The Flux model is objectively better at everything except speed and celebrity recognition (which is technically illegal anyway). People just need to familiarize themselves with the new style of prompting. Just write as if you were describing the image to someone with a mental deficiency and you’ll get some great images.
This is impossible from the users end; Any updates or additions to the model are done by the server owner. The only thing we can control are the prompt and the strength of the prompt.
No, we didn’t. Probably just an update/refinement. Still the same model…
Spot on analysis to be fair. It’s frustrating waiting for new gens, but I think the quality tradeoff is worth it, but understand those that don’t.
I think it will speed up over time, the devs will probably upgrade their servers at some point and the model itself will get much quicker - they always do!
Good luck! I’m still refining the art styles and getting better results as time goes on. Hopefully the model continues to be updated as the devs promised. Should improve things further…
I’ve put together a few Art Styles that work quite nicely with the new model. One of which is an Anime style. Feel free to give it a try and copy if you enjoy it…
Don’t get me wrong, it’s not terrible compared to some other models but compared to the previous SD 1.5 base, it takes about 10x longer.
I’m sure it’ll get better with time. Either way, it’s better once you figure it out…
The model is actually really fantastic, but you have to completely change the way you think about prompting. I’ve set up some base styles on my generator here: https://perchance.org/opentext2imagegenerator
If you give that a go, and be very descriptive, writing in natural sentences; you’ll see some great results. My only minor gripe is with the speed of gens, not the model itself, it’s a definite improvement…
Try this generator, the styles are much more effective: https://perchance.org/opentext2imagegenerator