AI digital art (Dall-E 2.0, Midjourney etc)

I went to a dinner party last weekend and the host was playing with (non adobe) online AI to create a promotional poster for some flute players (also guests). They generated the first image and the flutes were munted up with wrong keys on the instrument (as you’d expect). And so the image was unusable with the guest saying ‘we can’t use that! besides, I know the guy on the right’.
I was shocked that was even possible, to generate a living persons likeness. Wild Wild West

 

So I've ponied up money for ChatGPT plus for work. I use it mostly for text based stuff. I'm currently a science teacher, so it can save a huge amount of time, and more importantly the principal wants me to use it extensively.
But I got to say, AI sucks for technical images. Getting it to generate diagrams of even simple things, even with advanced models, is, terrible. ChatGPT is quite good at creating image prompt though, for other image generators.

However, my Kylie Minogue and Gillian Anderson detective series is coming along excellently. It has no issues generating that.

 

Got back into this and figured out why Flux models weren't working. I had to change from Automatic111 to Forge - https://github.com/lllyasviel/stable-diffusion-webui-forge - Getting it running was a task! Incompatible python, torch, cannot create venv, etc. Lots of googling at each roadblock but I got it up and running. Now I can use Flux successfully. Then some extra googling to find out why Lora's weren't applying.

 

Those are great.. Im nuking everything and starting from scratch..

 

this AI thing doesn't work very well...

 

Where'd you get that picture of me Mike

 

Looks extremely like me as well... uncanny stuff.

 

Brother?

 

I don't think so... but maybe the ocau lifestyle moulds our DNA to a perfection that others can't see?

 

whatcha talkin bout willis
Elmf RnR

 

You forgot the letter 'o' towards the end of that word

 

oyes that's for sure.

 

How is it getting the words right? Is this a new generation of AI? I've never seen an AI that can consistently produce coherent English.

 

it mostly got everything right because i issued it the command, but missed the . in .com.au and the mug says OCAN (and looks weird). feet bursting through bottom of case and issues with the tower. overall pretty shit but i think it's pretty random. i've asked it to fix stuff and it's tweaked with everything (even stuff i specifically said not to).

 

Text has been possible for many months now with Flux and SD3 local generations (June last year?)
Online services have improved it further to the point you can generate a whole page of coherent sentences.

.

 

That image and text Immediately reminded me of Half-Life..

 

I was having too much fun with the chatgpt filters.



Haters gonna hate just cos I used a filter and they think their opinions are worth more than mine and think i am stealing as a result... christ fuck off already. I am just having some fun, i ain't making bank on these images.



sorry to the Optimum fans out there

 

I've been trying a lot of AI resizers, something that can do batches..
I want to have more detail, more stuff in the image, and at higher resolutions. It tends to smooth the faces a lot. .But I am having good success with detail.


Im using AIArty image enhancer.. I want more offline tools, because I do find that the online tools vary wildly, and when they are updated, often turn in wildly different results.

 

Anyone able to get Flux models working well on SwarmUI on 7900XTX?