It does not guess who is in a photo. It only rearranges light and shadow.

Each sample uses standard relative-luminance weighting, then maps into a character set from dense to sparse. It is not face recognition and does not analyze identity.

Monospace characters are usually taller than wide, so the row count compensates for their shape. This is a visual approximation, not lossless reconstruction.

Source images, banner text, samples, and exports exist only in the current browser session and never enter a server-side artwork library.

TXT contains only characters and PNG only the generated image. The site adds no watermark and grants no extra rights to source material.