<h1 id="oversized-load">Oversized Load</h1><h3 id="engineered-for-when-the-standard-chassis-simply-isn't-rated-for-the-payload.-for-when-you-want-women-with-dramatically-wide-hips-oversized-rounded-glutes-strong-rear-projection-and-thick-proportional-thighs."> Engineered for when the standard chassis simply isn’t rated for the payload. For when you want women with dramatically wide hips, oversized rounded glutes, strong rear projection, and thick proportional thighs.</h3> A body-shape LoRA focused on exaggerated lower-body proportions while minimizing unwanted changes to faces, hair, clothing, backgrounds, and unrelated anatomy. It supports clothed, nude, neutral, and explicit adult scenes. Example prompt:<pre><code>
pawg, photorealistic full-body photograph of a confident adult woman, viewed from a natural rear three-quarter angle. She has a narrow waist, broad hips, exceptionally full rounded glutes with strong backward projection, and thick proportional thighs. Relaxed over-the-shoulder pose, one leg slightly forward, realistic weight distribution, anatomically coherent hips and legs, natural skin texture, subtle imperfections, authentic smartphone photography, gentle depth of field, tasteful premium creator aesthetic</code></pre> Lower strengths produce subtler proportions. Higher strengths increase width and projection but may exaggerate anatomy.<h3 id="prompting-tips">Prompting Tips</h3> For stronger results, add:<pre><code>
wide hips

rounded glutes

strong rear projection

thick proportional thighs

rear view or side-rear view</code></pre> For better control, describe the waist, thighs, clothing, pose, and camera angle explicitly. Works best from rear and side-rear angles. Front views may show a subtler effect.

Oversized Load — MOSS/Krea 2 PAWG LoRA

Here's MOSS for all of you so I'm not ancient and only running Krea. This was trained on Moss Base, however I prefer it with SFTI have not tested this extensively yet, I wanted to get it out to you all asap, so please if you have any feedback let me know!My results:<ul><li>I had the best results using 0.7 strength. going up to 1 gave me bad limbs. I'll update this the more I experiment. </li><li>Use a Uncond skip (SLG) node for better results. I used 22 single layers and a 0.5 end percent.</li></ul>

Files with SHA256: c8b98894117734e09b4824179fa73c5cbb10d637854bae858e9feeae7b88d1ae