Please Select Your Location
Australia
Österreich
België
Canada
Canada - Français
中国
Česká republika
Denmark
Deutschland
France
HongKong
Iceland
Ireland
Italia
日本
Korea
Latvija
Lietuva
Lëtzebuerg
Malta
المملكة العربية السعودية (Arabic)
Nederland
New Zealand
Norge
Polska
Portugal
Russia
Saudi Arabia
Southeast Asia
España
Suisse
Suomi
Sverige
台灣
Ukraine
United Kingdom
United States
Please Select Your Location
België
Česká republika
Denmark
Iceland
Ireland
Italia
Latvija
Lietuva
Lëtzebuerg
Malta
Nederland
Norge
Polska
Portugal
España
Suisse
Suomi
Sverige
<< Back to Blog

DALL-E 3 Introduces Image Editing in ChatGPT: Does It Achieve Precise Results?

VIVE POST-WAVE Team • April 15, 2024

|

2-minute read

In recent days, OpenAI has released an intriguing new feature—paid ChatGPT users can now "fine-tune" AI-generated images using DALL·E 3, showcasing the potential of AI Image editing. As demonstrated in a demo video, by simply circling the ears of a dog and typing the instruction "add a bow tie," a new element was seamlessly integrated into the original image, resulting in a flawless and natural-looking photo retouching enhancement.

 

 

According to the official introduction, this capability extends beyond adding new elements to an image. Users can also remove objects, alter patterns, and even change the expressions of cats in pictures, all part of the graphic design software's functionality.

Users can also remove objects, alter patterns, and even change the expressions of cats in pictures

This seems quite miraculous, suggesting that if DALL·E 3's AI Image editing results are viable, we might no longer need to use other graphic design software for post-processing after generating images with AI. Could we be looking at a future of direct image output? With the spirit of experimentation, let's put DALL·E 3's new feature to the test and see if it can withstand the rigorous demands of finicky clients in terms of photo retouching.

Hands-On Test: DALL·E 3's Image Editing on ChatGPT

First up, let's try an 8-bit style cat hero in a side-scrolling game scene.

Simply press 'Select' in the top right corner of the image you wish to modify(Simply press 'Select' in the top right corner of the image you wish to modify.)

Looks good, but it would be even better if we could change that orange cat in the top right corner to a black one using AI.

Circle the area you want to adjust, then enter the command(Circle the area you want to adjust, then enter the command.)

Let's see the result👇👇👇

Wait, what happened to the black cat - Why did it just remove the orange cat

 

Wait, what happened to the black cat? Why did it just remove the orange cat?

Alright, let's try changing the pattern of the orange cat in the bottom right corner instead by using the AI photo retouching feature.

What - The orange cat in the bottom right corner disappeared too

What!?! The orange cat in the bottom right corner disappeared too.

It seems altering patterns is a bit challenging for the current AI image editing feature. So, let's just remove the heart and the '20' in the top right corner; these details are a bit distracting.

So, let's just remove the heart and the '20' in the top right corner; these details are a bit distracting

After deletion, the AI has added some strange patterns on its own! At least it made the requested changes. Now, let's try turning the five cat heroes into dogs!

Now, let's try turning the five cat heroes into dogs

The result looks a bit half-baked; each dog looks copy-pasted, and the image is reminiscent of a low-budget production. It appears that (for now) DALL·E 3, even with its new advanced graphic design software capabilities, will still be unable to meet the expectations of demanding clients.