Bakunyuu Bod

CREATOR'S NOTE: This merge was specifically created for the ease of quickly creating high-quality female characters with larger bust sizes, without necessitating the extra GPU hit from adding additional LORAs needed to do so. If your complaint is that tags related to bust size produce breasts that are too large, or that you can't seem to generate characters with small enough busts, then you're barking up the wrong tree by using this checkpoint. This checkpoint was unapologetically created for breast lovers.

---

Updated: v.2 (10/14/25)

New and improved. Bigger and fleshier!

Updated: v.1+VAE (10/11/25)

Baked in the VAE to the checkpoint for users who were having issues with their renders (without having a VAE automatically load).

---

A fairly diverse checkpoint merge that handles all sorts of animated styles for creating characters. Very dynamic size and color options (venus body, tits on a stick, bbw, muscle girl, colored skin, furry, etc). Not great for images solely of landscapes/objects (you have to tag the negative section with character tags, just use something else for still life renders).

Primarily created to create curvy female characters, but might work decently for male characters (I would just go with a model designed for male characters, instead). Although this merge wasn't created with furry/anthro characters in mind, it's a fairly popular niche on here. So yes, I've tested, and it does that stuff pretty well also.

Results vary HEAVILY based on quality tags and style tags used in the positive and negative prompts.

• Will produce more realistic renders with tags such as "realistic, photorealistic, realistic skin texture" and extensive quality and lighting tagging. Add embeddings for even more polished results.

• Can produce very simple, toon-looking images when prompting with few quality tags. For more cartoonish results, try using tags such as "anime, toon \(style\)". Add the tag "painterly" for a subtle watercolor look.

• Adjust the weighting of any of these tags to enhance your desired outcome.

---

You're free to experiment with whatever settings you want (to see if you get better results). However, these are the results I typically use, you can try starting there:

Sampling Method: Euler A (SGM Uniform)

Sampling Steps: 25, CFG Scale: 4.5

Hires.fix: 4x_foolhardy_Remacri (Upscale by 1.5)

Hires Steps: 15, Hires CFG Scale: 4.5, Denoising Strength: 0.5

Adetailer: face_yolov8n

Best quality images in Portrait dimensions (896x1152, etc)

---