OMG!
I tried to evolve my Beira pic - kept the same seed # and changed a couple of keywords in the prompt and sheās the perfect crone Beira!
I tried to make her summer or fall faces but changing the prompt to anything other than wintry changes the whole face too much. But Iām happy with her Spring vs Winter face so let me stop that.
Okay. I have an idea for a cool graphic. Be right back!
I tried a few different concepts of using the young and old Beira together and I think this one creeps me out the most which means Iām on the right track.
Still not sure how to get the lips to blend properly since theyāre not in the same position on the old Beira. Maybe I can clone them or something to move them up a bit.
The original images are here, btw. Iām making a collection of the interesting ones.
Warp tool with enough opacity to see the younger lips below it.
Also, old lips naturally lose their plumpness. Something Iām starting to notice on me at 40, so the shape is what Iām noticing being off.
But if you really want to up the wrinkles more, you can overlay the right side of the old face onto the left and blend them together to get more wrinklesā¦
The old face is below the young face. This saves me from having to go through the process of creating eyes for the old Beira. Plus some details rendered quite strangely in the old one. If you look at the collection I linked, sheās got some weird cotton balls float in a circle in front of her. What is that supposed to be? I had to fix her hair (with a clone tool) since I wanted to show off the receding hairline but a cotton ball was in the way. lol
Iām using Pixlr/e for this. Iām not sure if they have a warp tool. Maybe liquify would do the job?
I never had plump lips to start with. I donāt even want to imagine what theyāll turn into laterā¦
Every face has been run through that face blender: the distorted ones are just the program trying to read masks, and the more realistic ones is those distorted ones blended with a similar face, to just fix the issues.
Confirmed, Liquify tool has a push option. It does the job. But Iām going to have to redo it from scratch because unfortunately, Pixlr doesnāt store layers properly when you try to access the same project the next day. Very glitchy. Sigh. But it shouldnāt take too long. I know what Iām going for this time. I just need to push around the lips and then cut out a bit of the face to blend the two looks and it will be done.
The biggest reason why I have to redo it though is because I need the full size images.
I was researching ideas for how to position the title when I have such a busy image and I found an amazing blog. Maybe I should post a whole thread about it because this really opened my eyes at the design process.
In any case, the Rule of Thirds and the Fibonacci sequence led me to this design (which will require the squarely sized original image, which Iāve got - just need to do the Pixlr work on it again). The title doesnāt disturb the image and the image doesnāt make the title invisible. Both have their space. This is awesome.
GIMP is free. The GIMP fiel allows you to access everything youāve done to different layers as long as you havenāt merged them, and what you see on the screen is what it becomes as jpg file.
Well, it distorts the AI artifacts that are in a badly āreadā face, so it looks more real, and could even be confused for an art piece of āreal peopleā in the one that looks more human.
I read a good article recently that discussed how thereās an organization that advocates to implement rules to the machine learning. Copyrighted works were used to train the algorithms and no one ever bothered with getting the artistās permission.
Theyāre advocating for creating a space for the artists to opt in or out of being used for the algorithms and to also give them the ability to choose which images to use.
AI art is definitely a very shaky ethical ground. As a fellow creative, I do think about that often. Itās not only the problem of copying the work of others but also biases that are being imprinted into the algorithms.
Iām feeling a bit pessimistic about us ever being able to control machine learning process since itās not one organization but many and there are no industry standards in place at all. Itās a brand new industry after all and everyone is in a race to create the best AI generator on the market. Itās going to be while before things settle down and in the meantime ethics are being trampled.
Hereās a site where artists can check if their works were used to train the algorithms. Type in the name of your favorite artist and weep.