10 Must-Try Tricks to Edit DALL-E Images Directly in ChatGPT

With inpainting now a part of ChatGPT with DALL-E 3, users will have more tools in their arsenal when it comes to editing their newly generated AI images. The “inpainting” allows users to edit the images directly and instantly.

We already have several similar AI models such as Midjourney and Stability, but DALL-E 3 Inpainting provides us with a special in-demand feature of editing generated images at one glance.

But how are AI enthusiasts, editors, and developers making the most out of this feature? In this article, we have curated some of the examples shared by tech enthusiasts online and looked into them in-depth.

10 Examples of ChatGPT’s Editing Images Feature

Below are some of the use cases of DALL-E 3’s new Inpainting feature that we have gathered from several users worldwide:

1) Adding a Speech Bubble

X user Umesh, an AI enthusiast, discovered an interesting use case in which he added a comic-like speech bubble to a black-and-white cartoon image generated by DALL-E 3.

To add the speech bubble, he selected the area of the image to be edited and then continued to give the prompt ‘Add a speech bubble that says “It is so fun”’.

This is quite a beneficial use case as now image editors can edit their generated images to make them look comic-like with a speech bubble. For a long time, writing and drawing comic images has been a hectic and troublesome task for several artists and editors worldwide, but with Inpainting we can say goodbye to this hassle.

2) A Series of Edited Images

Denny Klisch, a Product Manager. generated an image with OpenAI and used Inpainting on it to add a series of edits. His first generated image shows a simple tree. He then edited the tree image by adding apples to it first. He continues to edit further as he adds an aeroplane and birds in the third and fourth picture respectively.

This is a great feature as being able to edit the same image repeatedly, allows editors and developers to make impromptu adjustments to their image and scale it according to their preferences.

3) Reducing Brightness or Saturation

Kristi Hines, a Content Creator, tried experimenting with OpenAI’s Inpainting feature. She first generated an image of a gothic library interior using GPT-4. However, the image was a little more bright than she wanted.

So, she used the Inpainting feature and asked GPT-4 to make the image darker. Surprisingly, the chatbot did an excellent job of darkening the image! The overall brightness was highly reduced and the gothic library interior looked more mysterious and tranquil.

Adjusting image brightness and darkness is an in-demand feature that every image editor tool should have. And it’s highly beneficial for the editor community to have this feature from Inpainting.

4) Adding More Objects

We already know that GPT-4’s editing capabilities include adding objects to the AI-generated image. An AI enthusiast on X, named Teodora wanted to put this feature to the test. She first asked GPT-4 to generate an image of a baby panda.

She then began using the inpainting feature and asked GPT-4 to add a “bow” to the head of the panda. She then continued to select a specific region in the panda image and added an ice cream cone object to it.

These are just examples of various objects that can be added using OpenAI’s Inpainting feature. Editors and developers can think of several objects that can be added to their image and customize it overall.

5) Changing Facial Reactions

Ann, a programmer, also put the new tool to the test! She generated an image of an unhappy girl in an emotional black-and-white comic book-style portrait. GPT-4 successfully generated the image encapsulating the emotions in the girl’s facial expressions.

She then used the Inpainting feature and asked GPT-4 to change the girl’s reaction from sad to happy and content. She gave the prompt as “change expression to happy”. The resultant edit was amazing from the chatbot, as it generated a new image of the girl where she could be seen smiling with wide-open eyes.

It is a high advancement from OpenAI to develop an image editor tool that can alter facial reactions, without changing the subject’s details and ratio. This is optimal and efficient image editing at its peak.

6) Altering Image Texts

We can also use the inpainting feature to alter texts in images. Pierrick Chevallier, an expert in AI & Design, shared an experiment video with OpenAI’s Inpainting feature on X. He asked GPT-4 to generate a rum label image first.

He then decided to change the text in the rum label image logo. First, he selected the text portion “CAARBBEAN COVE” and then gave the prompt ‘text Caribbean Cove’. In the resultant image, the text was changed according to the user’s prompt. In the same way, he changed the “AGED RUM RUA” to “AGED RUM”.

Altering text in images has been a highly difficult task until now as basic editing tools can’t identify texts from images. But OpenAI’s Inpainting feature is here to just change the landscape forever.

7) Changing Subject Color

A future news media company named Testing Catalog News shared an image of an experiment they conducted with OpenAI’s Inpainting feature.

They first generated an image of two futuristic robots with the help of the chatbots. The robots in the images were mainly a mixture of both grey and blue colours. They decided to change the robot’s colour to red and gave the prompt “Make it red”.

GPT-4 responded excellently, providing an image of the previously generated robot with a vibrant red colour. The chatbot even asked if the user liked the image or not.

This shows that now you can even change the colour of the subject matter in the generated image using the Inpainting feature. Often other tools mess up the texture and quality of the image while changing the color of the image subject, but ChatGPT seems quite efficient in this task.

8) Editing Image Backgrounds

The same news media company pointed out another interesting feature of OpenAI’s Inpainting tool. They used it to alter the background of a generated image. They first generated an image of a robot in a serene landscape illuminated with soft and warm lighting. GPT-4 responded excellently with two amazing images.

Then they used the editing functionality. They gave the prompt “Draw an urban background”. ChatGPT responded amazingly by removing the objects in the image background and adding modern-style skyscrapers. The resultant edited image looked amazing and it looked as if the same robot was in a different background.

This use-case brings to light the fact that now editors can change image backgrounds at ease without having any hassle or troubles at all. Users had a tough time cropping out the image subject from the background and editing them separately, but Inpainting is here to help you out with this.

9) Changing Image Landscape

Asutosh Shrivastava, an AI enthusiast on X used OpenAI’s Inpainting feature to change the landscape of a generated image. He first generated an image of a green field landscape in a wide aspect ratio. ChatGPT gave an excellent landscape image capturing the prompt details.

He continued to edit the image landscape by selecting a portion of the green landscape and then he gave the prompt “River flowing”. ChatGPT made an excellent edit to the image by adding a river to the landscape. The interesting part is that the river was flowing exactly in the same way as the user had made the selection region in the image landscape.

This is a groundbreaking feature as editing image landscapes before required highly efficient designing tools with multiple reference axes and orbits. But with OpenAI’s Inpainting, all that you need is a simple prompt.

10) Removing Subjects Completely

The same user continued to use the Inpainting feature to remove a subject completely from a generated image. He first generated the portrait of a 30-year-old man using the chatbot. GPT-4 did well in generating the image but by mistake, it produced the same portrait again in the image background.

The user first selected the unnecessary portrait portion from the image background. Then he gave the prompt “Remove this man”. GPT-4 did well in removing the portrait but still, a few traces were left.

He selected the traces and then gave another prompt “clear this section”. This time the subject was completely gone and the blank background looked as if the portrait never existed in the first place.

View the previous video again from the timestamp 0:23, to witness the removal feature in action.

So, we can make use of Inpainting to completely remove image subjects and objects at ease, without worrying about leftover image traces and parts.

Conclusion

ChatGPT’s new in-built editing features for images are already a big thing among developers, editors, and AI enthusiasts. The generative AI community needed a skilled image editor and now we have one, thanks to OpenAI. Its diverse use cases and functionalities give us a strong idea of how far this tool can go in terms of image editing and enhancement.