A LoRA trained to generate doggystyle compositions with the subject facing the camera. This model was trained on a 61 photos of porn stills, it's a little rigid so I am working on an expanded datasetMain Trigger Word:<code>ffdoggy</code> (essential for activating the primary positioning logic). While it will produce results without the trigger, using<code>ffdoggy</code> is strongly recommended for consistent composition. Additionally using <code>her hips pressed against her ass</code> helps a lot to keep the man perfectly behind her if the composition is complex.Core Conceptual Keywords:<ul><li>Positioning:<code>facing camera</code>,<code>man standing behind a woman</code>,<code>ass up</code>,<code>on elbows</code>,<code>on hands</code>,<code>breasts pressed down</code>,<code>face down</code>,<code>looking down</code>,<code>man's hand on head</code></li><li>Action/Detail:<code>bouncing breasts</code>,<code>breast hanging down</code>,<code>hair pulling</code></li><li>Body/Face:<code>curvy</code>,<code>petite</code>,<code>huge breasts</code>,<code>large breasts</code>,<code>medium breasts</code>,<code>small breasts</code>,<code>closed eyes</code>,<code>open mouth</code>,<code>grimacing</code>,<code>pale skin</code>,<code>dark skin</code></li><li>Accessories:<code>pierced nipples</code>,<code>fishnet leggings</code>,<code>glasses</code>,<code>necklace</code>,<code>tattoos</code><ul><li>pierced nipples works inconsistently.</li></ul></li></ul>What it's capable of:<ul><li>Consistent Framing:<ul><li>Excels at maintaining the "woman facing camera" perspective, keeping the man behind her out of focus or partially obscured.</li><li>Key positioning terms like<code>on elbows</code> or <code>on hands</code> work reliably to define the base pose.</li></ul></li><li>Body Diversity:<ul><li>Trained on a range of body types from<code>petite</code> to<code>curvy</code>, with a good amount of data for various breast sizes, from<code>small</code> to<code>huge</code>.</li></ul></li><li>Can generate different skin tones effectively using<code>pale skin</code> and<code>dark skin</code> prompts.</li><li>Facial Expressions:<ul><li>Responds well to facial prompts like<code>closed eyes</code>,<code>open mouth</code> (for pleasure or shock), and<code>grimacing</code> to convey a wide spectrum of reactions but is not limited to those.</li></ul></li></ul>Known Quirks:<ul><li>It's sometimes to rigid in the positioning making more dynamic posing difficult to achieve</li><li>The exact strength may need to be adjusted (typically 0.7-1.0) to balance the pose with other elements in the prompt.</li></ul>

Facing Camera Doggystyle - Z Image Turbo

This version is a little rigid in posing, working on a larger expanded dataset to make it more flexible.

Files with SHA256: a267de2ff85cb381bdab40cdb9c8e5e544ffaea212abdf6c6bbd53e1606252be