DALL-E users can now edit AI images in ChatGPT

OpenAI has added some new features to its artificial intelligence (AI) image generator DALL-E, specifically its ChatGPT interface. Two new features announced Wednesday will allow users to edit images created by DALL-E in ChatGPT and allow users to see style tips for creative inspiration. These two features appear to be aimed at improving the tool’s accessibility and making it more friendly to casual users. Notably, on Tuesday, the artificial intelligence company also made ChatGPT accessible to everyone, even those without an OpenAI account.

The announcement was passed by the OpenAI official account postal on X (formerly Twitter). Inline editing capabilities in ChatGPT have been extended to the web client and Android and iOS apps. Users with a paid subscription to OpenAI now have access to ChatGPT, and after entering a prompt to create an image, if they are not satisfied with the results, they can open the image, which will now display an edit icon.

The edit icon precedes the download icon. Once clicked, users can select a circle to represent the area they want to edit. After selecting a region, the user will have to provide another prompt explaining what they want to change. After input changes, the AI ​​will regenerate the image. Recently, Microsoft’s Copilot Designer also included a similar inline editing tool.

For those without a graphic design background, using DALL-E on its native platform can be tricky, as the tool is quite complex and requires detailed tips including style weighting to achieve the desired results. This problem was solved when OpenAI integrated the AI ​​image generator with ChatGPT after the launch of DALL-E 3 in September 2023. Now, with editing tools, users will find it easier to fine-tune the images they create.

In addition, OpenAI will display style tips above text fields in ChatGPT to provide users with inspiration. Styles such as woodcut, close-up, low angle, artificial lighting, etc. are shown and hovering the arrow over them will show an example of that style. Users can click on the style they want and then generate images of the same style with simple prompts.