Taronites

Funny eyes variations!

Yes the Eyes were fun to do for sure I learned a lot doing those.

Took me a long time to find my way to this but it was fun getting there. Stil a lot of AI mistakes. Just pretend the guy on the left got hit by a turkey while he was eating Jello

" I want to know whose bright idea it was to have Thanksgiving outside???"

I wonder what they will be thanking for!?
Quite horrifying and hilarious. A total success!

Heh, It was too fun for me to ignore, mistakes and all Thx Taron.

I decided to revisit the Magical Beer Tap image, as I have written a vignette (finished but not yet done, almost a paradox, but it needs smoothing, and the impact is not there yet. So percolating) But there is definitely a story now tied to it. I wanted to see two things. How close could I come to the original image, maintaining it’s Magic and Spirit? and How the change of AI will affect things? The original which I posted earlier in this thread was done in Artbreeder Collage. The AI I am using here is Mage AI and you can do a lot with the free version. https://www.mage.space

Take any image, personal work after you have familiarized yourself with the controls. Click on the picture icon at the end of your description and import the image you want to adjust. Once its in the import box you have the option of adjusting how much the image x description ratio should be. I went with 50/50. Higher values for either / or could be 75% image and 25% description etc. Finish importing and look at your controls and start playing with them and rendering variations homing into what you want. Expect to make lots of adjustments as you try and adjust the description. Use remix as the images are done. It will maintain the same seed you are working with. Rerun works to provide different looks.

I tried a lot of Negative Prompts (down at bottom of the control field) such as Mangled fingers, Overlapped fingers, Fused fingers, Crooked Thumbs, and still couldn’t quite get what I wanted. In my favorite the fingers were still having probelms. But I managed to keep his spirit and general look. AI is definitely improving overall. By the way, the finger problem is easy to correct in Photoshop but I left as is to show AI progress.

I included a couple of variations along the way.

Next Post (next time I get bored lol) I will go into how you can create a Stable Diffusion model on your own computer and train it with your style by feeding it your images or at least I can point the way. By doing this you can maintain copyright as long as you use royalty free images and place emphasis on your personal style(s). Now to bring that Thanksgiving image into Mage for a challenge

First image the is the closest I could get to what I wanted but I liked the variations.

SOmehow it looks to me like a magical espresso machine, especially the last one, which I wouldn't mind dreaming about, hehe. The first one, too, looks utterly magical and quite delicious! While I couldn't properly guess what drink comes out, I want to have it!
The concept of training your own diffusion model seems really fun, by the way. Probably also rather frustrating, considering the amount of training it probably needs in order to positively surprise you, hehehe, but then... I honestly wouldn't know?!

I'm glad you did get bored, though!

Cool results!
And cool to see someone using AI!

Taron, The Vizcom xperiments you and Pilou have been doing are outstanding. I loved the dragons and the story.. and also really like the ghoul you did! And Pilou has been on a roll lol

I went back into my grab bag of bad images I did in the 90s but have held onto for inspiration. I did the texture on Goth .. she isnt wearing any 3d clothes. And remember spray nozzles in Painter in the back when? I created a large number and tried an experiment (for me anyway) and put it together in Photoshop in layers. The base layer consisted of a generic Bryce sky water with mountains in the background scene. Then I took it all into Painter and sprayed trees in the background then trees to the side, a new layer had a cutout of Goth, then the last layer Bryce plants I made a spray nozzle from. Not being as good as i am now at setting up 3d scenes I was happy with it

I did the AI version in mage and this was the closest I could come to her and the wolfs pose and have it look good. The AI ignored the funky gremlin sitting in the bushes smoking a cigarette and so did I I am sure you can see the ugly little bugger. This was the closest I could come to the original without suggesting Any artist. The prompt was A Female Adventurer and her Wolf on a cliff. There is an image to image option that took my original to this image in the settings and exactly at 5 setting after the original imported. I was surprised it came as close as it did.


and the original.. this wasnt exactly my best from the 90s lol


And before I forget here is a link to training a stable diffusion AI model. They refer to Photos but I think anyone who can sling a pen can do this and illustrate their style. I have more links but will post in the future when I know enough about the subject to discuss it+.
https://tryolabs.com/blog/2022/10/25/th ... own-images

Fresh air!

Thanks Pilou, That was something I wanted it to feel like in the 90s. Looking at it closely I see Jagged cut pixelsaround the mountains. They were another spray nozzle I made for Painter
This next I did using a trained Lora on an abstract I did in verve many years ago. It seemed to maintain most of the symnetry in unexpected ways. 1st the original abstaract posted in my Verve thread and then the AI... maintained at 50% Image/Prompt. Prompt used was Prince of the Dreamtime. Stable Diffusion model was majicmix-fantasy-v3.

That came out GREAT! Yeah, it's so dangerous, the temptation of the A.I. ...I can't stop goofing around with it either. But it's a brilliant process to learn, really! Along the way you also realize the weaknesses of A.I. as it still is. You begin to recognize their reference pool and tendencies, what sort of material it has a lot of and what not. Vizcom, for example, is jam-packed with nude, asian girls, fashion accessories and mechanics, happy, beautiful people (mostly asian) and all around advertising like material. If you want something gritty, harsh or unhappy, it tends to fail miserably. I mean, the word "angry" in the prompt is offensive?!
I haven't tried Dreamtime in a while and never in my current context of doodling with Vervette and then letting the AI figure out what it is, HAHAHA... so fun! And really beautiful, too. Surprisingly often it's not safe for work, though, so to say, hahaha...

Anyway, great result up there!