I am now using AI to generate images for almost three years. At that time, I have noticed that it goes from a fun game that can repeat the subject you are asking (just the emphasis about here), for a movement that is close to human artwork.
Nowadays, we have found a bunch of AI image generator to choose. In fact, under AI Sun, every company has just been made about one. But in those early days, there were two who stood outside: Dal-e (now part of the chat) and midzorney.
When Midzorney launched for the first time, it was my favorite AI image generator. This can make an exciting picture and, where AI was at that time, it could do them well. However, when I tried the latest upgrade, V7 of Midzorney, it felt like this behind the competition.
so what’s wrong? Well, after reading a thread of reddit users, how I do not know how to indicate for mid -jorney, I spent some more time on the AI tool. I realized that, while it is not the best option for most people in any way, it can just be the most fun.
What is the problem with midjourney images?
The AI device has gone from a niche interest to a device that every person you know. With this change, we have seen that models grow to understand human signals, fill the spaces you can miss.
This is an area where the chat thrives completely, especially in images. When it comes to the image generation, the chatgpt is not a risk. It is easy to use, requiring very little effort in terms of indicating, and you will generate what you ask in the style.
On the other hand, midzorney, risks. This can be seen alone in its adaptation. Before you can use it, you need to rate more than 200 images, helping the model understand your personal choice for art styles.
There are sliding scales for how strange your images are, as well as change how styled they will be or there will be variety between them. Unlike Chatgpt, midjourney produces multiple images at once and then lets them upscale, make them strange too, and try again.
I asked both models to generate the cyberpank world, which work best for each model. Both created accurate and truly interesting images for this. I tried again again, asked for a slight variation of the image.
Midjourney was styled again. Although it was still accurate for the given signal, it was a completely different form, trying to do something new. On the other hand, the chatgate, the nearest but highly wide images, got stuck in a cycle, which I asked was that the boat was avoided to be pushed far away.
Midjourney is not consistent as a chat in any way, and even with highly wide, thought-out signs, it can still misunderstand things, smear the face or leave huge blurs where there should be objects. However, the midzorney is also between a training period. In theory, its quality should be improved as version 7 develops.
If you want to embrace the bizarreness of AI images, the fantasy world, bizarre art styles and diving to the headfrebs in storytelling, may be a great place to go to midgarney. Especially with the option of saving images in the collection, and the best artwork of mid -jorney users has completed with its signals for inspiration.
If you want accuracy and have the ability to achieve consistent results from a simple signal, Chatgpt should remain the image of your choice. There is no learning state here, and experience is comfortable, even if you have no experience with AI.
More than Tom’s guide
Back to laptop
