Reliably produces what appears to be a linear change in areola size in the expected direction (larger negative weights -&gt; smaller areola, larger positive weights -&gt; larger areola) within a range of values (and the range depends on other factors; one such factor is breast size):*In testing “medium sized breasts”: a clear linear areola size increase from -1.0 to 2.0 with the difference between -1.5 and -1.0 and between 2.0 and 2.5 being a more subtle, but in limited testing, real difference. Also, despite no prompt change, breast size changed at the extremes (-1.5, 2.0, and definitely 2.5) especially. I am not claiming I’ve come close to breaking the association between breast size and areola size but I believe the LoRA at leasts bends the association in real and I believe useful ways.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/2a7fb556-9d73-47fa-8d67-ece153d37789/original=true/2a7fb556-9d73-47fa-8d67-ece153d37789.jpeg" />Instead of describing what happened when testing “small sized” and “large sized” breasts I’ll let the images mostly speak for themselves.*Small sized breasts:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/623c0d61-01df-4946-bf1d-8e2553bab7a0/original=true/623c0d61-01df-4946-bf1d-8e2553bab7a0.jpeg" />As you can see, pushing above 2.0 led to nothing good. I find the 1 to 1.5 strength to be interesting for those looking for relatively small breasts with relatively large areola.*Large sized breasts:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/36420b1d-f477-4fce-b262-4c30f0f6ada0/original=true/36420b1d-f477-4fce-b262-4c30f0f6ada0.jpeg" />The slider seems least useful with large breasts generally with values from 0 to 1.5 showing nominal changes. Its most promising use with large breasts imo is at the low end. Less than 0 weights definitely decrease areola size. Id be more enthusiastic if the nipples were a bit more prominent and the areolas a bit darker…I suspect but don’t know that prompting could bring out the nipples more but may not make the areola darker but its an empirical question Ill leave up to you to test…of course YMMV based on a myriad of factors. Have fun and I hope some of you will find this useful!

Files with SHA256: d55fee2fef871af8468399417a9dfdf9b9b01de975f24bb459a84b9b860b6444