Back to style dna
style dna
2026-02-15

The Algorithms: How Kombinlio Digitizes Style (2026)

Look under the hood of the Visual Hero™ Engine. See how Generative Adversarial Networks (GANs) and Biometric Data Points create your perfect wardrobe.

Beyond the "Sticker"

Patent-Pending Technology We don't overlay. We re-render. Last updated: February 2026

Most "Virtual Try-On" tools are just 2D stickers. They take a photo of a dress and paste it over your photo. It looks fake because it ignores physics.

Kombinlio is different. We built the Visual Hero™ Engine, a proprietary Generative AI model that understands Drape, Texture, and Lighting.

🧬 Experience the Engine: Download the Kombinlio App to run your own biometric scan.


1. The Visual Hero™ Architecture (GANs)

We use a Generative Adversarial Network (GAN) to synthesize photorealistic images. A GAN consists of two neural networks fighting each other:

  1. The Generator: Tries to create an image of YOU wearing the dress.
  2. The Discriminator: Tries to detect if the image is fake.

They loop millions of times until the Generator creates an image so perfect that the Discriminator cannot tell it's AI.

The Pipeline

  1. Pose Extraction: DensePose (R-CNN) maps 32 skeletal joints and surface semantics (skin, hair, background).
  2. Warptime Transformation: The garment is "warped" to fit your Body Architecture, accounting for your Somatotype (Ectomorph vs Endomorph).
  3. Refinement: The GAN resolves high-frequency details (texture, shadows, folds).
Technical diagram of the Visual Hero GAN pipeline. Input Image -> Pose Estimation (Skeleton) -> Warp Warping Module -> Generator Network -> Final Output. Data flow visualization.
Figure 1: The Visual Hero Pipeline. From skeletal mapping to photorealistic synthesis.

2. Style DNA: The Biometric Matrix

Style DNA is not a quiz. It is a Vector. We convert your physical attributes into a 128-dimensional embedding.

A. Color Quantization (Munsell Algorithm)

We do not guess your season. We measure it.

  • Input: We sample 50 points of skin, 20 points of iris, and 50 points of hair.
  • Processing: Key pixels are plotted in the Munsell Color Space (Hue, Value, Chroma).
  • Output: Your precise coordinate determines if you are a "True Autumn" or "Soft Autumn."

B. Morphology Calculation (LiDAR)

As detailed in our Calculator Guide, we use LiDAR depth sensing to measure:

Visualization of a user's face and body with biometric data points overlaid. Color swatches extracted from skin/eyes. Geometric lines measuring shoulder/hip width. Code snippets floating.
Figure 2: The Biometric Matrix. Converting biology into data.

3. The Digital Wardrobe (Graph Theory)

Your closet is a Knowledge Graph. Nodes = Items (Blue Shirt). Edges = Compatibility (Matches with...).

The Compatibility Score

Our "Outfit Planner" uses a recommender system to predict compatibility.

  • Score 0.0: A Clashing Outfit (Orange Top + Purple Pants).
  • Score 1.0: A Perfect Outfit (White Shirt + Blue Jeans).

The AI traverses the graph to find the "Shortest Path" to a maximum score for every day of the week, automating your Capsule Wardrobe.

Network graph visualization. Nodes represent clothing items (Shirt, Pant, Shoe). Lines connect items that 'match'. Clusters form 'Capsule Wardrobes'. Futuristic UI.
Figure 3: The Wardrobe Graph. Solving the Traveling Salesman Problem, but for outfits.

4. Frequent Technical Questions

Yes (mostly). Privacy is paramount. We compute the skeletal map and biometric vectors on your device (Edge AI). Only the heavy texture rendering is sent to our secure cloud TPU for processing.

Can it detect fit issues?

Yes. The physics simulation includes a "Tension Map." If the garment turns red in the heatmap view, it means the fabric is stretching beyond its GSM Elasticity Limit, indicating a size too small.


5. Trust the Code

Fashion is art. Style is science. We built the science so you can focus on the art.

Download the algorithm.