🎯 Config: Euler A / 8–10 Steps or DDIM 8–15 Steps
📌 Base Model (ZepMix) -> https://pixai.art/en/model/1966966196364629102
—
📌 A. Prompt Structure:
{subject|1girl|1boy|etc}, {char_name}, {any tags + clothing}, BREAK|<empty>, {quality prompt}
💎 B. Quality Prompt Example:
(masterpiece, amazing quality, absurdres)
🚫 C. Alt Negative Prompt:
scan artifacts, jpeg artifacts, sketch, (worst quality, bad quality, cropped, ai-generated:1.1), lowres, jpeg, artifacts, signature, username, text, logo, bad anatomy, artist name, artist logo, extra limbs, (extra digit, cropped), extra legs, extra legs and arms, disfigured, missing arms, too many fingers, fused fingers, missing fingers, unclear eyes, blur, film grain, noise
⚠️ Neg 2:
bad quality, worst quality, low quality, worst detail, bad anatomy, extra fingers, simple background
—
📌 Note:
🗣️ ILv2 and NX2→ supports limited natural language
🗣️ ILv3(Oriana-β)→ support limited style mixing from NoobAI
📌 Extra Note:
→ Models perform better for NSFW
→ For best initial texture Oriana variant B(alpha ver), ILv2, and NX2-A3 is better, for LoRA stability choose Oriana-β
✨ ILv2 Version is more versatile and has better initial quality but might produce oversatured image
✨ NX2 Version is the more balanced — also NX2 had some minor issue on fingers
✨ ILv3NB Version had slightly more style support and more stable LoRA usage for the f2 version, The downside is F2 version had more flat/generic default style than the Alpha version
