🔴Patreon | Stable difussion and Content Creation | PANyZHAL
━━━━━━━━━━━━ ◦ ❖ ◦ ━━━━━━━━━━━━
This model create pixel images with a gride, it's not perfect and i will upload more version soon so, stay tuned.
RECOMMENDATIONS
If it seems like your model isn't working as expected, you can try adjusting the weight to 1.1 or even higher, like 1.3 or 1.4. Sometimes, setting a weight of 2 or 3 helps to identify where the image breaks, and then you can lower it from there.
Realistic models often fails.
If you don't specify a mouth, usually, the generated image won't have a mouth.
Yes, it can do NSFW.
hope you like it and if you want you can leave me a review. :3
Description
FAQ
Comments (9)
This is a great idea! the issue with most pixel art loras/TIs is that while they look good at a glance, they don't actually translate to pixel art because they aren't truly on a grid. This solves that! you can grab the center pixel and get the true color.
It would be great to get more sophisticated pixel art, like from SNES games, capcom arcade games, etc..., to flesh out more art styles and possibilities. That might confuse things though. This looks great too.
Thank you very much for your comment, I based my model on the pixel art of the "braceletbook" page, I will try to do something like what you tell me, just to be sure, it's something like this? batmanpixel.gif
Finally an actual usable model for pixel art! As Jellai said, pixel art isnt simply pixelated art. Training models with a grid makes it much more precise and, when actually transferring the image to a pixel art editor, only the center pixel color of the grid is needed. Maybe the best way to train these are in specific grids and styles, like a model for 32x32 grids, 64x64 grids, etc.
Also, the best game pixel art to use imho is SNES era or newer, specially sprites from Capcom and SNK.
@hhks yes you're right, when I searched for pixel art Loras I noticed the HakuImg extension does the same job and without affecting the style, but the grids make the difference, so I'll try what you say: train a model of small grids.
PD. (sorry for my english)
Coincidentally, I posted my PixelNet ControlNet model the same day this was posted. PixelNet is controlled with a checker pattern.
@thomaseding omg :c it's the second time this happend to me XD.
I just checked your project and it looks very interesting, it seems that it's better to create more complex scenarios .
@PANyZHAL Don't lament. The more models, the merrier. I'll have to check out what your model does with and without my controlnet.
@thomaseding Any interesting cross-breeds to report? What works best?
@mnemic I haven't had much time recently to play around with SD. I haven't tried my controlnet model with loras yet. That said, I tend to like using any of the Realistic Vision model versions with it.