Introducing PixeLynx, a LoRa model, meticulously trained on a dataset of 1000 images from a diverse pool of AI artists. This collection predominantly features artworks initially conceived on Midjourney, an AI platform renowned for its artistic prowess, which were then skillfully refined using Photoshop and Lightroom.
The central inquiry of my venture with PixeLynx was to explore the impact of dataset size on the quality of generated art. The objective was: to harness the power of PixeLynx to create artworks that are not just visually captivating but also boast superior compositions, echoing the distinct style of Midjourney creations.
The journey with PixeLynx yielded some fascinating results. Images produced by this model have a certain allure, distinctly setting them apart from those generated by standard SDXL models. In terms of weighting recommendations, PixeLynx demonstrates versatility. For most subjects, applying a weight between 0.3 and 0.6 enhances the visual appeal significantly. However, for more eccentric themes, like creatures, aliens, and cyborgs, a higher weight ranging from 0.8 to 1.2 adds an intriguing level of complexity and eccentricity to the artworks.
