fathur_shaon47

joined 1 month ago
[–] fathur_shaon47@lemmy.world 1 points 1 month ago

Buying servers and bandwidth for long term will literally cost anyone worth of 2 mortgages in city centers, hardware costs are borderline ridiculous these days, let alone maintaining them for 24/7 in a air conditioned room. Those who provide the hosting services are robber barons - they weigh your needs and play the good buddy; and when you are completely reliant on them, then suddenly one day, the price hikes without a proper explanation, because shifting to another provider would cost downtime and extra $$ for file and model transfers.

[–] fathur_shaon47@lemmy.world 2 points 1 month ago

Stable Diffusion 1.5....... the previous tutorial is still here, don't know whether its canon for everything or specific to the model before

https://perchance.org/text-to-image-plugin

[–] fathur_shaon47@lemmy.world 5 points 1 month ago

I got a degree in English, and my prompt paragraphs, which I wrote, scrapped, rewrote and fine tuned for hours a day which led to months of brain drain to get somewhere stable with satisfactory results, is now all undone, ruined and tarnished. Either the new model itself is illiterate or we now have to resort to biblical text format(s).

[–] fathur_shaon47@lemmy.world 4 points 1 month ago

They replaced a well trained and resourceful Stable Diffusion 1.5 model with an untrained and more expensive FLUX-1.Schnell model.........

[–] fathur_shaon47@lemmy.world 1 points 1 month ago

Add to that -- warped hands, adolescent child, bony legs and a stomach loaded with worms

[–] fathur_shaon47@lemmy.world 3 points 1 month ago (4 children)

This was my refuge for guilty/carnal pleasures. As someone who personally prefers the imaginary over the real thing, I kind of feel disabled/handicapped. The new model is overtly realistic with hideous mutated output, while the old one maintained a fine tuned balance between realistic and imaginary.