PurposeThis model will generate a photo of a woman tit fucking. I tried to fix the following issues I saw with Chroma.<ol><li>Breasts size was mostly overly large regardless of whether you asked for large or not. </li><li>Body type defaulted to a bit heavy regardless of what body type you mentions. </li><li>Far too many images had the women's mouth open often in weird expression.</li></ol>PromptingThis model was trained on descriptive prompts generated by JoyCaption Beta 1, with manual edits for accuracy and reducing wordiness. So descriptive prompts are best.Trigger words:titfucking This is the main trigger.small breasts, medium-sized breasts, large breasts Will size the breasts more appropriatelycloseup Focuses on the woman's face and breastsWoman's face is not visible Combined with closeup zooms in on the woman's breasts.Thin Has a higher chance of a thinner face and body, especially if breast size is small or medium-sized.Strength:The LoRA is designed to be used at 1.0 strength. But seems to produce good images between .8 and .2.I haven't thoroughly tested it out yet with other LoRA's but the few prompts I created seem to work well with concept LoRA's and nude character LoRA's with a strength of .4 for this LoRA and .75 for the Character / Concept Lora.I'll post examples later once I test more.

Chroma Titfuck

PurposeGenerate a realistice image of a woman tit fucking while also fixing the following issues I saw with Chroma<ol><li>Breasts size was mostly overly large regardless of whether you asked for large or not.</li><li>Body type defaulted to a bit heavy regardless of what body type you mentioned.</li><li>Far too many images had the women's mouth open often in weird expression.</li><li>Body position often off.</li></ol>PromptingThis model was trained on descriptive prompts generated by JoyCaption Beta 1, with manual edits for accuracy and reducing wordiness. So descriptive prompts are best.Trigger wordstitfucking main word.small breasts, medium-sized breasts produces more appropriate sized breastscloseup Focuses on the woman's face and breasts.woman's face not visible Adding this with closeup zooms in on the woman's breasts.thin Has a higher chance of a thinner face and body, especially if breast size is small or medium-sized. Thin with Large is a bit more inconsistent.StrengthThe LoRA is designed to be used at 1.0 strength. But gives good results between 0.7 and 1.4.I haven't thoroughly tested it out yet with other LoRA's but the few prompts I created seem to work well with concept LoRA's and nude character LoRA's with a strength of .4 for this LoRA and .75 for the Character / Concept Lora.I'll post examples later once I test more.FailuresTrying to get aside image of the woman's body and legs as she kneeled in portrait mode occasionally seemed to focus on closeups and chest up. This could have been overtraining or bad descriptions.I may train a second one for more kneeling from the side and waist up.I left a few examples of bad hands and the woman's breast size still showing up huge in a few to show a bit some of the failures.

Files with SHA256: 378ff22c657a81a84463849f5420ae945b967bef4084a48a5b0dcfd728077465