!!!THIS IS NOT A MODEL FOR CASUAL USE!!!
I made a disclamer, so download it and use only if you struggle with my frames Loras (or any other, that use 2x2frames or 3x3frames tags). In short - it's a model that knows a basic concept of frames as a composition class, if that makes sense. Supposedly it gives more control of grids Loras, and at the first glimpse it really does.
This model was made like this:
Firstly I trained a model on about 500 images of frames;
Then I clean it from any traces of the model I trained it on through merging;
Then I merged it with perfectDeliberate model a couple of times to get this result.
Basicaly it's a perfectDeliberate model with frames, but I found a couple of downgrades in quality on my model when generated casual (not frames) pictures. So if you want to use as a standalone model - just use perfectDeliberate instead, it's just better.
How to use it:
Download any of the grids Loras;
Use this model as "guide" in txt2img (just generate a couple of pictures in 512x512 (2x2frames tag) or 768x768 (3x3frames tag), till you are satisfied with the results, there will be artifacts, because images are small);
Then change the model to your liking, set denoising str 0.5-0.6 and lower the Lora weight to 0.4-0.6, and upscale it as much as you can;
Also you can make gifs whith this model alone, the best results I get was with hiers fix on 0.5 str.
Guide on how to use my grid Loras.
Know issues
Sometimes here and there appears strange "glowing" artifacts. I guess it's because of the baked in perfectDeliberate VAE, or maybe I did something wrong when merging models. I'm trying to get rid of it.
P.S.
Feel free to ask any questions here in comments, or in my discord channel.
Also I am making games with AI arts. They will be free in the future, but if you want to participate in making or have an early access - you can support me on patreon)
Also my wet hair Lora was banned here, I dunno why. You can download the last version of it from my patreon for free.
Description
FAQ
Comments (10)
Hi thanks for this , Iahvent tried your other gif making Loras, but I was wondering do they work well with using another character Lora with them? Or do they only work well when genning a character from scratch? thx!!
I wondered about this as well
@Derpasta @Swamii7 It depends - for some Loras character Loras is usable, with some not) I am trying to make them versatile, but then people start complaining about being hard to create a grid. y mModel should help in txt2img, at least it worked for me with some character Loras to the certain degree) Lower the frames Lora when txt2img, most likely you'll get something similar to a character) Then go to the img2img with that generated image and lower the weight of the grid lora again. Optionaly - you can change the model on this step. That should do)
@aDDont Thanks for the detailed response. I will try to make some time to try it out later this week.
Hello! How do you get your gifs to look so smooth (Like in the t-shirt lift and missionary frames examples)?
@classypastry flowframes) https://nmkd.itch.io/flowframes
I just wondering..can we make frames without lora with this checkpoint? just by telling what we want . Great work by the way . This checkpoint is a brilliant idea
@Zhjsjsjsksk27 Well sort of) you can use 3x3frames tag, and make a open pose guide with controlNet to make a grid with anything. But results are similar to the img2img batch processing. Less flickering though)
Hello,
Just wondering whether its OK for me to prune this model. Will it still work if I prune it?
I have no idea, but I'm not against it if you try)