How your color report is made

You can submit one to three photos. Indirect daylight, a neutral setting, minimal makeup, no filter, and an unobstructed face give the system the best chance of seeing your natural color relationships. Multiple photos can reveal whether a reading stays stable when the scene changes, but they cannot rescue several consistently poor inputs.

The first structured vision model reviews the submitted photos and estimates CIELAB values for visible skin, iris, and hair. It returns both observed color—the appearance recorded by the camera—and a white-balanced estimate intended to reduce the influence of the scene. The corrected values use a D65, 2° reference convention so the internal comparisons share one frame of reference.

The model considers evidence across multiple visible areas, image quality, lighting, highlights, makeup, and agreement between photos. Visible eye whites or another neutral reference can support white-balance reasoning, but they are not mandatory. These values remain photo-based model estimates, not readings from a calibrated colorimeter or spectrophotometer.

After the color-evidence step, deterministic code derives four traits used by the rest of the report: temperature, depth, clarity, and facial contrast. Temperature and depth come from the white-balanced skin estimate; clarity combines feature, skin, and iris signals with a LAB-chroma fallback; contrast compares the estimated lightness of skin and hair.

This layer is deliberately rule-based: the same accepted inputs produce the same derived traits. It keeps important transformations inspectable instead of asking a language model to invent every conclusion in one pass.

A second structured vision model receives the photo evidence, the derived traits, and the submitted images. It selects one label from a 12-season framework and records its confidence. This is a model judgment built on structured evidence—not a direct laboratory fact and not a simple lookup from one skin-tone number.

The same stage creates wardrobe-core colors, seasonal highlights, signature colors, and caution colors, with HEX and LAB values plus guidance about where a color works best. Local weather can shape timely clothing suggestions, but it is not allowed to change the underlying color readings, white balance, or seasonal boundaries.

The report can expose several kinds of confidence: whether a color sample appears usable, whether white-balance reasoning had a trustworthy reference, whether multiple photos agree, and how strongly the final evidence supports the selected season. Quality flags explain common risks such as mixed light, shadows, clipping, blur, makeup, or limited reference areas.

A neighboring season may be shown as useful context, but it is not necessarily a statistically ranked runner-up. When evidence is weak or contradictory, the practical response is to read the palette cautiously and retake the photos under cleaner conditions.

Automatic white balance, HDR, compression, display settings, dyed hair, colored contacts, foundation, self-tanner, filters, and reflected colors can all change the evidence. More photos help with consistency; they do not turn a camera into a calibrated instrument.

Seasonal color analysis is a styling framework, not a medical or biological classification. Different frameworks—or a careful human draping session—can place the same person in a neighboring category. If repeated real-world draping in indirect daylight conflicts with the report, trust what you consistently see and use the report as a hypothesis to test.

Photos are sent to the AI providers needed to generate your report and are stored securely while your account exists. Style Look Lab does not sell your photos or use them to train its own models. Contact support to request verified account deletion and removal of associated photos.

The service does not create a faceprint or facial-recognition template, match your face against an identity database, or use your analysis to identify you. The links below describe the current privacy and biometric-data practices in full.