English 
Introduction
This article is a continuation of the previous article, where ChatGPT and Dall-E were used.
How Chat GPT and DALL-E can help 3D artists in their workThere we examined various cases of use, but the topic of generating pictures and textures was most interesting.
In ChatGPT, its own generation of images has become interesting to compare whether it became better compared to Dall -E?
A small spoiler - yes! It got better. Below I will give comparisons before and after.
Ah, and of course I also tried my photo in Ghibli style.
Generation of concept arts
Prompt: Modern Container House in Foggy Forest Near Lake with Warm Lightning, Realistic Style.
So ChatGPT coped with him:
And what DALL-E had:
In my opinion, progress on the face. ChatGPT generated a clearly more attractive preview. Put +1 ChatGPT score.
Creating textures and materials
Prompt: Realistic Seamless Texture of Rusteed Metal.
And here is the result that I received from ChatGPT the first time. I also checked for seamlessness and yes, the texture is perfectly glued.
The result with DALL-E was like that. I think not with an armed eye shows that the textures will be poorly glued.
Well, in general, the result from ChatGPT is more like what we expected. So here is +1 ChatGPT score.
More tests
With the generation of textures, I spent more time and also tried various types of surfaces.
I tried to generate a brick texture. The result turned out to be very clear and the gluing of the X was perfect, but the texture did not seamless by Y.
Nevertheless, this was the first option, it is possible if you try several iterations, the result will be better.
I asked to create a texture of the rug. The gluing is barely visible and this very easily runs in Photoshop, so we can assume that the first time it turned out excellent.
Here, the wood texture coped poorly, in X and Y the gluing is very visible.
Then I drag&drop the image with a rug and asked to make a texture out of it:
It turned out pretty well, despite the fact that the gluing is visible, this can be corrected in Photoshop. The main thing is that there is a good texture and there is something to work with.
DALL-E at that time did not know how to do this.
I threw the rug more complicated and I was surprised how ChatGPT coped, preserving all the patterns as much as possible. Although there is no gluing here, but I thought to be honest that the result would be much worse.
Creating Alpha channels
Prompt: Texture of smoke from Tea on Black Background.
Smoke texture, which every time you have to look for in Google, now you can get free.
The result from ChatGPT:
DALL-E result:
Here I will also give a score +1 ChatGPT.
I tried to create black and white wood textures, which is usually used for an alpha channel on the plane so as not to insert a real tree and get soft shadows.
And ChatGPT coped.
I remember DALL-E, I could not generate any option adequately, even with different prompts.
Generation of models
Prompt: Create 3D Cartoon Character of Girl.
ChatGPT result:
The result is DALL-E:
The result from ChatGPT surpassed all expectations, it really looks impressive. DALL-E, as I did a good job in many ways, I liked the first and last options.
Therefore, I will give them a +1 point.
Conclusion
Let's calculate the points.
ChatGPT: 4 points
DALL-E: 1 point
I covered only a small part where the generation of pictures of the 3D artist can be useful. But I think the answer is obvious that ChatGPT made a big jerk forward.
Options were affected that can be used here and now, right from the box and for free, let it with a limited amount on generation.
Having made a chat tandem that understands the context and generation of pictures well, Open AI managed to make a tool that essentially guesses 100% of your wishes.
Well, as the last argument, Fooocus vs ChatGPT.
If you go to the Fooocus website and enter Prompt: texture of smoke from Tea on Black Background.
Fooocus result:
ChatGPT result:
I can say with great confidence that I will use ChatGPT in my work as the main assistant - to generate ideas, clarify technical moments and even visual tips.
He does not just answer, but really helps - quickly, precisely and with an understanding of the essence of the task.
In conjunction with the generation of images, this becomes a powerful tool that closes several tasks at once.
In general, the future has already come. And it writes, draws and tells - everything is in one chat window.
{{comment.text}}