Well maybe more than 12 hours ago, all text rendering are almost perfect; finally some straight and clean line art of flat or drawing styles instead of some gritty intricate overcooked “masterpiece”.
Well done dev, thank you!
weird. i was hoping it was the bad parts that would be permanent. i wonder who will be right!
I’ll be tiger 🐯
Head up, MinoriMirari was probably banned on official discord and speaking nonsense.
Oki? I wouldn’t waste my time making this up. I was Dismissal from it yeah. Also, “Heads up”…
Heads"
Yea but if i wanted to add text to my images it could just do it the old way its really easy its all the other stuff that is hard/impossible without ai
The Tigger has lost it…
‐‐dude
‐dude’s details
‐ extra dude stuff
‐‐Wearing
‐blue‐jeans with some patches yo
‐plaid T‐shirt for the squad
—.
{(‐dude says “curly brackets reduce noise issues between different entities and boxy brackets stacked on eachother impedes noise bleeding from boxied details” speach bubble)}.
yuh you have to backslash…
I second this feeling. I use custom styles (output from a script, long story), and the results are a lot more consistent across generations. Very nice.
what script… wild… Im so bored without multi modality you have no idea.
I mean I figured out how to get text working with out any issues even when using formula blocks. it has to do with proper systen chain and or python prompt layering separation of the details related too text perssy too avoid noise issues. plus I had designed some cool text tools but for now those are all out the window as "multi modality is offline! the dev is likely just doing this temporarily because improvements on other things “text ai /character Ai” are being worked on… Enjoy the literal bland sketch work art if that’s your taste its about all the modal can do in it’s current state. 🤣🤪😁🙃
when ever it comes back online GG,… I will / may drop a post on something called RltDif the easy environments stream lined fine tuning system. although I will still use formula blocks here and there “that’s just me, you or if you do know me then yep” As for RltDif: Reality Definer tool kit will be released likely on lemmy world along with some information, tooling examples, and how too build novel Reality environments. Which for you intermediate prompters “those between casual natural language and Advanced formula architects like myself” will have a huge playing field to work with! end of disclosure for now 🤣😅😁🙃
just critique it over too screencap “styles” You lazy Tigger!
Its only “good” if all you make is anime/hentai. All the rest of us who enjoy ANYTHING else than cartoons/anime are out of luck.
Just to get photoreal images we have to use alot of tokens, reducing space for the actual prompt wording.
and without weights its even harder!
sure realistic style might be damaged by the amount of negative feedback reports showing. But as far as I can tell they are still thriving everywhere.
