Failed / crappy Lora collection. (But I suppose people will be able to do something with it?)版本Squirmles (ID: 365568)

Objectif :

Here will be a "failed" Lora directory, which I did not post for many different reasons or because I was simply not satisfied.

I know the Loras are failures, no need to give me a single star.

Some may say to me, “Why post them if they’re a failure?” Well... I guess they're duds in my opinion. Not for everybody ? Make a great generation out of shit but at least make that shit shine if you can do it? I don't actually know anything about it.

Black moment pic: Black moment pic: I guess this Lora was supposed to create a square image of a character with a little text or word underneath. The typical weeb photo that doesn't find a wallpaper at the time. (Read the comment of stapfschuh to have best result)

Cover me with my waifu: A Lora who was supposed to create a background with your favorite waifu. EXCEPT ba... it's shit huh. The good result happens too randomly. I recommend using ADetailer (left to right with separation between your prompts).

Dark Obscura (I certainly had a big lack of inspiration for the name of this Lora): If I remember correctly. This Lora was supposed to generate outfits with a strange aura, something dark emanating from them. EXCEPT... This Lora only works on realistic and I didn't want to get comments from people like "Yes I tried on..." NORMAL IT'S TRAINED WITH REALSIME . In short, it served me personally and fits perfectly into this category of Lora.

Gradient Led : This Lora was supposed to create a LED gradient effect (logical given the name). Why is he here? Because my dataset included crop images by myself (don't do that, it's a very bad idea for your Lora). Regardless, I think there's still a way to have 5 to 10 minutes of fun until it breaks your balls.

Soul Eater Cover : Soul Eater Cover: This Lora is used to create a manga cover in the style of Soul Eater. Well, I clearly didn't find the result conclusive to post it so thank you to this category for allowing me to do so.

Squirmles : Here is THE lora where I struggled the most (for not much eh)... He is supposed to create the famous Squirmles. Well... The AI is clearly creating mutants Squirmles. She certainly can't understand where the head is, where the eyes are etc etc... I tried it several times a few months ago though! But impossible.

Leaf Clothes: This Lora allows you to create leaf clothes and simply natural clothes... Why didn't I post it? Because it just doesn't work all the time. Sometimes you'll have pretty plant clothes and other times... Just a green outfit. (You can occasionally create mannequins. I'm talking about mannequins: puppet).

Face Melting : Face melt is supposed to make a melting face effect. The principle is good, the concept could even be great in fact! Except when the training goes poorly and the dataset is bad (yes, I'm clearing myself of a problem when in fact it's clearly me who messed up). In short, the Lora does shit 9 times out of 10. The Lora eats up absolutely all the styles to put on an ugly realistic style. It also greatly deforms the hands. It doesn't work with all checkpoints (otherwise it would be too simple).

Waifu Phone : This Lora had the concept of being able to create a real character who holds a phone with the face of an anime character. Well, already the HANDS: You will very often have zoiberg hands, ugly and misshapen. The ears are sometimes also very poorly placed and the forehead is extremely large. Nothing more to say I think. (can help !)

Moth Costum: This Lora's goal was to create a cape/costume of a butterfly with wings (without adding wings). But by dint of doing tests, reducing the Lora or increasing it, it was really really unstable. Some photos look really pretty but for the next generation it becomes a disaster.

Close-up Sharingan : Well I don't need to explain the principle of Lora I think.

Enjoy <3

描述:

训练词语: worm_god