ooo hi again! i am on the cusp! of major changes! but im not just there yet sadly.
ever since sunday i have been doing this!
training the LoRA overnight/at work -> assessing the outputs -> making the fixes / figuring out the problems -> preparing for the next iterations of the model -> repeat
i am on version 20 now! (the one you are familiar with is version 11) and... each version takes about 6 hours to train... sadly...
so why am i making a poll?! great question!
my main question is: should i... just put all my effort into training now? and best case scenario have something decent by the weekend? and then REAP THE REWARDS !!!!!!!!!!!!!!!!!!! (very good rewards)
or still focus on posting for the meantime (training is slower, but still some frequent-ish posts! rewards reaped in maybe 2-3 weeks instead of a couple days (best case scenario)
i'm asking this because it feels beneficial to take the risk! sort of like ignoring everything before a deadline to get it over with! but on one hand, i'd imagine more posts /content is better yes posts!
for some perspective on why this is even a decision, the previous technological jump i had made was from pony (previous base model) to illustrious (current base model). commission wise, i made the switch between commissions 89 (vicky) and 90 (zani)! i'd like to think you can see the differences! is it worth the slight downtime to result in much much better content (my opinion)? personally i think so! but best to ask you ahahahah
to describe what the differences are!
pony: travelling somewhere based on known land formations or landmarks blindfolded
illustrious: travelling somewhere with a map with blurry vision
anima: travelling somewhere with google maps with 20/20 vision LMFAO
anima actually understands language and has spatial awareness! (knows what left and right is), it does have some limitations, but it is vastly overshadowed by the pros. i'd essentially have full creative freedom! (whereas now its around like?... 40%?)
here is an example of base anima's capabilities, the entire goal of the dataset and training is to transfer the style of my images to it (very hard). normally these images would've taken me about 30 minutes - 1 hour MINIMUM! to recreate right now, these took 10 seconds each GULP! it also lets me be 10x more creative! especially with clothing
and the creatures as guys LMFAO (previously very difficult)
the three subject images would've taken maybe 1-2 hours each ahahahah
anyways with this in mind, what do you think? this is a risk! it may take a day, three days, 1.5 weeks, or longer! regardless if anima works, the base illustrious model WILL improve a ton as well!
whatever has majority i will do! thank you for reading and voting! no option is wrong!