Anyscene AIOpenAI's newest image model for creating and editing: readable text, faithful edits from up to 16 reference images, 1K to 4K.
Try an example
Real results from this generator: the prompt exactly as it was sent, the settings, and what the same generation costs today.
Both lines of text came out spelled correctly and where the prompt put them, over a screen-print pagoda, cherry blossoms and a low setting sun.
Prompt
A vintage travel poster for Kyoto: a five-storey pagoda among cherry blossoms at sunset, the title "KYOTO" in bold letters at the top and "Cherry Blossom Season" at the bottom, flat screen-print style
We gave it the Nano Banana Pro reading nook and asked for night and snow. The reader, the cat, the books and the mug all stayed where they were; only the light and the weather changed.
Prompt
The same room at night: the lamp is the only light, snow falls outside instead of rain, keep everything else the same
The screen shows Lisbon, 24°C and Sunny as asked, and fills in an hourly strip and a five-day forecast by itself. Behind the phone, a yellow tram and the river match the city.
Prompt
A weather app screen on a phone held in one hand, showing "Lisbon", "24°C" and "Sunny" on clean rounded cards, a blurred sunny street behind
The headline came out spelled correctly in a clean serif, with the cup, the steam and the soft window light composed like a café poster. Nothing was edited afterwards.
Prompt
A minimalist poster that reads "SLOW MORNING" above a steaming coffee cup
GPT Image 2.5 is OpenAI's image generation and editing model, the successor to GPT Image 2. This version is built for sharper images, more natural light and texture, and edits that stay true to the subject you started with, even after several rounds.
On Anyscene you can start from a prompt or from up to 16 reference images, choose Auto or one of eight aspect ratios, and render at 1K, 2K or 4K for 10, 16 or 26 credits.
Headlines, labels and short slogans render as real text. Quote the exact words in the prompt and say where they go.
Change the background, the light or the outfit while the person or the product stays recognisable.
Give it the logo, the product and the mood board in one request, and name what each image is for.
Work out the prompt at 1K for 10 credits, then render the keeper at 2K or 4K.
Resolutions, lengths, modes and prices side by side, taken from the settings the generator offers.
| Feature | GPT Image 2.5 | GPT Image 2 | GPT Image 1.5 |
|---|---|---|---|
| Credits per image | 10–26 credits | 8–22 credits | 6 credits |
| Resolution | 1K / 2K / 4K | 1K / 2K / 4K | 1K |
| Aspect ratios | 8 ratios + Auto | 1:1, 9:16, 16:9, 4:3, 3:4 + Auto | 1:1, 3:2, 2:3 |
| Reference images | Up to 16 | Up to 1 | Up to 1 |
| Available on | Any plan or credit pack | Any plan or credit pack | Any plan or credit pack |
Credits cover one image at each resolution the generator offers. The Generate button shows the exact price before you start.
All 28 models run in the same generator on the same credits, so you can switch whenever a shot needs something else.
Every plan unlocks all 28 models and every tool on Anyscene, this one included. Yearly billing saves 20%, and a credit pack tops up any plan.
Yearly prices, shown per month. Monthly billing and credit packs are on the pricing page.
What people ask before they generate with GPT Image 2.5 on Anyscene.
OpenAI's image model for generating and editing images, and the successor to GPT Image 2. On Anyscene it works from a prompt, or a prompt plus up to 16 reference images, at 1K, 2K or 4K.
10 credits per image at 1K, 16 at 2K and 26 at 4K. It unlocks with any plan or credit pack, and the Generate button shows the price for your settings before you start.
GPT Image 2.5 offers more aspect ratios (eight plus Auto, against five plus Auto) and takes up to 16 reference images, where GPT Image 2 takes one here. It costs 2 credits more at each resolution: 10, 16 and 26 against 8, 14 and 22.
Readable text is one of the GPT Image line's strengths, and the poster above was spelled right on the first try. Put the exact words in quotes, keep them short, and check the spelling before you publish.
Yes. Add your photo as a reference and describe the change, such as a new background or different lighting. Say what must stay the same, like the face or the label, so the edit keeps it.
1K while you work out the prompt, 2K for web pages and social posts, and 4K for print or large banners.
The generator at the top of this page is already set to it. Write a prompt, check the price on the button and generate.
