A single Lora Locon to allow prompting of mini giantess size people that appear 6-30+ feet tall. The goal is to teach models to tweak the scale of the background and provide full body poses.Flux 2 Klein 9b:<ul><li>Trained only using mesogts and macrogts as separate concepts, macrogts is maybe undertrained. </li></ul>Z Image Turbo:<ul><li>Trained only macrogts and mesogts as separate concepts, macrogts is maybe undertrained.</li></ul>Flux 1.D:<ul><li>Similar training to PDXL, trained all trigger words as separate concepts.</li></ul>Illustrious XL:<ul><li>Trained similarly to Pony, possibly undertrained, not certain if it will actually produce better results than the SDXL base lora, but worth a shot.</li></ul>Pony LyCoris:<ul><li>Re-trained on PDXL v6, see version details, as enabled flip augmentation which was missed during SDXL training.</li><li>AIO Concept, trained as a single concept separated by tagging, prompts easier with more unpredictable sizes.</li><li>Separate Concepts, separated sizes into their own concepts, more consistent sizes but may not trigger at as low strengths.</li></ul>SDXL LyCoris What's New:·      Re-trained as a SDXL LyCoris using kohya_ss instead of dreambooth.·      Tested on SDXL base, <a target="_blank" rel="ugc" href="https://civarchive.com/models/112902/dreamshaper-xl10">Dreamshaper XL</a> and <a target="_blank" rel="ugc" href="https://civarchive.com/models/125703/protovision-xl-high-fidelity-3d-photorealism-anime-hyperrealism-no-refiner-needed">Protovision XL</a>·      Slightly suspicious of Kohya resizing to rank 64, so the much larger original trained version is included although all prompts are with the resized v1.·      For older versions and more details, checkout the <a target="_blank" rel="ugc" href="https://civarchive.com/models/9214">pre-SDXL version</a>.Trigger words are tagged to better prompt approximate scaling using mg-sg guide:<ul><li>tallgts: Taller than average</li><li>minigiantess: Giantesses 7' to 12' tall</li><li>mesogts: Giantesses 12'to 50' tall</li><li>macrogts: Giantesses 50' to 500' tall</li></ul>Also training data is tagged by subject:<ul><li>solo: one subject</li><li>couple: two subjects</li><li>group: one minigts subject and a group of people</li></ul>

Minigiantess (ZIT/FLUX/Pony/SDXL)

<ul><li>Expanded to 1200 images, now roughly 50/50 between giant subjects and 'can stand up indoors'.</li><li>Re-trained as a Pony Lycoris.</li><li>AIO = All-In-One concept, uses the same tagging and additionally BLIP2/WD14. The result is it prompts more diverse/easily but more unpredictable subject size and pose.</li></ul>

Files with SHA256: b0f4bc8f017576c2c388d8acb8d9c8c121a3dbcd3657866b9e1536b13000ad0c