'cover', 'cover page', 'magazine cover', and 'spine (book)' are key tags for this LoRA, as much of the data comes from scans of SABU magazine covers, including spines in wide-resolution images. Expect some unintelligible text, though I transcribed English text when possible and included 'english text' and 'japanese text' tags where relevant. Some grit and scan lines are inevitable despite my efforts to find quality data. If you have higher-quality sources for training, feel free to DM me :)

This LoRA is designed exclusively for Pony XL v6, trained on 103 images alongside their booru captions. The example images generated were prompted using those same captions, though you're welcome to experiment with different prompting styles. No specific source or artist names were included in the tagging process, so there's no trigger word, and it performs well even without specifying any negative prompt. For guidance on prompting and best practices, please refer to the Pony model's page.