Thank you for many using. We look forward to your fascinating works of art.SDXL用のオリジナルマージモデルです。推奨作品：イラストレーション全般、女性の描写に適していて、光沢や反射の表現力が高いですが、アニメにはやや不向きかもしれません。LoLAの使用によりリアリスティック作品の生成にも対応しやすくなります。推奨パラメーター：サンプリング法：DPM++ 2M Karras、サンプリングステップ：30～38、CFGスケール：7.5～9.5Recommended values:Sampling method: DPM++ 2M KarrasSampling step: 30-40CFG scale: 7.5-9.5Clip skip: 1-2Note:This model can generate high-quality works with few generation outputs.However, it has a strong tendency for the depiction to collapse when continuous generation is continued.For better results, try generating a completely different prompt while generating with the intended generation prompt. Alternatively, it may be possible to delay the deterioration of the model's quality by changing the tokens of the prompt each time it is generated.Also, a slight difference in the arrangement of the syntax may give unexpected results... If you do not get the results you want, try generating with the minimum number of tokens and gradually increasing the elements.Although it is free for personal use, please note that it may be difficult to obtain the same quality as when generated with Seaart when generating with Stable Diffusion in a local environment.Unauthorized reproduction of the model itself is strictly prohibited, so please use it with good manners and morals.
Thank you for many using. We look forward to your fascinating works of art.
SDXL用のオリジナルマージモデルです。
推奨作品：イラストレーション全般、女性の描写に適していて、光沢や反射の表現力が高いですが、アニメにはやや不向きかもしれません。
LoRAの使用によりリアリスティック作品の生成にも対応しやすくなります。
【推奨パラメーター】
サンプリング法：DPM++ 2M Karras
サンプリングステップ：30～38
CFGスケール：7.5～9.5
【Recommended values】
Sampling method: DPM++ 2M KarrasSampling
steps: 30-40
CFG scale: 7.5-9.5
Clip skip: 1-2
Note:This model can generate high-quality works with few generation outputs.However, it has a strong tendency for the depiction to collapse when continuous generation is continued.For better results, try generating a completely different prompt while generating with the intended generation prompt. Alternatively, it may be possible to delay the deterioration of the model's quality by changing the tokens of the prompt each time it is generated.Also, a slight difference in the arrangement of the syntax may give unexpected results... If you do not get the results you want, try generating with the minimum number of tokens and gradually increasing the elements.Although it is free for personal use, please note that it may be difficult to obtain the same quality as when generated with Seaart when generating with Stable Diffusion in a local environment.Unauthorized reproduction of the model itself is strictly prohibited, so please use it with good manners and morals.
