【 AniReal 】
Welcome to AniReal! a latent diffusion model Trained and Fine-Tuned on Photorealistic High Quality anime illustrations using the Danbooru tagging dataset aswell as Blip I made it so that it understands some natural text description alongside danbooru tags It may not work as well though but give it a shot! The model itself its made to output generally anything with an anime art style if u can think of it u can prompt it!
Above pictures are examples generated with my model!
If you'd like to support my work leave a review <3
More examples below both SFW/NSFW
I've blurred NSFW images! this is just to showcase that my model is capable of NSFW though that was not my intent when I made this model!
I Only plan to release models in SafeTensors format for now if you're looking for a diffusers version head to my HuggingFace Repository!
AniReal VAE = kl-f8 anime2VAE
Main model page :https://huggingface.co/Hosioka/AniReal
The base model for training was Baka-Diffusion which is my other model you can find here: https://huggingface.co/Hosioka/Baka-Diffusion
Description
FAQ
Comments (12)
adding prompt, maybe?
Most of the prompts I took are from other ppl's gens most of them had "mksk style" which I believe is a LoRA but I forgot and left it out on[I didn't use any LoRA on my showcase] most of my gens I made here to avoid confusion I just deleted the prompts if theres anything specific u want lmk
What is blip?
an automatic captionning method used to describe the training images. when you use the sames word or sentences you will get a part of the style of the training image.
what is the base of the model? did you train it on sd1.5? Anything? or from scratch?
Base Model is Baka-Diffusion mentioned on the model page above!
Is the model free of Unet merges? no other models mixed in? If yes, this model has potencial to be the best base model for mixes.
I have not used MBW merge on this model yet but the current released version is a merge of 2 different models trained on the same dataset using the Automatic1111 merger
This model can create very beautiful images!
Does this model contain elements of ANYTHING series?
Baka-DiffusionV1 model has anythingV3 in it aswell as NAI,GAPE60, then theres BakaV2 which was a finetune of bakaV1 i trained those 2 with the same dataset of AniReal and merged them.
where to find : Baka_Baka-DiffusionV4A(FP16)outattxn - ca455ce14f ? i am particularly interested in your mix :D
It's named to Baka-Diffusion[S3D]
https://huggingface.co/Hosioka/Baka-Diffusion