This LoRA is designed to improve lying poses and all-fours poses, focusing on anatomy stability, pose realism, and focus on hips.<h2 id="supported-poses">Supported Poses</h2><ul><li>Lying poses (and variations)</li><li>All-fours poses (and variations)</li></ul><h2 id="benefits">Benefits</h2><ul><li>Reduced anatomy issues (arms, legs, hips, spine)</li><li>More natural body proportions and posture</li><li>Improved pose consistency across generations</li></ul><h2 id="versions">Versions</h2><ul><li>Base version (for comfy) &amp; MLX Version (Optimized for apple silicon, can be used with MFLUX)</li></ul><h2 id="usage-notes">Usage Notes ⚠️</h2><ul><li>This LoRA is slightly overtrained, therefore it is recommended to use a lower LoRA strength than usual.</li><li>Best results are achieved when the base image is a close-up (selfie or face-focused image).</li><li>When applying this LoRA to full-body images, it is strongly recommended to:<ul><li>significantly reduce LoRA strength</li><li>use a very detailed prompt to guide pose, anatomy, and proportions</li></ul></li><li>Excessive LoRA strength may increase the likelihood of anatomy artifacts, especially in full-body generations.</li></ul><h2 id="comparison-examples">Comparison Examples</h2><edge-media url="2a58c954-0496-4923-bb76-449439155825" type="image" filename="image.png"></edge-media><edge-media url="794bb106-23f6-4766-b6fd-8c306a2a737e" type="image" filename="image.png"></edge-media>

Files with SHA256: c9aada0a32253cb60b3b40ad31535ebfbb5a7e9b29e68958d07403058a210950