This model is designed to recreate the aesthetic of Kintsugi
the art of repairing objects with gold. We have intentionally moved beyond the cliché of simply showing cracks on cups or tableware (though the model handles these flawlessly) to demonstrate the full potential of this technique when applied to complex biomechanical and organic surfaces.
The test presets showcase a combination of 3D scarification, Māori-inspired patterns, and gold inlay. The model captures the physical properties of gold leaf, effectively "pouring" molten metal into the grooves of the skin’s texture, while generating authentic ambient occlusion and studio-quality highlights.
Recommended Settings and Weights
Kintsugi LoRA Weight:
0.60 (an optimal balance that preserves the underlying skin geometry while ensuring a convincing gold inlay).
Tattoo / Pattern LoRA Weight: 1.00 (to maintain a crisp, well-defined structure).
Aspect Ratio: 2:3 (recommended for portrait-oriented studio shots and high-resolution output).
Prompting for Texture:
Using terms like "scarification," "Māori patterns," and "embossed depth" will trigger a deep, 3D-inlay effect.
Workflow Details (ComfyUI Workflow Included)
Embedded within the image file is a complete workflow, featuring:
A tested, verified prompt.
Two standard LoRA loaders.
A custom API node, the "Orakul HF Loader":
Allows for the direct loading of up to 4 LoRAs from the Hugging Face repository, eliminating the need for manual local downloads.
Includes a toggle for memory management: choose between keeping the model in RAM/VRAM during batch generation or performing a one-time download followed by immediate memory release.
A repository containing my free API models:
https://huggingface.co/OrakulStorm/Zero-Disk-VRAM-Streaming-Engine
My toolkit for training models:
Description
The model is trained using full, non-quantized weights with a 2048x2048 dataset.