A new architecture : No LLM, NO IA, is a substrat tech. priorietary technology designed by me!
Free download available :
BitsDuJour: https://www.bitsdujour.com/software/vaudryllis
eznegative is a negative prompt integration designed to apply an anime style, compatible with all templates.
--- to be placed in the folder: embeddings**this is not a lora**---
the examples do not contain any negative prompts, see the image prompts.
automatic activation by prompts: 1girl or 1boy
This embeddings is not heavy, it is even quite light.
it will suit all your creations, reducing the prompts necessary to have a good image.
he nevertheless corrects the problems coming from the vae/lora, he was trained on a wide variety of characters and styles of clothing and different hair.
Description
FAQ
Comments (8)
Like and follow me :)
new version having had a better care on the choices of the dataset and having a solid base of regularization. it includes keys on the natural language and is able to bring an appreciable correction on your realistic, anime, semi-realistic creation.
I don't get on how to use v2
for v1, a simple EZNegPONYXL-neg in negative prompt will do
but for v2, we need to include all the prompt "1girl,girl,breasts,1boy,boy,man-neg"?
these are common prompt used, wouldn't putting it in negative prompt causes issue?
no, it means you have to add it to the negative prompt and it activates in addition with the prompts I provided. a negative always activates with positive prompts, like a lora
when you put items in the positive prompt, the negative will activate with them, it supports natural language and danbooru style prompts
alright, thanks for your explanation
@n0utis Honestly, this sounds stupid. Instead of giving a keyword that only works with your embed, I need to use a set of commonly used tags. What if I don't want boys and men in the image?
@Unicom whether it is a man or a woman or other or objects, using tags or natural language it will activate! each image has been trained 2 times by tags and natural language in maximum number of words.
indeed, I hadn't seen that the file name was incorrect, a comma was embedded in the worst place.. the activation has also been modified for your understanding :)
have a great new year
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.