This LoRA is primarily trained for scenes featuring two characters in a single composition. As more generations are tested and its characteristics are better understood, the information here will be refined and updated over time.To get started, try remixing prompts from the gallery, especially those you find interesting or appealing. You are also encouraged to remix and test your own works that include two or more characters. As suitable examples emerge, recommended works will be pinned for easier reference.

[Lora]DUO futa

DUO CORE is designed as an identity-layer LoRA rather than a strict recreation model. This series focuses on preserving the atmosphere, facial impression, and presence of the character while allowing the base checkpoint’s own style, lighting, and rendering to remain visible. The model is intentionally tuned for flexibility and checkpoint compatibility rather than hard overwrite behavior. DUO CORE also carries interaction-aware enhancement behavior: facial balance, eye contact, mouth separation, skin response, and multi-character harmony are softly reinforced to improve natural scene composition without forcing a fixed art style. Recommended weights: 0.3–0.45 → checkpoint-heavy / soft influence 0.5–0.6 → balanced CORE mode 0.7+ → stronger character influence

Files with SHA256: a5fe301a893009b9a635d2f6fb458453db32aafe6524a7d8b8bd665164bb6aeb