GPT Image 3.0GPT Image 3.0

GPT Image 3 workspace

GPT Image 3 Generator

Describe the image you want, add a reference photo if it has to look like something real, and get a finished frame in seconds. The day GPT Image 3 drops, it opens here.

Text to image, reference editing, readable text, up to 4K

Try a prompt

8 free credits, no card · failed generations are refunded

Poster with a headline
Portrait
Move a person to a new scene
Packaging concept

What is GPT Image 3?

GPT Image 3 is OpenAI's next image generation model. OpenAI has not published a release date, an API id or pricing for it yet, so there is nothing to install or subscribe to in advance.

Until it ships, this workspace runs GPT Image 2.5 — OpenAI's current image model — so you can create today and keep the same account, credits and history when 3.0 arrives.

Independent third-party workspace · Last updated

Examples

Made with GPT Image 2.5, with the prompt that made it

Every image below was made with GPT Image 2.5, the model this workspace runs today. Open any card to reuse its prompt, or copy it and change one detail.

Poster with a headlineModel · GPT Image 2.5
Open in generator

Poster with a headline

Type inside the frame, not on top of it: a headline, a date line and a small credit block that all stay legible.

Prompt

Editorial poster for a late-night listening party. Large serif headline in the upper half reads exactly "OPEN LATE" on one line. Below it a thin sans-serif date line reads exactly "SAT 21.06". Bottom-left small credit block reads exactly "DOORS 23:00 · STUDIO 9". Warm amber and deep teal palette, subtle paper grain, generous margins, flat print design, no other text.

Product shotModel · GPT Image 2.5
Open in generator

Product shot

A clean studio frame with believable shadow and reflection — the kind a product page needs before a photographer is booked.

Prompt

Studio product photograph of a matte black ceramic espresso cup on a travertine slab, soft window light from the left, gentle falloff, subtle contact shadow, pale warm background, no text.

InfographicModel · GPT Image 2.5
Open in generator

Infographic

A four-step diagram with consistent icon weight, aligned labels and enough white space to stay readable on a phone.

Prompt

Clean isometric infographic of a four-step onboarding flow. Four rounded nodes arranged left to right, numbered exactly "01", "02", "03", "04", connected by thin arrows. Each node has a simple 2px line icon and a one-word label: "Create", "Connect", "Invite", "Launch". Pastel lavender and cream palette, generous white space, flat vector look, crisp legible type, no other text.

PortraitModel · GPT Image 2.5
Open in generator

Portrait

Skin texture, catchlights and a shallow depth of field, without the plastic look that makes an AI portrait obvious.

Prompt

Close-up portrait of a woman in her thirties in golden-hour light, visible freckles and fine skin texture, catchlights in the eyes, shallow depth of field, natural colour, no retouching, no text.

Restage a real productModel · GPT Image 2.5Reference: Restage a real product
Open in generator

Restage a real product

The reference photo keeps the cup, handle and glaze. Only the surface, light and background change.

Prompt

Place the same cup from the input image on a pale oak table in a bright Scandinavian kitchen. Keep the cup shape, handle and matte speckled glaze exactly the same. Change only the surface, background and light: soft morning daylight from the left, subtle contact shadow. No text.

Move a person to a new sceneModel · GPT Image 2.5Reference: Move a person to a new scene
Open in generator

Move a person to a new scene

Face, hair and expression stay the same while the street, weather and light around them are replaced.

Prompt

Put the same person from the input image on a rainy city street at night. Keep her face, hair, expression and pose unchanged. Change only the surroundings: neon signs, wet pavement reflections, light rain, shallow depth of field. No text.

Social creativeModel · GPT Image 2.5
Open in generator

Social creative

A square ad frame with a readable offer, deliberate negative space and room for a logo.

Prompt

Square social ad for a coffee subscription. Bold sans-serif headline in the upper left reads exactly "FIRST BAG FREE". A single matte bag of coffee beans sits on a deep green background, soft studio light. Generous empty space in the top right for a logo. Clean modern layout, only the headline text, no other words.

Packaging conceptModel · GPT Image 2.5
Open in generator

Packaging concept

A shelf-ready label with small print, a barcode block and a consistent brand colour across the set.

