IMZ- Qwen Portrait Pro

์„ธ๋ถ€ ์ •๋ณด

ํŒŒ์ผ ๋‹ค์šด๋กœ๋“œ (1)

๋ชจ๋ธ ์„ค๋ช…

๐ŸŽจ IMZ-Qwen Portrait Pro (LoRA)

IMZ-Qwen Portrait Pro is a specialized LoRA model designed to enhance portrait generation with superior detail, depth, and realism. Fine-tuned with 12,400 training steps across 5 epochs, it excels in producing lifelike skin textures, balanced lighting, and expressive facial featuresโ€”making it ideal for artistic portraits, character design, and high-quality editorial imagery.

โœจ Key Features

  • ๐Ÿ‘ค Portrait Specialization โ€“ Optimized for face and upper-body compositions.

  • ๐ŸŒŸ Enhanced Realism โ€“ Smooth textures, natural tones, and balanced highlights.

  • ๐Ÿ–ผ Creative Flexibility โ€“ Works well across styles while retaining realistic quality.

  • ๐Ÿงฉ Trigger Word โ€“ ImzQwenPortraitPro

โš™๏ธ Recommended Settings

  • Lora Weight: 0.6 โ†’ 0.8

  • Sampling Method: Euler

  • Sampling Steps: 20 โ†’ 40(โญ Best: 25)

  • CFG Scale: 4 โ†’ 7 (โญ Best: 5)

  • Upscaler: 4x-UltraSharp

  • Hires Steps: 10 โ†’ 25

  • Denoising Strength: 0.3 โ†’ 0.45

โš ๏ธ Disclaimer

This model was trained on a dataset of stock and AI-generated portrait images. All data is entirely fictional and not based on any real person.

  • ๐Ÿ–ผ AI-Generated Content: All portraits are artificial, with no intentional reference to real individuals.

  • ๐Ÿšซ Non-Commercial Use Only: Provided for creative and educational purposes, not for commercial exploitation.

  • ๐Ÿ“œ Policy Compliance: By using this model, you agree to follow Tensor Artโ€™s policies. Misuseโ€”including generating explicit, harmful, or illegal contentโ€”may lead to restrictions or legal consequences.

  • โœจ No Celebrity Training Data: No celebrity likenesses or copyrighted content were included.

  • โŒ Prohibited Usage: Generating or selling celebrity likenesses or unauthorized content is strictly prohibited.

    ๐Ÿ’ฌ Feedback & Support

If you enjoy working with IMZ-Qwen Portrait Pro, please share your feedback. Your support helps me improve future LoRA releases and keep building better models!

์ด ๋ชจ๋ธ๋กœ ๋งŒ๋“  ์ด๋ฏธ์ง€