Comparison of GPT and Midjoni Chat • DigiKala Meg

With the emergence of artificial intelligence and discussing the best artificial intelligence, Now the two artificial intelligence giants are in front of each other. Midjourney and ChatGpt have recently released new versions of their artificial intelligence -based image production capabilities. Historically, these two artificial intelligence have been one of the best in the image production space. But which artificial intelligence offers better illustrations? Midjourney V7 or the pictures that ChatGpt 4o makes? The winner of the comparison of GPT chat and Midjonny is specified.

In this article, ChatGpt and Midjourney are used against each other using seven different peripheits; The purpose of this confrontation is to see what is the best producer of artificial intelligence image. These experiments examine everything from the model’s ability to understand the concept to recreation of complex forms and creative thinking to create images.

ChatGpt and Midjourney v7 comparison rules

While both models of these artificial intelligence can produce an image, it is difficult to hold this competition fairly; Because the volume of settings that Midjourney allows the user to change. However, considering this, we will start the early stages of competition. In this analogy for Midjourney, we used version 2. This is the latest version but it is still in the trial stage. We also tried the test of this experiment with the personalization capability (settings that add your favorite art styles to the images). Midjourney produces four copies of each image for the user compared to an image ChatGpt. In all cases, comparison of GPT and Midojani Chat, we selected the best image from the four proposed Midjourney works and downloaded the high quality version.

1. Photorulism (Realisticism)

Left: ChatGpt / Right: Midjourney

Prampt: A photographic image of a puffy bird that creates over the rock with the water at the bottom, while the bird flies. There is a mountain range in the background. The day is sunny, and at the bottom of the puffin bird, two people look at him with binoculars.

ChatGpt result:

This image responds to almost all the desires in our Permapt. Although the image may be slightly over -saturated (high Sacrifies), it is still a photorealist. A puffin bird flies over a rock, there is water at the bottom, and a mountain range is seen in the background. In addition to the tips, the image produced also includes the two men who look at him with a binocular camera. Of course, they do not look directly at the puffy, without Grefen’s view of this image is quite accurate.

Midjourney result:

There are many things in this picture. We cannot deny that everything he had in Prambt is considered. The mountains in the background, a puffy bird, two with binoculars, and even water and mountains in the background. However, let’s look at the main point of the image. The puffin bird in this image is giant and can fight Godzilla if necessary. The image is also not really a photorealist and looks more like a large oil color painting. Even with the problems of the bird size, we still think that ChatGPT has understood the symptoms more accurately. Both made water models at the bottom of the cliff, but Chat GPT understood the general concept of Pram.

Win: ChatGpt In this case, it wins in almost every way. Midjarni has misunderstood the concept of the puffin bird in the sky, and on the other hand, ChatGpt has fully fulfilled our desire, so the winner of the GPT chat and Midjarani is undoubtedly chatgpt.

2. Complex crops

Left: ChatGpt / Right: Midjourney

Praps: Consider a large market with a fruit booth, a clothing booth and a ceramic booth. There is a forest in the background of a river. A man gives a woman to a woman in front of one of the booths and two children in the market. There is also a warm air balloon in the sky.

ChatGpt result:

There was a lot of information in this crop, and some of them were easy to ignore for the artificial intelligence model. However, there are all the key details in the GPT chat response. The hot air balloon, the two children in the market, and the man who gives money to a woman. It is also clear that the air is warm and you can see a market that sells fruit, ceramics and clothing. Despite all the required details, ChatGPT produced a high quality and very accurate image.

Midjourney result:

While Midjourney produced the image of Prapete, its details were very small and were not accurate. For example, when zooming, the faces are not perfect, the hands of the two are merged and the background is blurred.

Winning: The board of this stage of comparing the GPT chat and the ChatGPT is ChatGpt. While both images look correct at first glance and include most of the requested features, Midjourney has lost much detail.

1. Matching with real images

Left: ChatGpt / Right: Midjourney

Praps: Turn this image into a Renaissance portrait.

ChatGpt result:

With this adaptation, ChatGPT basically turned the image into a Mona Lisa -style image. Still, the work of this artificial intelligence cannot be done. Exactly the photo we presented to him in the Renaissance style. It has also well maintained features such as headphones, backgrounds, and clothes wearing, while at the same time adhering to the main topic.

Midjourney result:

This was the best attempt by Midjourney to us. We see what the model was trying to do. But it couldn’t reach it. During the test, we even tried to change the Permapce a little to find that we wanted it in the style of painting, but that made things worse. We optimally assume that the brown margin is supposed to match the subject, it is really difficult to identify it.

Win: ChatGpt He had proven that his new model was very successful in making creative changes to personal images, and this experiment is another positive. He did exactly what we wanted. Midjourney seems to be halfway through. So the winner of the comparison of ChatGT and Midjarni is clear.