Prompt

Packaging concept photograph for a small-batch olive oil. Tall matte dark-green bottle with a cream paper label. On the label, small serif type reads exactly "OLIO NUOVO" and below it "HARVEST 2026". A barcode block sits at the bottom of the label. Warm studio light on a stone surface, shallow depth of field, no other text.

Prompts are shown as they were sent to the model. The model label on each card is the model that produced it.

Why this workspace

Four things you cannot get from a screenshot

Not a gallery with a waitlist. A working generator, built around the model that is actually available today.

01

OpenAI's newest image model, live now

Every prompt runs on GPT Image 2.5, the current OpenAI image model. You get today's text rendering and prompt following instead of a placeholder that waits for a launch.

02

Flare for speed, Sunburst for detail

Flare is the fast default for drafts and iteration. Switch to Sunburst when a frame has to be final: sharper detail and tighter control over the prompt.

03

Reference editing that keeps the subject

Upload your own photos and describe the change. Faces, products and layouts stay recognisable while the scene, lighting or background is replaced.

04

Same account the day GPT Image 3 drops

No second sign-up, no migration, no lost history. When GPT Image 3 goes live it appears in this workspace and your credits stay where they are.

How it works

From an idea to a download in three steps

The first image is free to try. You only pay for generations that succeed.

  1. 01

    Write the prompt

    Name the subject, the light, the style and any words that must appear inside the picture. Short, concrete prompts beat keyword lists.

  2. 02

    Add references if it has to be real

    Drop in your own photos when the result has to keep a specific product, person or layout. Describe what changes and what must stay.

  3. 03

    Generate, compare, download

    Choose a ratio and resolution up to 4K, run the prompt, then download the frame you want. Every result stays in your history.

What people use it for

Four jobs that used to need a studio

The same workspace covers campaigns, catalogues, decks and prototypes.

Posters and in-image text

Headlines, dates and small print that render as type instead of squiggles, ready to drop into a layout.

E-commerce product images

Put a real product on a new surface or in a new room without re-shooting the whole set.

Social and ad creatives

One idea, several ratios: square, story and landscape frames from the same prompt and reference.

UI mockups and infographics

Diagrams, interface concepts and explainer visuals with labels that can actually be read.

Questions people ask before signing up

Short answers about the model, the credits and what happens on launch day.

When will GPT Image 3 be released?

OpenAI has not announced a date. The moment GPT Image 3 goes live, it becomes available in this same workspace — no new sign-up and no new balance.

Can I use GPT Image 3 for free?

New accounts get 8 free credits with no card, which is two 1K images on the current model. Credits you already hold stay in your account when GPT Image 3 arrives.

Which model creates my images right now?

GPT Image 2.5, OpenAI's newest image model as of September 2026. The generator labels it by name, and GPT Image 3 has not been released yet.

Do I need to pay again when GPT Image 3 launches?

No. The workspace, your account and your credit balance stay the same. You pay for the images you generate, not for a model switch.

Can I use the images commercially?

Commercial use follows OpenAI's terms for the underlying model and this site's terms. Read both before you publish, especially for products, ads and print.

Does it render readable text inside images?

Yes. Text rendering is a strength of the current OpenAI image model: headlines, labels and short paragraphs come out legible, and reference editing keeps existing text in place.

Can I edit my own photos?

Yes. Upload one or more reference photos, describe what should change and what must stay, and the editor keeps faces, products and layouts intact.

Is there a GPT Image 3 API?

There is an API for the model this site runs today, gpt-image-2.5. Create a key on the API Keys page and call the image endpoint with your prompt, ratio and resolution.

Pricing

Start free, pay for what you generate

Eight credits on signup cover your first two 1K images. After that, buy a credit pack or subscribe — one balance covers every model in the workspace.

See plans and credit packs
  • 8 free credits on signup, no card
  • Credits are refunded when a generation fails
  • One balance for every model in the workspace

Be among the first to create with GPT Image 3

The day GPT Image 3 drops, it is live here on the same account, the same credits and the same history. Start now and your first two images are on us.

Start creating freeRead the FAQ