I quite like the new image generator
and sadly i must be the only one... anyways. i've been tinkering with self-hosted models the whole day and nothing has come close to perchance. Perhaps someone knows what model/loras they use? i'd really like to self-host... Thanks in advance~
Yes... I tried some self hosted models (=> ComfyUI), but I miss a secret sauce. The results are far below perchance.org T2I results regarding... well... the "overall feeling". Most of them are plain and simple bad at quality, composition and consistency. Z-Image-Turbo was the only model which performed good (to a certain extent), but the training material seems to have been heavily filtered for "all good looking and shiny things and especially people". It's quite incapable of generating some more down-to-earth or slightly ugly stuff. perchance.org seems less biased here, which results in a wider variety of "person types" you could achieve. If the devs would disclose, what they exactly do, to modify or enhance a model - that would be quite interesting.
Same results here. So far the best results i've gotten with Nova anime XL and toony lora, but the prompt comprehension just isn't close, especially for art style. + if i then tried to gen something realistic i'd have to change the checkpoint/loras... so i'm guessing they have multiple engines running, and they are picked based on the prompt? either that or some they use checkpoint/loras that cover really broad spectrum of... everything imaginable. So far i haven't found anything like that. According to text-to-image plugin it only takes 3gb, but i doubt that. Old info i guess.
A small peek behind the scene provided by the dev would be awesome. I guess, there is something to be learned :-)...
It is pretty good once you figure your way around it yes, Infact it has more technical tools, but again the formatting is a bit difficult.
hmm updates on this info. people could track down that info, and make lists and or not, but is fundamentally free models then. you can also build models in several diffrent ways but you'd have to skim basically all the technical stuff and for formal pi torches the perchance.org discord server they have been referenced and updating those references with gemini should be possible and easy enough for some. As for deepfakes type stuff the laws around it are 1. if you know them 2. if it is hard conflicting political propaganda & 3. the celebrity laws are the most grey area technically because you have to understand marketing that and potential things the celebrity may find as defaming themselves and ground over it, The biggest one though in the area of defaming is none meme related uploads for the majority of this although is where deepfaking your friend or not that you know might not count as being protected. The actual greatest offence though is pretending to be an existing person using deepfakes, while also being a direct marketing deepfake violation. Now if Obv it is a meme then the only grey area is if they where a paid comedian. Lastly some actors supply thier modeling data and some do not while others can even actively make that harder to acquire, point being here ;study the individual actors stance on the matter with out persay bugging them. most free modeling goes under "true to real life look alike", and similar phrasing. Is a few others which are considered word & phrase itemization syntax and chaining, some of these are more or less either not formal or are a formal. I don't share much on the others as to not cause deepfake problems. Deepfakes used out side of memes should probably be kept to self that or posted very limitedly.
As for this yall could start a list of names that work with and or alternative lists that require formations of the key words true to real life look alike.
Any thing else like actual lora torches could possibly be used but start on the perchance.org discord server, confirm information with Gemini which is my go-to, You would simply need too likely perform some JSON prompting &or java scripting.