Civitai model page(Please note that is unstable):
https://civitai.com/models/938672NoobAI-XL(v-pred 1.0) based anime model with human artist style
Fine-tuned NoobAI-XL(v-prediction) without merging another base models, and merged SPO(Step-aware Preference Optimization) LoRA
他所のベースモデルを混ぜずに調整し、SPO LoRAをマージしたNoobAI-XL(v-prediction)
NoobAI-XL(v-prediction版)に特定画風を学習させたcheckpointを複数作成しマージしたものです。「典型的AIイラスト」感を抑えつつ、鮮やかで非常に美しいイラストを生成します。
This model does not have the "AI Masterpiece" style.
Features/特徴
- Greatly improved stability and quality.
- The base style is not strong and can be restyled by prompts or LoRAs.
- This model does not include any base model other than NoobAI (v-prediction), so it has the equivalent knowledge.
- The true v-prediction model with Zero Terminal SNR
- Excels in compositional diversity.
- 安定性と品質を大幅に改善
- 素の画風は強くないので、プロンプトやLoRAによる画風変更ができます。
- このモデルはNoobAI(v-prediction版)以外のベースモデルを一切含まず、それと同等の知識があります。
- 真のv-prediction + Zero Terminal SNRなモデル
- 優れた構図の多様性
Known issue(s)
- This model will very rarely generate errored images by certain prompts.
- このモデルは特定のプロンプトでごく稀に壊れた画像を生成します。
About 1.2.4
Requirements / 動作要件
- AUTOMATIC1111 WebUI on
dev branch / devブランチ上のAUTOMATIC1111 WebUI
- Latest version of ComfyUI / 最新版のComfyUI
- Latest version of Forge or reForge / 最新版のForgeまたはreForge
Instruction for AUTOMATIC1111 / AUTOMATIC1111の導入手順
- Switch branch to
dev (Run this command in the root directory of the webui: git checkout -b dev origin/dev or use Github Desktop)
- Use the model as usual!
(日本語)
devブランチに切り替えます(次のコマンドをwebui直下で実行します: git checkout -b dev origin/dev またはGithub Desktopを使う)- 通常通りモデルを使用します。
Prompt Guidelines / プロンプト記法
Almost same as the base model/ベースモデルとおおむね同じ
Positive prompt: Works good without quality tags / 品質タグ無しでおｋ
Negative prompt: Leave it empty(negative quality tags have no effects)
Recommended Settings / 推奨設定
Steps: >=16(best value is 20-28)
Guidance Scale: 2-5(best value is 4)
Recommended Samplers
- DPM++ 2M
- Euler
- DPM++ 2M SDE
- DPM++ 2M SDE Heun
other samplers are not tested/good.
Recommended Noise Schedulers
Hires.fix
Hires upscaler: 4x-UltraSharp, Lanczos or Latent
Denoising strength: 0.4-0.5(>0.55 for Latent)
Training scripts
Notice
This model is licensed under
Fair AI Public License 1.0-SDIf you make modify this model, you must share both your changes and the original license.
You are prohibited from monetizing any close-sourced fine-tuned / merged model, which disallows the public from accessing the model's source code / weights and its usages.
Diffusers format is not available, please convert it yourself.
