If you liked this LORA, follow me on <a target="_blank" rel="ugc" href="https://twitter.com/lewdslotta">Twitter</a> and <a target="_blank" rel="ugc" href="https://www.pixiv.net/en/users/81897676">Pixiv</a> (to see latest in LORA news and some cool generations), or consider subscribing on <a target="_blank" rel="ugc" href="https://www.patreon.com/lottalewds">Patreon</a>.The old SD1.5 versions should not be used.The new Illustrious version was trained with <code>swoswalpo</code> as the trigger word, but it is strong enough that it seems this is unnecessary.The generations you see were at 0.5 str.Helpful prompt tags: <code>saliva</code> <code>deepthroat</code> <code>irrumatio</code> <code>kneeling</code> <code>squatting</code>

LottaLewds' sword swallowing position (deepthroat)

These gens were made with ControlNet, and seems to work really well. Try playing with Guidance End between 0.25 and 0.5 (stops guiding the pic after 25%~50%, as I understand it). You might try latent coupling with top-half being lora, lower half being generic girl kneeling prompt.

Files with SHA256: d2a8fe6123d765df28479dd92415ba7f17a611d9be666a9deaa65f863755e668