How to generate an image in ChatGPT on Mac: steps for the new 2026 app

Start with the right ChatGPT app

Open ChatGPT from your Applications folder or the Dock. OpenAI released a new ChatGPT desktop app for macOS and Windows on July 9, 2026. It combines Chat, Work and Codex in one app. The older Mac app was renamed ChatGPT Classic.

Classic still gets model updates, bug fixes and security patches. It does not get the Work and Codex agent features. Switching is optional, according to Plain AI Daily.

For image work, use the new app. OpenAI's docs describe image requests made from the app composer, and this guide could not confirm the same steps for Classic. If you cannot find Work or Codex in your app, you are probably in Classic.

Generate an image from the composer

In the desktop app, image requests start in the composer. The composer is the text box where you type messages. You type your request there, just as you would type any normal question.

Start a new chat when the picture has nothing to do with your last conversation. A separate chat is easier to find later. It also keeps old context out of the new request.

Write a prompt the model can use

A good prompt names the subject, the setting, the style and the light. Short prompts leave the model to guess. A guess can give you a picture you did not ask for, so spend a minute on the wording.

Say what to leave out, too. A phrase such as no people or no text tells the model where to stop. Keep any text inside the picture short, and put the exact words in quotation marks so the model knows what to write.

Edit a photo with a reference image

To change a photo or an earlier image, attach it to the composer before you type. Then describe the change. For example, attach a product photo and ask for a plain white background, with the product in the same position.

OpenAI's docs say a reference image lets ChatGPT edit the picture or use it as visual guidance. For the attach steps, read how to attach a file in ChatGPT on Mac.

Reference images also help when you want the same character in several pictures. Images 2.6 lets you save a photo as a permanent reference, name it, and reuse it in other chats with an @mention. That feature is rolling out in phases, so you may not see it yet.

Open the result and compare versions

Click a picture to open it in an expanded viewer. The desktop app offers two views. Focused view shows one image large. Canvas view adds Comment and Multi-select, which help when you want to mark a detail or compare several pictures.

Images 2.0 can create up to eight images from one prompt, at up to 2K resolution, according to MacRumors. Expect a set of options when you ask for variety. Use Canvas view to sort through them before you choose one.

Refine the picture without starting over

Reply in the same chat when you want a change. Name what to change and what to keep. Asking for the same bicycle with a blue frame, and the harbour unchanged, gives the model two clear instructions.

• Keep one change per message when you can.
• Describe the change in plain words, not in a long list of rules.
• Use Comment in Canvas view to point at a detail, if you can see that option.
• Start a new chat when the old thread gets long and the results drift.
• Attach the best version as a reference before you ask for a new style.

Refining works best when you change one thing at a time. If several things change at once, you cannot tell which change helped.

Save the image on your Mac

Saving is where people get stuck. A 2024 forum thread describes generated images that would not save. Right-click showed no Save as option. The workaround in that thread is to double-click the image, open it in Preview, and export it.

That thread is a single report with no reply from OpenAI staff and no official fix. Your version may behave differently, so check the save options after each app update.

Exporting from Preview may give you a .webp file. A user in the thread said that format would not open in Photoshop. If the app you need accepts PNG or JPEG only, check Preview's export options before you save.

The Images section on web and mobile

On the web and in the mobile app, ChatGPT has an Images section. On the web, click the sidebar icon, select Images under Search chats, then upload an image or enter a description. Storyboard18 quoted these steps from OpenAI's Images in ChatGPT help article, published December 19, 2025.

That help text covers web and mobile only. It does not name a Mac path. This guide could not confirm an Images entry in the Mac sidebar, so do not expect one until you see it.

If you prefer a browser, open ChatGPT there and use the web steps above.

Know your plan's image limits

Image limits depend on your plan and workspace settings. The ChatGPT pricing page does not list numeric caps. It uses plain wording instead, shown in the table below.

Third-party sites publish daily counts for Free and Plus. Those counts conflict with each other, so this guide does not use them. Check the pricing page or your account for the current wording.

Images 2.6 was released on November 4, 2026. Reports say it is available on every plan, including Free, on web, desktop and mobile.

Mistakes people make with image requests

Many image problems start with the request. A vague prompt gives a vague picture, and a long pile of rules can confuse the result. Watch for these habits.

• Asking for long paragraphs of text inside the picture.
• Sending a second request before the first picture appears.
• Forgetting that a new chat does not show the older picture.
• Giving up on saving after right-click, without trying Preview.
• Attaching a reference image without saying what to keep.
• Assuming Classic and the new app have the same features.

Related steps to read next

If you would rather speak your prompt, read how to dictate a prompt in ChatGPT on Mac. If you want the same style in every chat, read how to set custom instructions in ChatGPT on Mac.

What the usage numbers tell you

According to Dataconomy, citing OpenAI, more than 4 billion images are generated each week across ChatGPT and the API. That volume means the image tools get frequent updates. Check the app version and the plan wording before you follow an older guide.

What this guide could not verify