Home Creators Posts Import Register Favorites Logout

Content

I DID IT! I FINALLY FUCKING DID IT! LOOK AT THIIISSSS DUDE! Can you see it now! CAN YOU SEE WHY THIS WAS DRIVING ME UP THE FUCKING WALL FOR THE LAST WEEK AND A HALF! But it's fine, It's FINNNEEEEEEEEEEE! I'm not mad at all! IN fact, I'm more elated now than I've ever been in my whole life! AND THAT'S WITH THE POSSIBLE SKIN CANCER STILL VERY MUCH LOOMING OVER MY HEAD! (which I will get check in 3 days from now, looking like a false alarm they said)So Obviously. Things Still aren't one to one. Play spot the difference long enough and you will find several things that are broken, BUT IT'S CLOSE ENOUGH.23 minuets short of the 100 hour mark and I finally fixed it. I know I said I was going to give up on several occasions, but that didn't happen at 30 hours, 70, or 90, so It should've been obvious to anyone just how stubbornly I was perusing this goal. I have a nsfw example going up in a bit, which I will not be leaving a big message on, just so you guys can see how much more accurate it is when there's fucking involved.So now, you may all be wondering. "Crit! What was wrong with your mix!"And to that I say, four things!---------------------------------------------------------------1. Tag mixing was broken in the update from 4.5 to v5. Any number between .01 and .99 no longer carried the same value. I released it had to be a systemic change and came up with this Equation to fix it:

1 divided by .1 is 10

1 divided by .2 is 5

1 divided by .3 is 3.33~

1 divided by .4 is 2.5

1 divided by .5 is 2

1 divided by .6 is 1.666~

1 divided by .7 is 1.43

1 divided by .8 is 1.25

1 divided by .9 is 1.111--------------------------------------------------------------

In the v5 update, Novel started counting the "heavy Emphasis" prompt in the undesired section towards your total character count. the Problem? If you character count exceeds the 512 limit, Novel will take it upon itself to erase any sets of duplicate tags you may have. Of which, Mine included not one, NOT TWO, BUT THREE SETS of those very same emphasis tags from where I copy and pasted them back into the software after losing my prompt in the past

For some reason, around the release of the Bea set, My vibe transfer settings were changed. I don't personally remember changing them, nor did I recall the values I used in the past, So i had to go back through all my old content to recover them. AND I FOUND EXACTLY ONE IMAGE FROM APRIL 18th THAT LET ME REVERSE ENGINEER MY OLD DATA BACK! Something the novel ai website SAYS IS IMPOSSIBLE!

Finally. Sometime in may, novel changed the way seeds work. So when I was cross referencing new data with stuff I made back in febuary, it was yielding different results because the pipeline fundamentally changed .

And now here we are! As far as I can tell, It's fixed. I haven't ran any extensive tests beyond a few Cross references, but Now my style should look like how it did back during sets!Specifically around the 50 - 60 range in the vault. In case you wan't a more straight forward example: The sets that used the style i just finished reverse engineering were:Herta X Ruan Mei, Rumi x Ochaco, Rimuru X Shion, Bowsette X rosaline,

and many more. They actually very heavily align with my most popular sets to date! I also backed all this information up, and began working with someone much more talented than myself, who can make a style lora for me! Maybe. We'll See. I'll name drop them once they finish and not now just in case things fall through~I'm going to make a quick test set to see if this is 100% going to work, and if it does, I will update you guys on what I intend to do from here on out in terms of content. Cause my stuff, desperately needs an upgrade!

Files