QuillworksIllustrious

While still Quillworks at its core, this newly checkpointed model represents an exciting evolution in both form and flair. At the foundation remains the flexible, expressive engine that Quillworks has become known for—versatile across genres, styles, and character designs. But this time, it’s been infused with a carefully curated selection from Hailoknight’s legendary dataset, known for its vivid stylization, punchy contrast, and character-centric nuance. The result is a synthesis of vision: Quillworks’ dynamic adaptability meets the iconic artistic fingerprint of Hailoknight’s earlier LoRAs.

This collaboration brings forward a new chapter. It’s more than a merge; it’s a refinement, a tuning of sensibilities that adds richness without overwhelming the original character. Expect a slightly more dramatic lighting curve, enhanced brushstroke emulation, and that subtle but unmistakable Hailoknight tone that makes eyes pop and forms read like stylized illustrations straight from a graphic novel.

Whether you’re a long-time Quillworks user looking for a bolder aesthetic or a fan of Hailoknight’s visual storytelling, this checkpoint aims to offer the best of both worlds. It was trained with care and intention, focusing on preserving coherence while enhancing stylization, so your existing prompts won’t break—but they might just shine brighter than ever.

This is a merge of Quillworks with a lora trained on a dataset sent by Hailoknight and WaiNSFW with my style Lora applied. The LoRA runs best at 5-6 CFG and using Euler A, Simple on Forge or Comfyui. It runs on standard Booru-style tagging and as a base illustrious model it runs quite quickly and even on lower end machines.

I like to run hires Lancoz at 1.3 times upscale for many images.

This Model is still based on the old quillworks dataset so its still got the same errors and issues as found in the base model. Hands often have an extra finger or a missing finger. Eyes sometimes needs adjustment. Certain styles cause dramatic shifts in output.