Initial identity-focused SDXL LoRA release. This version prioritizes facial consistency, realistic facial proportions, stable eye, nose, and jaw geometry, restrained expressions, and semi-realistic fantasy portrait rendering. The model was trained primarily on close-up and medium portrait compositions across varied lighting conditions and grounded fantasy environments in order to preserve identity across different contexts.
The training process emphasized realistic skin texture, natural facial features, layered academic and fantasy-inspired fashion, and strong identity cohesion without excessive stylization. The model currently performs best with realistic or semi-realistic SDXL checkpoints, portrait and medium-shot compositions, restrained styling prompts, and natural or softly cinematic lighting.
Future iterations may expand into stronger body and silhouette consistency, broader outfit and pose flexibility, more stylized fantasy rendering, and Pony-compatible variants while preserving the underlying facial identity established in this release.
