LOOK CLOSER

Every image
has a story.

Discover the colors, light and little details
hiding in your image. Start with the pixels.

A closer look.
A clearer picture.

NO ACCOUNT. NO UPLOAD.

01Select image

02Analyze pixels

03Explore results

IMAGE INPUT01

Private by default. Images are processed locally and never uploaded, stored or sent to analytics.

THE HIDDEN DETAILS02

PIXELS → PERSPECTIVE

More than
meets the eye.

A simple photograph holds a world of information. Let’s make it visible.

01

Image essentials

Dimensions, format & proportions

02

A palette with a purpose

Dominant colors & their distribution

03

The language of light

Luminance, contrast & tonal balance

Real pixel statistics.
No AI predictions or invented insights.

BUILT TO LOOK DEEPER

Color. Light. Perspective.

PHASE 01 / IMAGE STATISTICS

How the measurements work

Images are decoded locally. A worker reads original-resolution sRGB pixels, without resizing or blending colors. Images up to 262,144 pixels use every pixel; larger images use one deterministic random pixel per grid cell, weighted by cell area. Fully transparent pixels are excluded. Partially transparent pixels contribute proportionally, without a background color. Coverage is average alpha, not a count of nontransparent pixels. Animated images are represented by a decoded frame.

Colors are grouped in 5-bit RGB buckets, then clustered deterministically into up to six RGB families using weighted k-means. Shares are approximate and based on visible sampled pixels. Average RGB is calculated in encoded sRGB; relative luminance uses linearized sRGB and Rec. 709 weights. The histogram has 32 equally spaced luminance bins. Tonal spread is measured with population luminance standard deviation and range, not text accessibility contrast.

Sampling can miss rare colors or extrema. Larger images have approximate statistics; bucketed color families are approximate even when every pixel is read. Shadow/midtone/highlight cutoffs are product-defined bins, not standard exposure categories. Browser decoding and low-alpha rounding can affect results. This prototype reads metadata locally, with GPS hidden until explicitly revealed. Spatial maps are bounded-resolution explorations, not sharpness or quality scores. No semantic predictions.