Since the Christmas season is over, the cover illustration has changed, but the model has not.

日本語解説は下方にあります

Nonnette is a miscellaneous model that can be found anywhere.

I am still new to the Illustrious model, so you may see some odd output from time to time, but please bear with me there.

Basically, there is no problem if you use it according to how to use Illustrious eps models.

v010 has not yet been merged with the adjustment LORA, etc. so it will not have the same feel as the model I used to publish.

v020 model was created by lightly fine tuning the v010 model with my lametta output image, then merging the merged model with the Lora model created from the lametta output image.

output is now more animated than v010 with each adjustment.

“bright pupils,white pupils,” recommended for Negative prompts in v010, is not necessary.

How to

Sampling method Euler a or DPM++ 2M (You may be able to use something else (I haven't tried it))

Sampling steps >20

CFG Scale >3 (The adjustment was made in 5)

recommended that the Hires. fix be implemented.

ESRGAN:Denoising strength About 0.4~0.5

Latent:Denoising strength 0.6 and above

It is still easy to increase the number of fingers due to lack of adjustment. I don't know what to do. hummm…

As a recommended prompt

Quality prompts sample

masterpiece,amazing quality,very aesthetic,absurdres,newest, (Adjust to your taste)

Negative prompts sample

lowres,bad quality,worst quality,bad anatomy,sketch,jpeg artifacts,ugly,poorly drawn,blurry,watermark,simple background,transparent background, (Adjust to your taste)

v010 should add more " bright pupils, white pupils, "

merged recipe

v010

lilybreed_v10 + (kiwimixXL_v3 - daisybreed_v10) = A

A + catCarrier_v20 = B

catTowerNoobaiXL_v10 + whynotIlluWainoob_v30 = C

illustriousXLPersonalMerge_v30Noob10based + sakuraplatinum_v20241117 = D

obsessionIllustrious_v30 + holyMixIllustriousxl_v1 = E

daisybreed_v10 + dreammix_v10 = F

C + D + E + F = G

B + G = Nonnette_v010

v020

Fine tuning nonnette_v010 with 520 lametta output images = A

ntrMIXIllustriousXL_xii + rillusmRealistic_v11 = B

Merging B with Lora created from 60 lametta output images = C

sakuraplatinum_v20241117 + waiNSFWIllustrious_v70 = D

Merge Lora created from 60 lametta output images into D = E

A + C = F

A + E = G

F + G = nonette_v020

built-in VAE is liquid111vae_sdxl9745, change according to your preference.

Thanks to the authors of each of these wonderful models !!

-----

ノネットはどこにでもあるような雑多なマージモデルです,

私自身まだIllustrious系のモデルに触り慣れていない事もあり、調整不足から奇妙な出力があってもお許しあれ,

基本的にはIllustrious eps の使い方に従ってもらえれば問題なく使用できると思います,

v010はベースモデル的なやつなので本当にどこにでもあるような雑多なマージモデルです,

v020はv010をlamettaの画像でFine tuningしたものに、同じくlamettaの画像から作成したLoraを当てたマージモデルをかけ合わせて私好みの出力となるようにしました,

v020は以前推奨したネガティブプロンプトの「 bright pupils, white pupils, 」は基本不要です,問題がない限り入れない方が多分きれいです,

使い方

サンプリングはEuler a か DPM++ 2Mが推奨で、モデル調整は主にEuler aで行いました,

他のサンプラーは試していませんが何かしらの絵は出ると思います？

CFGは3～7あたりが推奨で、調整は5と7で行っていました,

Hires. FIXするとより良い出力が得られます,

この時ESRGAN系(4x-AnimeSharpとかR-ESRGAN 4x+ Anime6Bなど）はDenoising strengthを0.4～0.5付近にするのがおすすめです,

Latent系なら0.6～0.7あたりが良いと思います,

調整不足からかやっぱり指は増えやすいですね,どうしたものか,

推奨プロンプト

クオリティプロンプトとして

masterpiece,amazing quality,very aesthetic,absurdres,newest,

あたりは入れておいてください,適時調整してね,

ネガティブプロンプトとして

lowres,bad quality,worst quality,bad anatomy,sketch,jpeg artifacts,ugly,poorly drawn,blurry,watermark,simple background,transparent background,

こんな感じで、もちろん適時調整してね,

v010は「 bright pupils, white pupils, 」を追加で入れてください,

v020は「 bright pupils, white pupils, 」を入れると場合によって出力が劣化します,

マージレシピ

v010

lilybreed_v10 + (kiwimixXL_v3 - daisybreed_v10) = A

A + catCarrier_v20 = B

catTowerNoobaiXL_v10 + whynotIlluWainoob_v30 = C

illustriousXLPersonalMerge_v30Noob10based + sakuraplatinum_v20241117 = D

obsessionIllustrious_v30 + holyMixIllustriousxl_v1 = E

daisybreed_v10 + dreammix_v10 = F

C + D + E + F = G

B + G = nonnette_v010

v020

nonnette_v010をlamettaの出力画像520枚でFine tuning = A

ntrMIXIllustriousXL_xii + rillusmRealistic_v11 = B

Bにlamettaの出力画像60枚から作成したLoraをマージ = C

sakuraplatinum_v20241117 + waiNSFWIllustrious_v70 = D

Dにlamettaの出力画像60枚から作成したLoraをマージ = E

A + C = F

A + E = G

F + G = nonette_v020

内包しているVAEは「liquid111vae_sdxl9745」ですが好みに合わせて差し替えて使ってください,

上記の素晴らしいモデルを作成してくださった方々に感謝いたします,どうもありがとう！

描述:

Second version

I aimed for lametta-like output with fine tuning and Lora, but this is the first time I have created a Model with this approach, so I am not sure if I did it well.

训练词语: