You can create an original image in ChatGPT by describing what you want in a conversation. On supported interfaces, you may also see an Images option that provides a dedicated starting point. The feature is available on the web, iOS and Android.

The workflow is straightforward: start the request, write a specific prompt, choose the image shape, review the result and then save or share it. OpenAI documents ChatGPT Images as available across all tiers, although image creation has separate usage limits that can vary. See OpenAI’s Images in ChatGPT guide and the ChatGPT Free Tier FAQ for the current availability details.

Start an image request in ChatGPT

You can generate images in ChatGPT on the web or through the iOS and Android apps. Because menus can differ by platform, account and rollout, use whichever of these documented entry points appears for you:

  1. Open ChatGPT and start a new conversation or open a suitable existing conversation.
  2. Enter a description of the original image you want ChatGPT to create.
  3. Alternatively, select More and then Images if that option appears in your interface.
  4. Submit the request and wait for the image to be generated.

A complex request may take several minutes. You can leave most creative decisions to ChatGPT, but a more specific description usually gives it a clearer target.

Write a clear image prompt

A useful image prompt can include the purpose, subject, action, setting, visual style and any essential constraints. Not every prompt needs every element; include the details that materially affect the result.

A reusable structure is:

Create a [type or purpose of image] showing [subject and action] in [setting]. Use [visual style, mood or lighting]. Include [essential details or constraints].

For example:

Create a landscape illustration for a travel article showing a red passenger train crossing a stone bridge in a misty mountain valley at sunrise. Use a warm editorial style with soft light. Keep the train as the clear focal point and do not include text.

Start with one to three clear sentences rather than a long list of competing instructions. If two requirements conflict, ChatGPT may make an unwanted compromise. OpenAI’s image-generation guidance likewise recommends giving the model a well-defined subject and relevant visual details.

Choose an aspect ratio or orientation

Choose the image shape according to where you intend to use it:

  • Square works well when you need a balanced image with no strong horizontal or vertical emphasis.
  • Portrait is useful for tall compositions, such as a standing subject or mobile-oriented design.
  • Landscape provides more horizontal room for scenery, groups or wide compositions.

If ChatGPT displays an aspect-ratio picker, select the option that fits your intended layout. The picker may not appear in the same place—or at all—on every platform or account.

When there is no picker, add the desired shape directly to the prompt. For example, begin with “Create a portrait image” or “Create a wide landscape image.” You can also request a specific aspect ratio, but prompt wording should not be treated as a guarantee of exact dimensions.

Refine the generated image

Open or inspect the generated image before accepting it. Check the main subject, composition, setting, visual style and any important small details. If the image contains text or an intricate layout, inspect those elements especially carefully because they may require further refinement.

For a more controlled revision, request one or a few precise changes at a time and identify what should stay unchanged. Avoid a vague response such as “make it better,” which does not tell ChatGPT which part needs attention.

For example, the first image might place the train too far into the background. A focused follow-up prompt would be:

Make the red train larger and move it closer to the center of the composition. Keep the stone bridge, misty mountain valley, sunrise lighting and warm editorial style unchanged.

Review the revised result and repeat with another focused instruction if necessary. Do not assume that generated text, tiny objects or complex arrangements will be flawless. If your next goal is to make targeted changes after generation, follow the separate guide on how to edit an image in ChatGPT.

Save, copy or share the finished image

Once you are satisfied with the result, open the generated image to see the actions available on your device. Depending on the platform and current interface, these can include Copy, Save or Share.

  1. Open the finished image at its larger view.
  2. Select Copy, Save or Share, according to what you want to do with it.
  3. Check the copied, saved or shared result to make sure it is the intended version.

Before distributing the image, look for inaccurate text, unintended visual details or anything that could misleadingly resemble a real person, event or document. The exact action labels and their placement can differ between the web, iOS and Android interfaces.