Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

History

Image generation becomes part of the conversation

·

OpenAI added image generation to GPT-4o in ChatGPT.

A designer looks at two versions of an image in a conversation window.
Image generation became part of a conversation with follow-up instructions. Illustration: Denkbaar.

On 25 March 2025, OpenAI introduced image generation with GPT-4o in ChatGPT. Users could request images in a conversation and give follow-up instructions. The announcement covered topics including text in images and following instructions. OpenAI described the capabilities and limitations. A tablet shows blank conversation blocks and two images of a kettle.

An image and its explanation in the same conversation: that is the interesting starting point for a designer here. This gave people a direct reason to compare different versions side by side. It was not evidence that every requested image would be correct. An employee compares an illustration on paper with the same illustration on a phone.

A small or medium-sized business could try this with an original illustration for a public manual. Specify what must be visible and what must not change. Choose your own visual language, with concrete shapes and colours. Then have an employee check whether the image explains the action correctly. Examine text, quantities and proportions separately. Also compare the result on a small screen. Use the trial to establish where generated images help and where a real photograph or a drawn diagram remains necessary.

Sources

  1. GPT-4o image generation in ChatGPT

Topics: image/video