✨ Uma from The God Of High School
— She is a powerful warrior with a personality that's just as bold as her appearance. Despite her intimidating presence and fearsome strength, she's surprisingly playful, mischievous, and sometimes downright goofy. But when things get serious, her fierce loyalty and determination remind everyone why she's a force to be reckoned with.
>NO MAN'S LAND COLLECTION<
⚠️Turn off your X/XXX Filter to see hidden JUICY images and Loras
🖼️ How I Generate My Images
I made a full article that explains how I generate my images — including my settings, prompting tips, and even my potato PC setup 😅
👉 Check it out here if you’re curious or want to try the same method.
🔞 NSFW gallery (18+)
Check my account here to see the spicy showcase that I can't show here, if any.
🎨 Prompt Tips
These are optional starting points for handling a few behaviors of this version. You may still need some retries depending on the checkpoint, prompt, and LoRA strength.
👁️ Eyes
For her intended eye appearance:
✅ Positive prompt:black_eyes, white_pupils
The original manhwa is inconsistent with her eye details, so results outside this combination may vary.
🐂 Horn Control
Uma's horns are intentionally separated into controllable concepts.
Normal horns
✅ Positive prompt:black_horns, multiple_horns, cow_horns, demon_horns
This gives her the larger cow horns together with the smaller demon horns.
Broken horn
✅ Positive prompt:black_horns, multiple_horns, cow_horns, demon_horns, (broken_horn:1.2)
You can also add:
✅ Positive prompt:asymmetrical_horns
if the result keeps making both sides too similar.
Long horns
✅ Positive prompt:black_horns, multiple_horns, cow_horns, demon_horns, long_horns
This is useful when you want to push the larger horn variation more strongly.
Cow horns only
This one behaves a bit backwards.
✅ Positive prompt:black_horns, demon_horns
🚫 Negative prompt:cow_horns
For some reason, this gives me a better chance of getting the cow-horn-only appearance than directly asking for it. Yes, I know. Don't ask me why the cow refuses to listen normally.
Demon horns only
Not recommended.
The dataset does not contain enough examples of this variation, so the success rate is extremely low. You can experiment with the inverse-prompting trick used for the cow-horn-only version, but I would treat any successful result as luck rather than reliable LoRA behavior.
👗 Outfit Control
The LoRA has a noticeable bias toward Uma's default outfit.
If parts of it keep bleeding into another outfit:
Lower the LoRA strength
Put the unwanted default outfit tags in the negative prompt
Prompt the replacement clothing explicitly instead of relying only on the negative prompt
If the navel detail appears when using a more covered outfit, try adding:
🚫 Negative prompt:covered_navel
This can help suppress unwanted exposed-navel details when generating regular clothing.
✒️ Suppressing the Manhwa Outline Look
Because the dataset is fully manhwa-based, the LoRA can sometimes produce a slightly stronger dark outline style than intended.
To suppress this lightly, try adding
🚫 Negative prompt:outline, black_outline
This is not 100% effective, but it can help reduce the thicker manhwa-like outline tendency.
⚖️ LoRA strength and outfit flexibility
The LoRA has a noticeable bias toward her default outfit.
If parts of it keep bleeding into another outfit:
Lower the LoRA strength.
Put the unwanted default outfit tags in the negative prompt.
Prompt the replacement clothing explicitly instead of relying only on the negative prompt.
Lowering the LoRA strength too far can also weaken her character identity, so I recommend adjusting it gradually rather than immediately dropping it by a large amount.
📊 Lora Project
If you want to know what characters that I'm covering next you can check these link :
👉 Personal Project
👉 Commision Project
If this model sparked something in you, show some love with a like, review, or buzz.
Now go—
Be wild
Description
📚 Manhwa only reference
This LoRA was trained entirely using images from the manhwa.
Because this LoRA was trained entirely on manhwa images, some generations may still carry a slight manhwa-style outline bias, especially thicker dark outlines in certain results. This can be suppressed to some extent during generation, although not perfectly. I explain the practical workaround in the Prompt Tips.
🎨 Design Choices & Source Inconsistencies
Some parts of Uma's design are surprisingly inconsistent throughout the source, so I had to settle on a few details instead of trying to reproduce every variation at once.
- Armwear
Her armwear sometimes looks closer to bridal gauntlets, while other panels make them look more like detached sleeves. I decided to standardize them as detached sleeves for this LoRA.
- Loincloth
The bottom part of her default outfit also changes slightly between panels. Because of this, details such as the knot and side slit may appear somewhat randomly during generation.
- Eyes
Her eye details are especially inconsistent in the manhwa. The most common appearance I found was black eyes with white pupils, so I treated that as her main eye design.🐂 Horn Variations
The dataset contains enough variation for the LoRA to handle several different horn states, including:
- Her normal horn arrangement
- A broken-horn variation
- Longer horns
- Cow horns without the smaller demon horns
The last one behaves a little strangely and needs an unusual prompting method, which I explain in the Prompt Tips below.
Generating only the smaller demon horns is unfortunately not reliable. There simply aren't enough suitable examples in the dataset for the LoRA to learn that state properly.👗 Default Outfit Focus
This LoRA is primarily focused on reproducing Uma's default outfit.
Outfit bleeding can happen when trying completely different clothing, especially at higher LoRA strengths. It is still possible to change her outfit, but lowering the LoRA strength and placing unwanted default outfit elements in the negative prompt usually gives better control.As usual, the result can still depend on your checkpoint, prompt, LoRA strength, CFG scale, and generation settings.