BEWBS is a breast-size Concept Slider for SDXL.
Negative = smaller. Positive = larger.
The goal of this version was not simply to make a stronger size slider, but to make the change substantially more local to breast morphology. Earlier experiments tended to recruit correlated features such as overall body mass and body shape as breast size changed.
This version was trained with deliberately stripped-down pole concepts, breasts as the Target Class, and body type as the Anchor Class. In testing, that combination substantially reduced the whole-body migration seen in earlier versions.
The result behaves more like a relative morphological direction than an absolute size selector. A thin, athletic, average-build, or heavyset adult subject retains much of her prompted body type while BEWBS moves breast size within that existing solution. Consequently, the same LoRA weight does not represent one fixed absolute size across every subject.
Tested range: approximately −4 to +6 before significant composition changes or other high-weight effects begin. For ordinary use, start around ±1 to ±3 and increase as needed.
Clothing matters. Plain or anatomy-conforming clothing generally allows the underlying change to remain visible, while garments such as sports bras appropriately constrain and compress the apparent range. Viewpoint matters as well: profile views reveal the change especially clearly.
Prompt instructions still participate. Explicit body type, clothing, pose, viewpoint, and other contextual information can constrain how the checkpoint realizes the requested morphology. BEWBS is therefore not a fixed geometric transform—it applies semantic pressure to the existing generation.
No trigger word required.
Suggested starting weights:
Subtle: ±1
Moderate: ±2–3
Strong: −4 / +4–6
The gallery includes fixed-seed sweeps across body types, frontal/profile views, different clothing constraints, direct anatomy tests, and deliberately contradictory prompts to show both the range and the limits of the slider.
Use responsibly. :)


