1. Movie posters

Left: ChatGpt / Right: Midjourney

Prambt: Create an exciting poster for this movie: a cyberpunk movie that takes place in year 6. The story is told in a large, vibrant city. The film is about a detective sent to the past to prevent a future war.

ChatGpt result:

This is not the most exciting poster, but ChatGPT definitely did our will. Our detective is in the center of the scene, with a crowded (and relatively futuristic) city in the background. He also picked up Pramat as a completely literal text and added a slogan to the requested data. In general, it is impressive. Detective with a shadow of neon light on his back with detail, there is a bird machine in the sky, although a little crude is drawn. There are also many futuristic skyscrapers in the background.

Midjourney result:

In fact, it tries to make up for it with its style. Certainly skyscrapers look better here, and there are more things to see in this picture. Unfortunately, Midjourney stays behind Jet GPT in detail. The images are mixed on the ground, the car and the motorcycle are in trouble, and there are many strange details in the background. While this poster is more interesting, there are many problems.

Win: ChatGpt He made all our desires and made a poster that we could publish and no one could notice it (except the extremely boring film). On the other hand, Midjourney did a lot wrong here. However, compared to the GPT and Midjoni chat, we respect the path that Midjoni had taken.

1. Text production

Left: ChatGpt / Right: Midjourney

Post: Create an image of a poster written on the poster: “The” Artificial Intelligence Image Manufacturer “performs tonight at 9pm! The cover will perform all your favorite songs »Stylize the image as if this is a poster for a music band that performs in a popular place.

ChatGpt result:

There is a common theme in ChatGpt in the image production capability: Details often have superiority over style. This poster made all our desires, and more importantly, he wrote the whole text completely correct. ChatGpt had trouble with just a few months ago, so it is exciting to see how progress it has made. While this poster is a boring poster, it has fulfilled our will and has made a difficult challenge for artificial intelligence models.

Midjourney result:

We appreciate that Midjourney made this poster more like a poster mounted in a venue and background. We also like the energy that was sought in the middle of the poster by placing the group’s image in the middle of the poster. However, except for the words “Band”, not even a word of the text cannot be read. Compared to the ability of the ChatGpt to display the entire text, this poster is a little disappointing.

Win: ChatGpt It may not have been very interesting here, but it completely finished. As Midjourney has shown, dealing with text in images is not always easy for artificial intelligence models. So the winner of the GPT and Midjoni chat will be ChatGpt.

1. Hands

Left: ChatGpt / Right: Midjourney

Prampt: Create an image of a person’s hands while holding an orange in the left hand and holding a glass of water in the other hand.

ChatGpt result:

How progressive has artificial intelligence. When producing artificial intelligence for the first time, one of the easiest ways to identify it was hands. They had extremely loud fingers or had fingers that had fallen out of the wrong place. At present, while the hands here do not look completely human, its accuracy is really impressive. Both hands have the right number of fingers (a good start), the water in the glass correctly changes the image of the hand behind it and you can see the veins and nails.

Midjourney result:

Midjourney did a great job here. What we think is particularly impressive in this picture is its details. Hair on the arms, veins, bruises on the fingers and stretching lines on the hands. While the ChatGPT image is immediately recognizable as artificial intelligence, it can be recognized as a human hand. The only noteworthy problem is the finger behind the glass and does not seem to be completely natural. It is also a very strange way to keep the oranges, but everyone has their taste.

Win: Midjourney By overcoming one of the most well -known weaknesses of artificial intelligence, it won. This shows how much he has progressed. This does not mean that ChatGpt did badly, it was not just good enough compared to Midonery.

1. Food

Left: ChatGpt / Right: Midjourney

Prampt: Create a picture of a sea bowl that can be used for a food Instagram page.

ChatGpt result:

This kind of image is one of the same food photos we see in a cookbook and we don’t even doubt it for a moment. Although there is no external existence, you may want to eat it. We really can’t criticize the work of the artificial intelligence model here; He made all the demands of our requested, even if the random vegetable at the bottom is completely inappropriate.

Midjourney result:

Just like ChatGpt, Midjourney did a great job here. This image looks like a real pasta bowl served in a good restaurant. Even a few tomatoes and random garlic are scattered around, which we assume is to decorate.

Win: ChatGpt Thanks to a little better visual quality, it wins here, but like the case of the hands, the competition was closely competitive compared to the GPT chat and the Midjoni.

Final conclusion: ChatGpt win.

Unfortunately for Midjourney, competition was not even close. Of course, it is important to note that Midjourney has just released this release and is still in the test phase. However, the latest version of GPT image production is only one or two older weeks. While these models were once equal, ChatGpt was often superior to where Midjourney was not successful. We hope that Midjourney will see progress in the experimental phases of version 2, as it can become a great artificial image manufacturer.

Source: Tomsguide