🎨 LoRA Name: Sand Sculpture Style
🍵 Support my Work: Buy Me A Coffee 🥰
🔑 Trigger Phrase: In a sand sculpture style
⚙️ Suggested LoRA weight: 0.75 - 0.85 range
📜 Description: this Lora converts any prompt and any scene into a realistic sand sculpture. It works best with simple straightforward natural language prompts where you leave out color and texture details and only describe the scene in broad strokes.
🔍 Sample Prompts:
In a sand sculpture style, A lone hiker reaching the peak of a mountain, taking in the panoramic view. <lora:Sand-Sculpture-Style-SCG:0.85>
In a sand sculpture style, A group of friends racing each other on bicycles through a forest path. <lora:Sand-Sculpture-Style-SCG:0.85>
🚧 Known Issues:
The lora leans towards warm output. It really wants to put a sunny beach sky over everything. I don't think it's bad necessarily, but if you want cooler colors, you're going to need to lower color temps in post.
discoloration can appear around the edges and in dark colors, especially at higher lora weights
details can get sloppy, though overall coherence is actually a boost for base model. C'mon, you're working with sand, expect some chunkiness!
💡 Tips & Tricks:
Avoid describing textures and colors. If you describe stuff as "wooden" you'll likely get wood mixed with your sand. "A pirate walking the plank" versus "A pirate with red hair and a green jacket walking the wooden plank under a murderous red flag" will net you very different results with this lora, though not necessarily bad 😉
This is very controlnet friendly! I've included a few classic movie controlnet examples in the gallery below.
❤️ Like what you see? Please give it a heart and like the model.
📸 Share Your Creations: Your feedback helps me create better future versions!
🔗 Stay Connected:
🌐 Discord: Learn from and chat with me and other professional artists and creators on the SCG-Playground community!
📷 Instagram: Catch my latest creations.
🧵 Threads: Like X, but far less Nazis!
💬 Chat on Discord: Reach out directly.
Description
This is my first LORA trained with Civit using Buzz! Got to say, it's pretty easy to use ✌🏼
This is launch version 1.0. It was trained on 24 high quality synthesized images of subjects, locations and animals made of sand, then captioned using GPT-V (Find the instructions on my Discord Server). This is also my first training using DALLe3 output for training images. 🥳
FAQ
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.