Leshy (Inscryption)

🌟 What does your model do?

This LoRA generates stunning depictions of Leshy, the enigmatic character from Inscryption. It captures his eerie, surreal appearance while staying faithful to the horror aesthetic. Ideal for creating dark, atmospheric, or surreal artwork. 🦌🌌

⚙️ How can I use your model?

Example Prompt:

• Positive Tags:
  score_9, score_8, score_7, score_6_up, Leshy, solo, facial hair, mushroom beard, green hair, green eyes, glowing eyes, spiral pupils, colored skin, green skin, hair branches, leaf, sky, vines, full body, fur legs, hooves, (horror, creepy), skinny
BREAK
night, trees, full white moon, hyper detailed, dark atmosphere, (forest:1.2), surreal landscape

• Negative Tags:
  score_4, score_3, score_2, cute, beauty, ((glowing)), signature, artist name, watermark, simple background

Detailing Tip:

If you're using a face or eye detailer, it's highly recommended to include the following tags:
green eyes, glowing eyes, spiral pupils
This ensures consistency with Leshy's iconic features and enhances his eerie expression. Remove glowing eyes from the main prompt if you use this too, will help to have less glowing area on the main image.

Guide to Usage:

• Use Leshy-specific tags like mushroom beard, green hair, green eyes, glowing eyes for accurate character rendering.

• Add background/style tags after the BREAK for dynamic and atmospheric scenery.

• If hooves aren’t rendered with fur legs alone, add the tag hooves explicitly. However, include hands in the prompt to avoid hooves being mistakenly applied to the character's hands.

• Adjust LoRA strength between 0.7 and 1.0 for the best results.

🔑 Recommended Tags

• Character Details:
  Leshy, green hair, long hair|short hair, (glowing eyes:0.8), spiral pupils, fur legs, hooves, mushroom beard

• Backgrounds and Styles:
  dark forest, woodcabin, outdoors|indoors, surreal landscape, full {insert color} moon, dark atmosphere, night|day

• Horror Aesthetics:
  horror, creepy, hyper detailed

🌟 What is your model good at?

• Faithful Representation: The LoRA accurately portrays Leshy, with a strong resemblance to the original character.

• Customizable: Works well for full-body illustrations, portraits, or close-ups. Leshy adapts nicely to props such as a staff, camera, or other specific objects.

• Atmosphere: Excels at generating horror-themed or surreal compositions with dark, moody settings.

💡 Note: The dataset had limited visual references for Leshy's legs, making consistent designs for the lower body a creative challenge.

⚠️ What is your resource bad at?

• Artifacts: Small white spots may sometimes appear on the image.

• Unpredictable “Holding” Tag: While holding works for props like cameras or staffs, for other objects, use specific prompts like card in hand or hands at gamepad.

• Inconsistent "Mushroom Beard": This tag usualy creates mushrooms in unexpected places.

• Background Weakness: The model struggles with complex backgrounds unless explicitly described after the BREAK.

• Green Theme: Sometimes everything is very greenish, ensure you add colors.

🎨 What should it be used for?

• Horror Art: Perfect for eerie and atmospheric compositions.

• Dynamic Poses and Props: Leshy can be depicted holding specific objects or adapting to various clothing and poses.

• Surreal Settings: Combine Leshy with vivid forest, full {insert your color} moons, or descriptive backgrounds for dramatic impact.

🚫 How should it not be used?

• Avoid Overuse of Embeddings: Many embeddings can dilute the LoRA’s core character essence. Start without embeddings and experiment by adding them gradually.

• Hyper Detailing: Including hyper detailed in the prompt is often unnecessary, as the LoRA already achieves impressive detail.

• Expressions: Occasionally, the character’s face may appear overly comedic or sad. Tags like (serious face) or (horror face) can help refine the expression (i didn't use them on examples).

• Dont use upper body use portrait for that or lower parts will disappear

📂 Additional Resources

🔗 A JSON file with all training tags and their frequency is provided. Use these tags flexibly in both positive and negative prompts to fine-tune results.