AI Asian Influencer – Multi Model LoRA (NSFW + SFW Mix)

A high-quality, versatile LoRA designed to generate modern Asian influencer-style visuals across multiple base models. Trained on a carefully curated synthetic dataset to balance realism, aesthetics, and stylistic consistency.

✨ Supported Base Models

• Z Image Turbo

• Qwen Image

• 2512

• Z Image Base

📸 Concept

This LoRA focuses on generating:

• Asian female influencer aesthetics

• Social media-style portraits & lifestyle shots

• Clean beauty, fashion, and casual visuals

• Both SFW and tasteful NSFW outputs (non-explicit, model-friendly)

🧠 Training Details

• Steps: 3,000

• Learning Rate: 0.0001

• Dataset: Synthetic dataset generated using Nano Banana Pro

• Balanced mix of:

  • Portraits (close-up, mid-shot)

  • Lifestyle & candid poses

  • Fashion & casual outfits

  • Controlled NSFW variations (non-graphic)

🎯 Strengths

• Strong facial consistency

• Clean skin tones & lighting

• Influencer-style posing and framing

• Works well with:

  • Soft lighting

  • Studio portraits

  • Instagram-style compositions

⚙️ Recommended Settings

• Weight: 0.6 – 0.9

• Sampler: DPM++ / Euler A

• Steps: 20–30

• CFG: 5–8

🧾 Trigger Keywords (optional)

lav3nder, east asian girl, blue black mix hair

🚫 Notes

• Optimized for aesthetic output, not hyper-real anatomy accuracy

• NSFW capability is present but best results come from tasteful prompting

• Overweight (>1.0) may cause:

  • Over-smoothing

  • Facial distortion

💡 Example Prompt

lav3nder, soft natural lighting, instagram style photo, casual outfit, detailed face, shallow depth of field, high quality