I present my attempt to make a detailed anal cross-section model (since base models are too cervix-focused). The model is consistent enough that I'm uploading it as-is. <ul><li>This model is meant to create an internal that is not just a 2D side view (different angles, more depth, etc). </li><li>I tried to train this to have minimal influence on art style but there's only so much that can be done for that. </li><li>Compared to my other model this one does have a slight difference in the flatness of shading/color and that's due to the extra detail involved on the internal view itself during the training but it's minimal. </li></ul><hr />V1: Recommended model strength is 0.6-0.7 to avoid having it overinfluence the generation when combined with additional models (LoRA and LyCORIS).Main tags:<pre><code>mp-rectal, internal view, cum inside</code></pre> I have found that lowering the tag strength on internal view (like internal view: 0.8) can help with duplicate/doubling of the cross-section elements. None of my training images had the below cum concepts but you may optionally add them to negative prompt if you want the cum to be less messy (it's an Illustrious/booru dataset bias issue, not a LoRA issue): Negative tags:<pre><code>ejaculating while penetrated, cum in ass, cum overflow</code></pre><span style="background-color:rgb(26, 27, 30);color:rgb(141, 141, 143);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">The settings used for my generation testing were 4-5 CFG, 30 steps, DPM++ 2M SDE - Karras<span style="background-color:rgb(26, 27, 30);color:rgb(141, 141, 143);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px"> on Forge<span style="background-color:rgb(26, 27, 30);color:rgb(141, 141, 143);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px"> with an adetailer<span style="background-color:rgb(26, 27, 30);color:rgb(141, 141, 143);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px"> (face_yolov8s<span style="background-color:rgb(26, 27, 30);color:rgb(141, 141, 143);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">).

(Male) Detailed Anal Cross-section - Illustrious

Recommended model strength is 0.6-0.7 to avoid having it overinfluence the generation when combined with additional models (LoRA and LyCORIS).I have found that lowering the tag strength on internal view (like internal view: 0.8) can help with duplicate/doubling of the cross-section elements.

Files with SHA256: 605c841a1c1fc8a936c4a9a5a6666377d8ac0c54418867a427cf7c354a799122