This model has received a major update with a newly trained Z-Image Turbo (ZIT) version of Egwene alVere, built from the ground up using a refined dataset and improved captioning strategy.
While earlier FLUX and SDXL versions are still available, the Z-Image Turbo version is now the recommended and featured release.
Stronger character identity
Egwene’s face, expression, and overall presence are significantly more consistent across lighting, angles, and motion.
Natural defaults learned from the dataset
The model now reliably chooses appropriate hair and clothing without needing strict prompts, demonstrating that the character identity itself has been learned — not just caption obedience.
Much better handling of movement and action
Poses like walking, gesturing, riding, working, or casting simple spells retain identity and realism without stiffness.
Improved outfit flexibility
The ZIT version adapts cleanly to both canonical fantasy clothing and non-canonical modern outfits, while still preserving Egwene’s recognizable features. The single braid appears less often if the prompt is less "Two Rivers" and "Wheel of Time". Its always useful either way to throw the single braid soft trigger into your prompt.
Cleaner generalization
Works well with minimal prompts, narrative prompts, and cinematic scenes — not just tag-style inputs.
Best version: Z-Image Turbo (~1000 steps)
Trigger / name: Egwene alVere
(Used consistently in captions; no additional hard trigger required)
Soft, optional phrases users may find helpful:
her hair in a single tight braid
wearing a wool dress with a wool belt
These are not required, but gently guide results when desired.
The original FLUX and SDXL versions were early iterations and remain available for reference or experimentation.
The ZIT version reflects significant improvements in training approach, dataset balance, and refinement, and should be considered the definitive version going forward.
If you’ve used an earlier version of this model before, the Z-Image Turbo update represents a substantial step forward in quality, flexibility, and character fidelity.
👉 For best results, use the ZIT version.
👉 Its possible there will be a new FLUX version made from this dataset.
-This LoRA is not intended to resemble any real person, living or deceased.
No real photographs or celebrity likenesses were used in the creation of this model.
This character was created solely through prompted image generation, using a combination of:
Descriptions found in Robert Jordan’s The Wheel of Time series
Narrative cues and imagery from official book text
Inspiration from classic book cover art (such as Darrell K. Sweet and Michael Whelan)
Generic medieval visual language
The training dataset was generated entirely using AI image tools, guided by natural language prompts referencing Egwene al'Vere, a fictional character from a fantasy novel.
This project is an act of fan expression, intended to reflect the book-accurate interpretation of the character—not any real-world likeness.
It is shared here for non-commercial, artistic, and transformative purposes under fair use.
Improved dataset: Earlier versions were my learning phase
1. Dieses Modell dient nur Lernzwecken. Urheber- und Auslegungsrechte liegen beim Originalautor.
2. Bist du der Originalautor eines Modells, kontaktiere uns bitte zur Authentifizierung über unsere offiziellen Kanäle. Wir schützen die Rechte aller Schöpfer. Hier klicken, um es zu verifizieren.
