Trained on a collection of the front covers of 19th-Century Persian manuscripts, this model delivers rich colours, intricate geometric patterns, and bold floral motifs. For best results, use a high-contrast base image and ControlNet, and a LoRA strength between 0.6 and 0.8.