chilloutNebula – 1.1

chilloutNebula - 1.1

Run Model 使用GRAVITI Diffus 关于GRAVITI Diffus
Model Name chilloutNebula
Model Type Checkpoint
Base Model SD 1.5
Version Name 1.1
File Name chilloutnebula_11.safetensors
SHA256 2c572f1fe9
NSFW False
Trigger Words
Tags character, sexy, 2.5d
Metadata fp = fp16
size = pruned
format = SafeTensor
Size 5.80 GB
Version Description

The safetensors file.

Source https://civitai.com/models/30591
Author Ai_00_Girls

2.0 Update

Please share your creations. I love seeing how other people use my stuffs😊.

For 2.0 I have merged: orangeSlushyMix_v2, casHeartMix, AbyssOrangeMix 3, shinyOiledSkin, Makoto Shinkai Substyle, 晖映 Enhanced Backlighting.

Replaced chicVibe with ShojoVibe少女感, because chicVibe have too much influence on face.

Overall the NSFW compatibility have improved, and so does the skin. It is also more character LoRA friendly. When using multiple LoRAs / LoHAs be sure to use LoRA Block Weight to minimize the interference.

However the stability have slightly decreased especially when you don’t use hires fix or img2img upscale. If you do not wish to generate NSFW contents then there’s not really necessory to update.

2.0更新内容: 提高皮肤质感(尤其是油光、汗湿皮肤)、提高NSFW能力,同时降低预设脸型的强度。但在图片质量稳定性上有所下降,如果你不想要NSFW内容,则无需更新此版本。请在使用多个LoRA / LoHA 时,务必搭配LoRA Block Weight 插件,以降低多个LoRA之间的互相干扰。

2.0アップデート内容:肌の質感の向上(特にオイリーで汗ばんだ肌)、R-18コンテンツの描画能力の向上、およびプリセット顔の強度を下げる。ただし、画像品質の安定性が低下しています。R-18コンテンツが不要な場合は、このバージョンをアップデートする必要はありません。 複数のLoRA / LoHAを使用する場合は、必ずLoRA Block Weightプラグインと併用して、複数のLoRA間の相互干渉を減らしてください。

1.0 Introduction

Just a merged checkpoint that I experimented and found pretty great.

I merged a lot of LoRA and checkpoint. Some I cannot recall anymore. But below are those I do remember:

LoRAs: chicVibe, atomicHips, beautifulDetailedEyes, HiPoly 3D Model, Beautifuleyeslikeness, epi_noiseOffset,

Checkpoints: chillout_ as base, merged body shape from luckyStrike, color palette from beenYou.

Good at: generating asian face with beautiful body curve, using prompts about hip size, skin details, eyes and pupils. Without face LoRA (example image 1-4) you can still get pretty asian faces, but you can use other face to blend in their features.

美しいボディカーブを持つアジア人の顔を生成することが得意で、ヒップサイズ、肌の詳細、目と瞳に関するプロンプトを使用します。顔のLoRAがなくても、美しいアジア人の顔を生成できますが、他の顔を使用して特徴を混ぜることもできます。

擅长生成带有漂亮身材曲线的亚洲人脸,使用臀部尺寸、肌肤细节、眼睛和瞳孔的提示。即使没有面部LoRA,仍然可以生成漂亮的亚洲人脸(图1-图4),但可以使用其他脸部特征混合它们的特征。

Random tips:

If you generate upper body portraits, you don’t even need to upscale and the face and all will come out just fine. But for full body it often ruin the face, I guess some LoRA just does that, img2img upscale will easily fix that.

Be sure to check out these extensions:

MultiDiffusion and Tiled VAE

LLUL

Anti Burn

Lora Block Weight

NSFW capability not tested. Ok, I’ve tested and it did not work very well, so just a heads up if you want serious NSFW content this checkpoint might not suit you.

查看更多

© 版权声明

相关文章