AnidLora
詳細
ファイルをダウンロード (2)
モデル説明
Just a simple mix of AniDosMix_A and AnyLoRA. Really liked the style of animix but also it had some issues when tried to do nsfw or while using loras'so here we are.
Things I've noticed (might not be true) with Lora's:
-Euler A usually gives most consistent results if using low CFG scale steps should also keep steps low
CFG 3= steps 20-25
CFG 5.5= steps 20-40
DDIM and DPM++ 2M Karras Could often use few more steps than Euler A +(5-10)
Solid basic negative prompt: (worst quality:1.6), (low quality:1.6), (normal quality:1.6), paintings, sketches, (monochrome), (grayscale), animal ears, out of frame, lipstick,