Z-img Turbo 2.1 Dataset Version:
Same ZiT process using the smaller breast size 1.2 dataset.

Z-img Turbo XXL Version:
This is my first ZIT LoRA so hopefully there aren't any glaring issues but let me know if you notice anything. 0.75 was the approx. strength that I landed on but feel free to experiment!

XXL Version:
Added to, subtracted from, and manually modified the existing v1.2 dataset for that bigger impact. Purely a preference version compared to v1.2 if you like em big or really big. The closer you get to 1.0 strength the more likely alpha shapes will appear and the head will be out of frame. A strength of 0.75 seemed to be a good spot, but if you're ok with wading through some garbo gens, 1.0 will give you the biggest knockers.

v1.2:
Changed up the dataset to include more high resolution images with more swoop!

v1.2 SDXL:
Strengths near 1.0 will center your gens on the breasts, almost always cutting the head out of frame. Good for inpainting/detailing, bad for original gens. 0.6 to 0.75 is a good range to bring that face back into frame. Adding things like "eye contact" help a lot too. This version also responds very well to small/medium/large swooping breasts prompts ❤️

This is my second tit model and in this one, I tried to recreate the exceedingly pleasant slope that Swooping Breasts/Snoopy Breasts/Ski Slope Breasts/etc. have and their oft included slightly puffy nipples.

Three versions:
Mixed Slope (Formerly V1.0): Both datasets, may give you steep or shallow or in-between
Steep Slope
Shallow Slope (45:55 Ratio)

Original release description:

I can already tell you that there will be at least 2 more versions of this LoRA released. This is because, when I created this one's dataset, I did so with two distinct groups of tits. One group has a steeper, more dramatic slope to them. The other group has a shallower slope that is more in line with the 45:55 Golden Tiddy Ratio. I was hoping that, with different tagging for each group, this one LoRA could produce one or the other depending on your prompt. Now, I knew tags were less important with Flux, but I think I underestimated just how irrelevant they are. Live and learn I guess.

Therefore! This version is a bit of a grab bag at times, but mostly produces a middle ground between the two slope amounts. The next versions will each use their own of the two aforementioned datasets, resulting in a Steep Slope version and a Shallow Slope version.

Enjoy! And post your gens, it's helpful for future refinement!