How to generate AI images using ChatGPT

Since March 2025, ChatGPT has been able to generate images. After a period where it was briefly unavailable to free users, you no longer even pay for one of OpenAI’s subscriptions to use this feature. And while creating images in ChatGPT is easy, there are some nuances worth explaining. For example, did you know that you can ask ChatGPT to edit photos you have taken? It’s more powerful than you think. Here’s everything you need to know about generating AI images with ChatGPT.

How to Create Images with ChatGPT Using Text Prompts

You can start generating images in ChatGPT by simply typing in the prompt bar what you want to see. There is no need to think too much; as long as you have a version of “generate image” followed by a description of your idea, ChatGPT will do the rest.

Depending on the complexity of the prompt and whether you pay for ChatGPT, the chatbot may take a minute or two to complete your image request. Sometimes the process may take longer if OpenAI’s servers experience higher traffic than usual.

Late last year, OpenAI updated the model that powers image generation to make it faster, as well as improve text rendering and instruction tracking. At the same time, it added a dedicated “Images” section to the ChatGPT sidebar. Here you can see all the images you’ve created, along with sample prompts and suggested styles to try, making it a great place to start if you’ve never used an image generator before.

How to create images with ChatGPT using existing photos

In addition to generating images from text prompts, ChatGPT can edit existing photos or images you upload. This is my favorite way to create images with ChatGPT; I don’t need to describe the composition, I can use an existing one to guide the chatbot. To use an existing image as a starting point for a new generation, follow these steps:

1. Press the “+“, located to the left of the prompt bar.

2. Select Add photos and files.

3. Select the image you want ChatGPT to edit. If you’re uploading an image from your phone, you’ll need to first grant ChatGPT access to your camera roll.

4. Write a prompt describing your desired changes.

If you are generating from the Images section, tap “Add Photos” instead.

Keep in mind that any photos you upload to OpenAI’s servers may be used by the company to train future models. You can refuse to have your data used for training by following these steps:

1. Open the sidebar menu.

   1. On mobile, tap the two lines at the top left of the interface; on the desktop, click on the OpenAI logo instead.

2. Tap your name to access account settings.

4. Disable Improve the model for everyone.

How to edit images generated by ChatGPT

If you are not satisfied with the result of ChatGPT, you have two options. You can either invite it to create a completely new image or modify parts of the image it just generated. As always, the process for both is simply typing whatever you want into the prompt bar. On mobile, OpenAI offers users different ways to accomplish the same task.

To generate a completely new image:

1. Tap the three-dot icon below the image created by ChatGPT.

To modify part of an existing image generation:

1. Tap the image created by ChatGPT.

3. Use your finger to hide the section of the image you want ChatGPT to edit. The left slider adjusts the size of the masking brush. On desktop, masking is also available if you click on an image then click on the paintbrush icon at the top right.

4. Describe what you want ChatGPT to add, remove, or replace via the prompt bar.

ChatGPT can also mix one of your photos with an image it generated. To do this:

1. Tap an image created by ChatGPT.

3. Upload the photo you want

Like all AI systems, ChatGPT is non-deterministic, meaning that even if you ask it the same way multiple times, it won’t generate the exact same response each time.

Tips for Creating Better Images with ChatGPT

The best advice I can offer is to be specific when inviting ChatGPT. The more detail you can provide when describing what you expect, the better the results. And remember: ChatGPT can hallucinate – as you may have noticed in one of the sample images I included above. Like the tortoiseshell cat, not only is the tortoise not sitting on the windowsill as shown, but it is sitting on a table that doesn’t make much sense. So above all, be patient. Coming up with an AI model is not an exact science, and it may take a few tries before creating the desired result.

FAQs

How to access ChatGPT?

ChatGPT is available on web, desktop and mobile. To access it on your computer, open your favorite browser and navigate to chatgpt.com. OpenAI also offers dedicated Mac and Windows apps that you can download from the company’s website. On iOS and Android, you will need to download the ChatGPT app from the App Store or Google Play before you can start using the chatbot.

Since ChatGPT runs on OpenAI’s servers, as long as you can access the chatbot, you’ll be able to use it to create images regardless of the age of your phone or computer.

Can ChatGPT generate images for free?

Yes, ChatGPT can generate images for free, provided you create an OpenAI account. However, there is a daily cap and GPT-5 will take longer to create a free image. After March 27, 2025, OpenAI briefly limited free users to three image generations per day. The company has since relaxed this restriction, although it does not list a specific limit on its website. In my experience, you will be able to generate about six to seven images every 24 hours.

OpenAI offers three different subscription plans, each with its own set of image generation benefits.

• ChatGPT Go, which costs $8 per month, offers “more image creation.”

• ChatGPT Plus, which costs $20 per month, offers “expanded and faster image creation.”

• ChatGPT Pro, which costs $200 per month, offers “unlimited and faster image creation.”

Note: ChatGPT Go will be included in OpenAI’s upcoming ads pilot, which will see the company display sponsored content alongside ChatGPT’s organic responses. The company does not plan to display advertisements to Plus and Pro users.

Can ChatGPT generate an existing photo?

No. For copyright reasons, ChatGPT cannot reproduce real photos or events. For example, when I asked it to recreate the photo of Zinedine Zidane’s iconic header from the 2006 World Cup, ChatGPT refused.

“I can make an artistic reinterpretation inspired by the emotion or energy of that moment — for example, a stylized painting showing the tension and intensity of the competition, without depicting real individuals,” he told me.