light, gentle shadows, no text.
Sunlit home office with a wooden desk, a laptop and a green plant beside a large window on the left, warm neutral colors, wide-angle photograph, no people, no text.
Misty pine forest at dawn with a narrow trail, soft golden light through the trees, calm cinematic mood, no people, no text.
Ceramic coffee mug on a light oak shelf beside a small green plant, soft window light from the left, calm neutral palette, clean product photograph with empty space on the right, no text.
Close photograph of a bowl of ramen with a soft-boiled egg, sliced scallions and rising steam, dark wood table, warm side light, shallow depth of field, no text.
Flat vector-style illustration of a lighthouse on a rocky coast at sunset, layered orange, pink and navy sky, simple shapes, no text.
Friendly fox character sitting on a stack of books, soft 3D render, pastel colors, rounded shapes, plain light background.
Vertical soft gradient background in peach and lilac with a few paper-cut leaves along the edges, large empty space in the center for a caption, no text.
Each image was generated with GPT Image 2.5 Flare from the prompt shown on its card.
Start with a brief for a new image, or a reference for a specific change. Decide the composition around where the image will appear.
Give GPT Image 2.5 Flare three decisions: what to show, how it should look and where the empty space sits. Keep the headline for the page layout, where you can check every letter, and describe only the quiet area the text will occupy. The image on the right was generated from this brief with GPT Image 2.5 Flare.
Cozy cafe window seat with a latte and an open notebook, warm afternoon light, soft film look, one focal object, empty area at the top for a headline, no text.
Upload a supported image and say what should stay recognizable and what should change. GPT Image 2.5 Flare treats a reference as guidance, not as an exact copy, so compare the bottle shape, the cap and the shadows with the source before you use the result.
Three actions in the creator above.
Write the subject, style and composition in the prompt box, up to 4,000 characters. Add references only when each one has a job, such as a product shape or a room layout. The creator accepts up to 14 in total.
Open the model menu to pick GPT Image 2.5 Flare, or its High variant when the image needs more finish. Then pick the aspect ratio for where the image will appear, HD, 2K or 4K, the number of images and PNG or JPEG. Start at HD while you compare directions.
Review the actual image request before you confirm. The Generate button shows the estimate: 5 credits per HD image, 9 per 2K and 15 per 4K on Flare, and 18, 33 and 60 on Flare High, multiplied by the number of images. Check the result at its real display size, then download it or send it to an editing tool.
Four jobs where the fast Flare model, and its High variant, earn their place.
01
OpenAI calls GPT Image 2.5 Flare its fastest model for high-quality, everyday image generation. Use it for the images you make often, such as breakfast tables, product shelves and story backgrounds, and ask for two or four per request to compare directions before you settle on one.
02
Attach a product, a room or a palette and say what each reference is for. GPT Image 2.5 Flare accepts text and image inputs and reads references as guidance, not as an exact copy, so compare shape, caps and labels with the source.
03
GPT Image 2.5 Flare High is the same model at a higher quality setting. It costs 18 credits at HD against 5 for Flare, so test the idea on Flare and rerun the closest direction on Flare High. A higher setting is not better for every brief, so compare both on your own prompt.
04
Each ratio has a fixed size at every tier: a 16:9 image is 1280 by 720 at HD, 2560 by 1440 at 2K and 3840 by 2160 at 4K. OpenAI marks 4K as experimental on this model, so check a 4K result against your brief before you rely on it.
Both GPT Image models and Nano Banana accept text and reference images and offer HD, 2K and 4K. They differ in price and focus, so compare them on one brief before you choose. GPT Image 2.5 Flare High is the Flare column's model at a higher quality setting, for 18, 33 and 60 credits.
| GPT Image 2.5 Flare | GPT Image 2.5 Sunburst | Nano Banana | |
|---|---|---|---|
| Best for | Fast, high-quality everyday image generation and editing | Precise image editing with strong instruction following | Fast, capable image generation and editing |
| Resolution | HD · 2K · 4K | HD · 2K · 4K | HD · 2K · 4K |
| Reference images | Up to 14 | Up to 14 | Up to 14 |
| Credits per image | 5 HD · 9 2K · 15 4K | 5 HD · 9 2K · 15 4K | 23 HD · 34 2K · 51 4K |
Generate another image to test a different direction. Edit the one you have when the subject works and one part needs attention.
GPT Image 2.5 Flare is OpenAI's fastest model for high-quality, everyday image generation, with the model ID gpt-image-2.5-flare. It accepts text and image inputs and produces images. In Potto AI you choose it in the generator, with up to 14 reference images and HD, 2K or 4K output, from 5 credits per image.
GPT Image 2.5 Flare High is the same gpt-image-2.5-flare model run at its high quality setting instead of medium. It appears as a variant in the generator's model menu, not as a separate model, and keeps Flare's 14 references, ten ratios and three tiers at 18, 33 and 60 credits per image. Potto fixes the quality setting for each variant, so you pick the variant rather than a setting.
Flare is the fast GPT Image model for everyday images, while OpenAI positions Sunburst for workflows where editing precision matters most. In Potto AI both run at medium quality, accept up to 14 references and cost 5, 9 and 15 credits per image at HD, 2K and 4K, so run the same brief through both and compare the results.
GPT Image 2.5 Flare costs 5, 9 and 15 credits per image at HD, 2K and 4K, while Nano Banana costs 23, 34 and 51. Both accept up to 14 references and the same three tiers. They are different models with different strengths, so use one brief, the same references and the same ratio, then judge the two results against your criteria.
Everyday images from a prompt, with or without references: food and interior photographs, product shots, flat illustrations, character art and backgrounds with room for a headline. Describe the subject, the style and where the empty space should sit, and keep any headline in your page layout so you can check every letter.
Up to 14 in total. Add a reference only when it has a specific job, such as supplying a product, a room or a palette, and say so in the prompt. Several images do not guarantee that every detail is preserved, so compare the caps, labels and proportions with each source.
HD, 2K and 4K output, ten aspect ratios and PNG or JPEG. Each ratio has a fixed pixel size at every tier: a 1:1 image is 1024 by 1024 at HD, 1920 by 1920 at 2K and 2880 by 2880 at 4K. The generator marks 4K as experimental on this model, so results may vary and deserve a check against your brief.
When your closest draft deserves a higher quality setting. Test directions on GPT Image 2.5 Flare at 5 credits for an HD image, then rerun the strongest prompt on Flare High at 18. A higher setting does not guarantee a better image for every brief, so compare both results before you spend more.
No. Verify every word and compare the important reference details in the actual result. Keep approved wording in editable layout text. A higher tier, several references or the Flare High variant do not establish correct lettering, exact product fidelity or an identity lock.
Name the one unresolved issue, such as space for text or the lighting direction, and write the next instruction around it while keeping the working parts of the brief. To change one area of a finished image, use the AI Image Editor; a prompt to change one object is not an exact mask, so check the rest of the picture.
You can write a prompt, add references and choose GPT Image 2.5 Flare settings before signing in. Generate opens the sign-in dialog, and everything you entered stays in the creator.
The task explains why it stopped and offers a retry, and its credits are released. Stop a running request with the cancel control, since leaving the page does not stop it. Generating again after a finished image is a new request.
GPT Image 2.5 Flare costs 5 credits per image at HD, 9 at 2K and 15 at 4K. Flare High costs 18, 33 and 60. More images multiply the price, and the Generate button shows the total before you confirm.
Describe the image, add references only where they have a job, and review the quote before you generate. Flare is the fast, lower-priced way to test an idea, and Flare High is there when the closest draft needs more finish.
