Character LoRA for Matthew Kantakouzenos — an imagined version of a historic Byzantine noble navigating the tension between monastic life, political power, and personal conviction in the 1300s.
Trained on a curated dataset of painterly, cinematic images to capture:
consistent facial identity
contemplative and serious expressions
multiple roles (monk, noble, soldier)
Byzantine-era atmosphere and lighting
The model is designed to maintain identity across varied outfits and settings while preserving a cohesive medieval aesthetic.
Trigger Word
matthew_byz
Recommended Settings
LoRA weight:
0.6 – 0.9
~0.6 → more flexibility
~0.8 → best balance
~0.9 → strongest identity
Example Prompts
matthew_byz, young Byzantine man, dark robe, contemplative expression, candlelight, medieval monastery
matthew_byz, young Byzantine noble, ornate robe, palace interior, composed expression, cinematic lighting
matthew_byz, wearing light armor, holding spear, courtyard, serious expression
matthew_byz, simple tunic, no robe, natural light, neutral background
Notes
Works best with SDXL-based models
Strong identity retention across different roles and clothing
Trained with varied lighting (candlelight, window light, interior/exterior)
Designed for cinematic, painterly outputs
Limitations
May favor medieval / Byzantine aesthetics due to training style
Strong lighting bias (warm, dramatic tones)
Extreme modern prompts may require lower LoRA weight
Description
Initial release of Matthew Kantakouzenos LoRA.
Trained on a curated dataset of Byzantine-inspired, painterly images capturing multiple roles (monk, noble, soldier) and varied lighting conditions.
This version balances identity strength and flexibility:
- consistent facial features across angles
- works across different outfits (robes, armor, tunics)
- strong performance in cinematic and medieval scenes
Best results achieved around mid-to-late training epochs.