Blondie - UHD set
10mp + versions unedited for patreon tier 2 and up
2024-03-22 21:47:41 +0000 UTC View Post
DV posts
Style B Bloopers
105 additional images not selected
You may notice similarities in the styles, which will become more apparent.
There is a base style 80% of the art takes the line thickness, shading, and overall mood is the exact same all styles have to have similar coloring too.
STYLE B - STYLE RUNS (comedy anime style 1)
These are 7 to 8 styles; I will try to post 1 or 2 a day.
STAY TUNED
The illustration combines expressive faces with everyday situations to create a subtle sense of comedic tension. The character's features, such as puffy cheeks and wide eyes, hint at a humorous moment or a light-hearted misunders...
STYLE A - Bloopers
Generating 16 of each.
16 is the sweet-spot in AI to see stuff imo
STYLE A - STYLE RUNS
VOTING WILL OCCUR ONCE ALL STYLES HAVE ALL RENDERS.
I will pick a few to vote for after renders. But you get to see all styles and fails of them
STYLE RUNS -> STYLE A We're picking from 7 styles for a vote This is the first of the 1600 randomized-style runs. I was VERY BUSY. But damn, these are all goodโall...
2024-03-21 04:56:12 +0000 UTC View PostCOOOKING
Nina's styles i chose need one image per transformation to show so its being cooked as we speak 112 hours later we pick the 3 best I will leave that poll lol
2024-03-20 23:48:23 +0000 UTC View PostTech Sharing: ANTI-HIVE-AI-DETECTION-NOISE
I'm also an engineer and will give this tech to only tier 3 Patreons
Basically that nightshade stuff, I built one that does the opposite and spoofs itself to look like traditional art, so here is the deal I do not want this shit to spread because they can easily detect it if too many of us use this.
Update for backers #1
We're using Animagine v3.0
with A Lot of prompt engineering
https://civitai.com/models/260267/animagine-xl-v3
Previously, we used a custom model, NinasLeap V1.0
I will upload the model; give me some time....
process to pick style. blind.
So the goal is find style that expands well. randomized the style parameters that worked out 1600 to generate 100 supposedly good style choice, then blind shot per transformation, style of the game should it all, AND be appealing.
2024-03-19 13:40:03 +0000 UTC View PostStyle Search complete, randomized lookup, 7 out of 100 styles that supposedly worked. will be polled have to choose 3...
Grids to test
400 or so have the capacity to 'grow' so we have to see 60% of the max transformations anything extra is reworked by hand ai has its limits
2024-03-18 19:15:47 +0000 UTC View PostStyle Process (explanation)
unordered/uncorrected images, some will be reposted sans the watermark I have a script adding those ignore it; there is 1600 images thus far, picked less than 50 for this post, my hard Drive is dying lol.
We have been polishing with prompt engineering to get close to the style we wer...
2024-03-18 17:37:47 +0000 UTC View PostAI advancement, sketches to art?
Didn't know this was too posible. but with some tricks you can just autocomplete shit?
Back to the game: Open Pose SDXL Tasks
completes: https://trello.com/c/BY8OqsWI/453-nina-needs-baseline-for-control-net-pose-research-on-sdxwork2
Knew for a fact there'd be problems, I can take a photo and just have a pose??...
2024-03-17 22:21:16 +0000 UTC View Post
Office Cute (SD) - Ultra-HD tier Announcement.
New Tier 2 benefit is Ultra HD versions of posts that have 10MP. This allows me to give tier twos and threes something else on top of things. Only for arts above 10MP meaning above 10'000'000 pixels, you still get all the same you did before. This is something above that.
2024-03-17 18:21:38 +0000 UTC View PostOffice Cute (15MP 18MB Ultra HD ver)
New Perks Is Ultra HD versions of things. So there you have it
2024-03-17 18:02:26 +0000 UTC View PostWallpapers!
Looking for perks to add to patreons
Now I personally have never downloaded a wallpaper my phone wallpaper is my cat. it implies I may make more 'explicit art' for them too, it has to be something you'd want to download. measuring the temperature.
Morning (COFFEE)
sdxl practice I have to attempt to do stuff that is hard to do in prompts
Nina's Leap, style change incoming. We will vote later lol.
Result of research:
Creates tasks: https://trello.com/c/ws0EMpOT/478-nina-needs-a-new-style-with-pollresult-of-sdxl-not-having-anything-we-had-bef...
HD of forest boobs
Doing some retouching testing on sdxl, same issues at less of a problem, you still have AI putting attention to a cup 9 meters away, missing the point of drawing; drawing is about hierarchy; AI tends to still a lot of this.
+focus on head, and bust
+colors to clear detail of the forest I removed colors and blu...
2024-03-16 20:42:01 +0000 UTC View Post
guess its time..
Stable diffusion XL actually creates good nipples and other things. So I want to be real here; we need all the money we can get. It would have to be for something.
Patreon forbids me to give it on the free tier, just so you know. and this is a game dev patreon. We'd have to l...
2024-03-16 20:07:29 +0000 UTC View PostNavigating New Realms: My Odyssey with Stable Diffusion XL
completes https://trello.com/c/b1OdkAPV/476-nina-needs-you-to-figure-out-sdxl
In the ever-evolving tapestry of AI art, a new chapter has unfolded, marked by the emergence of Stable Diffusion XL (SDXL). This... 2024-03-16 19:11:19 +0000 UTC View Post
About Stable Diffusion XL
I'm doing some practice rounds of prompt engineering and coloring and correcting. Here's the gist of it. I can't use my old models in SDXL; they're not compatible, I still believe SDXL will give us much better specs. I have to really use the experience to understand if this is a good idea at all.
What is SDXL
Its t...
