IMAGEOpenAI · from 24 credits / image
The image model that lives inside GPT-4o
GPT-4o Image Generation and Editing
Readable text. Many objects per prompt. Edit up to 5 images.
OpenAI built image generation straight into GPT-4o, so it spells words correctly, keeps track of a long list of objects and learns from the photos you attach. On PicX Studio you run it per image, with no ChatGPT plan.
- 1K
Images that carry information
Signs, menus and labels you can read
OpenAI's launch focused on the useful kind of image: signs, menus, labels, diagrams. Put the exact wording in quotes and say where it goes, and GPT-4o Image places it on the picture. Keep it to a headline and a few short lines: very small, dense text is one of the limits OpenAI lists.
Long lists of objects, each with its own detail
OpenAI puts GPT-4o at up to 10 to 20 distinct objects in one prompt, against the roughly 5 to 8 where it says other systems start to struggle. Give each item its colour, position or label and it stays tied to the right thing.
Learns from the images you attach
Attach up to five photos and GPT-4o Image reads them as context: keep a face, borrow a style, or build a new design from several references. With images attached, the prompt can be as short as the change you want.
GPT-4o Image, from its maker
Official release video by OpenAI
GPT-4o Image
Readable text. Many objects per prompt. Edit up to 5 images.
Visuals that explain something
Infographics Labelled maps, guides and step charts.
Covers and posters A title, a name, a date.
Busy illustrated scenes Every prop placed where you said.
Photo edits Same person, new setting.
Reference sheets Grids of labelled items.
Why pick GPT-4o Image
- TextRendered in the imageHeadlines, labels and signs.
- 10–20Objects per promptOpenAI's own figure.
- 5Reference imagesEdit, restyle or combine.
- 3 shapes1:1, 3:2, 2:3Square, landscape, portrait.
- FlatOne price per imageEvery shape costs the same.
- No planNo ChatGPT PlusCredits, web or API.
What GPT-4o Image costs
Billed per image, at the resolution you pick. The same prices the generator shows before you run it.
- 1K24credits / imageOn Starter≈ 150 images / mo
Every model on every planCredits never expireSee plans
Pay for GPT-4o Image with crypto
Buy credits with Bitcoin, USDT, USDC, ETH, Solana and more — no card, nothing auto-renews, packs from $5, and 15% extra credits on crypto payments.
GPT-4o Image vs GPT Image 2.6 Flare
Both are OpenAI models in the same PicX generator. GPT-4o Image keeps it simple; GPT Image 2.6 Flare adds sizes, shapes and references.
Live credit prices are in the pricing table.
How to use GPT-4o Image in 3 steps
- 1
Attach references, if any
Add up to five images to edit, restyle or combine. Skip this to create from text alone.
- 2
Write it like a brief
List the objects and give each one its detail. Put any words in quotes and say where they sit, for example 'top left, a third of the way down'.
- 3
Choose 1:1, 3:2 or 2:3, then iterate
Square for covers, 3:2 for scenes, 2:3 for posters. To refine, attach the result as a reference and describe only the next change.
GPT-4o Image prompts to copy
Real prompts and what they made. Copy one, or open it in the generator with the model already picked.
Illustrated harbour guide
An illustrated guide to a fishing harbour at dawn, soft pastel watercolour. Number and label eight things: 1 lighthouse, 2 fishing boat, 3 crab pots, 4 harbour wall, 5 gulls, 6 fish market, 7 rope bollard, 8 tide board. A small legend box in the bottom right lists the numbers and names. Title at the top: 'Morning at the Harbour'
Bike parts reference grid
A square 3 by 3 grid on a white background showing nine parts of a green city bicycle, one per cell, left to right, top to bottom: brass bell, wicker basket, leather saddle, front light, chain, pedal, kickstand, mudguard, tyre pump. Each item has its name printed underneath in small black capitals
Desk with every detail placed
Anime-style illustration of a study desk by a rainy window at night: a green desk lamp on the left, a red mug that says 'FOCUS', an open notebook with 'chapter 3' written at the top, white headphones, three small succulents, a sleeping orange cat on a stack of books, and a polaroid of the sea taped to the window
Vinyl record cover
Vinyl album cover titled 'Midnight Ferry' by 'The Harbour Lights', a ferry crossing a dark bay under the moon, bold retro lettering
Templates to start from
Browse all templates
GPT-4o Image questions
What is 4o image generation?
It is the image generator OpenAI built natively into GPT-4o and announced on 25 March 2025, when it became ChatGPT's default image model. Because it shares GPT-4o's knowledge, OpenAI pitched it for useful images such as diagrams, infographics, menus and labelled designs, not only pretty scenes.
Is GPT-4o image generation free?
In ChatGPT, OpenAI rolled it out to Free users as well as paid plans. On PicX Studio it is a credit model: each GPT-4o Image render costs the same credits whatever the shape, and the one free image new accounts receive is made with Nano Banana 2 instead. You need no ChatGPT subscription to use it here.
How is 4o image generation different from DALL·E?
OpenAI describes 4o image generation as significantly more capable than its earlier DALL·E 3 models. It renders text more reliably, follows long, detailed prompts and can take your own images as input to transform them. DALL·E is not offered on PicX.
How do I use GPT-4o image generation for editing?
Attach up to five images, then describe the change, for example 'same face, plain grey studio background'. With images attached, a one-line instruction is often enough. PicX runs each edit as a single request, not a chat, so to refine further, attach the result and describe the next step.
How do I write good GPT-4o image prompts?
Write it like a brief. List every object with its colour or position, put exact words in quotes and say where they go, and name the style. OpenAI notes the model can crop tall images such as posters too tightly near the bottom, so leave space under the last line of text on 2:3 images.
Can it make Ghibli-style or anime images?
OpenAI trained it on a wide range of styles, so anime, watercolour, vector and photo looks all work from a prompt or a style reference. Describe the look in your own words, for example soft hand-painted backgrounds and warm light, rather than relying on a studio's name.
How do I use the 4o Image API?
Use the PicX Studio API with your PicX key: send a prompt, optional reference image URLs (up to five) and a size of 1:1, 3:2 or 2:3. Credits come from your PicX balance, at the same flat per-image price as the web app. See the developer docs for the request format.
What can GPT-4o Image not do on PicX?
There is one output size (about 1 MP) and three aspect ratios. The chat-style back-and-forth refinement and the transparent-background option OpenAI showed in ChatGPT are not available on PicX yet. OpenAI also lists dense small text, precise charts and non-English text as known weak spots.
Create with GPT-4o Image
Readable text, busy scenes and photo edits, per image, no ChatGPT plan.