SSLQ-StoneOcean-LoRA-v1

SSLQ-StoneOcean-LoRA-v1

<Gallery />

Model description

SSLQ-StoneOcean-LoRA-v1 is a style LoRA trained on Stable Diffusion XL 1.0. It generates environments inspired by the Stone Ocean aesthetic from the SSLQ dataset which are towering stone columns, cavern temples, ocean-eroded structures, and teal atmospheric lighting.

This LoRA is designed to transform ordinary landscape prompts into the unique Stone Ocean look. The SSLQ dataset includes caverns and pillar formations with a blend of natural erosion and geometric, almost architectural shapes. These structures are difficult to prompt reliably using a base model alone, making a LoRA ideal for capturing the style.

To activate this LoRA, include either of the following keywords in your prompt:

• sslq

• stone ocean

Different CLIP and model strengths will shift the stylization from subtle texture changes to full Stone Ocean transformation. Currently, the v1 model is overfitted. It can still give results, but they won't be varied among differnet seeds. Examples below show how CFG and CLIP strength influence the results to counter the overfitting. Future versions are going to be designed for less overfitting.

Recommended prompt

Prompt

stone ocean with rays of sunshine coming through the top

Negative prompt

(bad anatomy), text, error, missing fingers, extra digit, cropped,
signature, watermark, username, deformed face, water, ocean, waves

SDXL base 1.0 (no LoRA)

SDXL base 1.0 + StoneOcean LoRA (strength_model .50, strength_clip .25)

(strength_model .50, strength_clip .85)

(strength_model .95, strength_clip .85)

Additional prompt example

Prompt

stone ocean rays of sunshine coming through the top, deep inside cavern green and blue stone

Negative Prompt

(bad anatomy), text, error, missing fingers, extra digit, cropped,
signature, watermark, username, deformed face, water, ocean, waves

(strength_model .80, strength_clip .50)

Random seed exploration (sslq only) Prompt

sslq

No Negative Prompt

(strength_model .80, strength_clip .50)

(strength_model .80, strength_clip .50)

(strength_model .80, strength_clip .50)

Trigger words

Use one or both of the following to activate the Stone Ocean style:

• sslq

• stone ocean

Download model

Download them in the Files & versions tab.

Version notes

v1.0 — Initial functional release

Trained via kohya_ss over Runpod on an A6000

This is the first fully working version of the Stone Ocean LoRA.

Planned improvements for future versions:

• Less overfitting
• Better textures if possible
• Expanded dataset for multi-biome Stone Ocean variations (temple zones, flooded caverns, open-ocean monoliths)

Future releases will be labeled v1.1, v2, etc., depending on the scale of changes.