This was a test to see how well a multi-concept lora, based on some of my previous Lora datasets, would work on Pony and was generally successful so am releasing it for others to use. I had either 25 or 50 images for concept. Concepts are listed below, sorted from most coherent to least coherent from my test results:<ul><li><code>bangs pinned back</code> generates bangs pinned back decently well. Note that it may appear in other gens, should add to negatives if unwanted.</li><li><code>lap pillow</code>, suggested that you use tags <code>lying</code> and add <code>pillow</code> to negatives for better results, as well as using a landscape orientation as this leads to better results</li><li><code>69</code>, generate with <code>lying, cunnilingus, fellatio, licking, ass grab, lying, on back, hetero</code> and landscape orientation for best results. Can mix with yuri, futanari, or yaoi tags to get alternate concepts. May generate limbs in incorrect positions, requiring inpainting or img2img to fix.</li><li><code>horn grab</code> works sometimes, may take multiple gens to get correct or fixing via inpaint or img2img. Using <code>pov</code> generally increases success rate, as well as specifying type of horns.</li><li><code>low-tied sidelocks</code> is improved but generally only shows up on close-ups. May appear unexpectedly in other gens.</li><li><code>scales</code> and <code>scaled humanoid</code> shows signs of improvement, reducing the number of more anthro-like dragons, but not a complete success</li><li><code>fuckboi</code> was an attempt to make male characters with curly-esque hair and annoyingly smug faces. Unfortunately it didn't really work, but can be used to make characters more masculine.</li></ul>

Files with SHA256: 601f690d81621df3ecd495b791f2920266d906802de957113356cfe0d163998a