AI portrait with Astria

Today I played with AI. It is called Astria and it allows from a series of photos to create portraits inspired by artists or various graphic universes.

For 5 US dollars, I was able to make about 40 series of portraits. In the heap there are interesting things and others, let’s face it completely missed but sometimes we get some nuggets.

If you try, I suggest you don’t use photos where you smile, AIs currently have a lot of trouble with the representation of teeth and hands.

I will post you some series that I was able to obtain with the successes and the failures. On many I have a distorted face but maybe I didn’t provide enough photos for the AI to work. (I provided 20 where the AI asks for between 4 and 40)
Finally I removed the too daring images. For some reason that I don’t know on certain series I have been represented super sexualized and several times topless…

Here are pre-recorded series that the AI produces automatically without being asked anything:

80’s Portrait:


Woman Fashion Portrait:

The results are pretty good, even very good, but I guess the developers had to work on its terms for a long time and if they offer them as a base, it’s probably because it’s advantageous for the AI, let’s see what that is. gives on personalized topics.

Personalized topics:

Demon Hunter:

As a Diablo fan who can’t wait for Diablo 4, I wanted to try it and here it is, the images are nice but I don’t recognize myself.

Elf Woman:

I wanted to see myself as an elf. When I was talking about hypersexualization rather why am I in my underwear on last one!!!

There is some nice stuffs ans some are really random I think.

Police officer:

Wanna see me as a police woman cause I’ve played a lot to Counter Strike with friends.

I kinda like it.

I think I have been a little too generic in my requests so far so I wonder how the AI would do if I offered it more marked universes by asking it to represent me in universes of films or tv shows.

Nilsa as an actress of cinema?

Blade Runner:

What else than one of my best movie of all time?

I’m a little disappointed with this series. It is a bit random or I dont reconnize myself.

The lord of the rings:

I’ve tried with Game of thrones but the AI juste faceswap me with Emilia Clark so I’ve prefer to post this ones.

Space marines in Alien movie:

I love Alien’s univers in Alien 2 and in Aliens vs Predators video game series so I’ve test it.

There is some default but I found me so badass on those! I love it!

Something that doesn’t work:

I wanted to see if the AI was able to “imagine” me as something other than human, but for the moment it can’t. When I wanted to become a demon he just undressed me. (I don’t know what to think of the fact that to demonize a girl this one undresses her ^^) (there was nude pic on other try so I’ll didn’t post it. ^^’)

Then I wanted to see myself as an angel but I did not have a better result. I tried to stick wings on my back but I’m still looking for them

In the end, this AI is interesting for making avatars for our social networks, but I don’t really see what else it could be used for now.

  • I find that at the time of 4k and soon 8k the images are of very low resolution.
  • The framing is sometimes really rotten. I only provided portraits and I end up with images without my head!
  • I think that for an optimal experience I should have provided as many photos as possible with as many different angles of view as possible, which would have minimized errors, however I don’t know what happened to the images we provided.

If you wanna try it, there is the link:

An AI draws for you, soon the end of artists? !!!

You may have noticed that for some time now, my articles have been illustrated with paintings and that I have not posted any artist links in source. The simple reason is that they are mine. Finally from me with a little help, that of an AI: Midjourney.
Midjouney is an AI that creates paintings from a sequence of words that you give it.

Here are some examples:

To use it is very simple, just go to this address:

Sign up with a discord account (a chat program) and start giving him instructions. There is a free demo and 2 subscription models to not be limited in the number of images generated.
He ends up generating 4 images, from there you can ask him to start over, to launch variants of an image he gave you so that he comes up with 4 others approaching it or upscale one of the results.

I used it recently to illustrate my articles. It is an interesting tool for me because even if I like to draw and paint, the time that I will spend making the illustrations would be time that I will not spend writing.
Until then I used google images but these are not necessarily copyright free and above all they are not original which is important to me.
The other solution could have been to go through an outside artist but this has a cost that unfortunately I cannot afford, especially when trying to post one article to you per week.

Hey, but Nilsa, are the artists going to disappear?

The answer is no, for several reasons:

  • The sequence of words that we use is of paramount importance to generate the images and depending on what we choose, the result can be very different. Then we choose an image among 4 that we choose to modify or not from the AI. Those are already artistic choices.
  • The images generated sometimes have bugs and would often need to be retouched so that they are less shocking to the eye.
Headless gorilla
  • Finally I think it can be a great tool for all designers / painters. Midjourney is a great way to get inspired and I think the images generated will mainly serve as a base for repainting over them as we already do with photos in what is called photobashing .

There are no copyright issues: is such a creation the property of the artist?

That’s a delicate subject. Not being a lawyer specializing in intellectual property, I would not be able to say who owns the image generated by code which is the intellectual property of the developers. Is it their property or is the AI ​​considered a mere tool and thus the ownership of the work produced belongs to the user of the AI?

I do not know enough about the law to be able to decide on the question but this one is very interesting. It would be necessary to search the CGU of Midjourney and I admit to you that I was a little lazy.

After when the artist repaints over the generated image, he appropriates the image and creates his own work and in this case the final rendering becomes his property.
As in photobashing, if you have repainted and modified the image enough so that we do not recognize the original photo used then there are no copyright concerns.

I am not sponsored for this article if I tell you about it it’s just that I find this tool really magical.

I know it’s different from what I usually do, but I hope you enjoyed this article.

Take care of yourself and your loved ones and see you soon.