That Body

This LoRA is focused exclusively on female body proportions and anatomy.

It was trained only on body data to provide consistent, natural-looking proportions across a wide range of styles, outfits, and scenarios. Faces were intentionally not trained, allowing this LoRA to remain highly compatible with most base models and face LoRAs.

The goal of this LoRA is clean structure, believable curves, and stable full-body results, whether the subject is clothed, armored, stylized, or minimally dressed.

---

✨ What This LoRA Does Well

• Natural body proportions

• Defined waist, hips, and thighs without exaggeration

• Stable anatomy in full-body shots

• Works across realistic, semi-realistic, 3D/CGI, and anime-leaning styles

• Plays well with clothing, armor, and fashion prompts

---

⚠️ What This LoRA Does Not Do

• Does not control facial features

• Does not lock in age, ethnicity, or expression

• Is not a character LoRA

This is intentional to keep it flexible and mixable.

---

🔑 Usage Notes

• Activation word: thatbody_ref

• Recommended weight: 0.7 – 0.9

• Best results with full body prompts

• Compatible with most SDXL base models and other LoRAs

---

🧪 Training Notes

• Trained across multiple epochs with controlled testing

• Same seed and prompt structure used to evaluate consistency

• Final release chosen based on body stability, proportion accuracy, and versatility

---

📎 Notes from the Creator

This LoRA is meant to be a foundation piece — something you can rely on when you want the body to “just work” without fighting the model. Mix it, stack it, push it stylistically, or keep it subtle.

If you want exaggerated or hyper-stylized results, higher weights can work — but its strength is in balance and consistency.