Upload two thumbnails and enter your video title. We'll analyze contrast, faces, text density, and predict a winner.
Despite its introductory promise, this source does not detect faces or text density. It reads each image pixel into a canvas and computes luminance as point two nine nine times red plus point five eight seven times green plus point one one four times blue. An opaque red pixel gives seventy-six point two four five, green one hundred forty-nine point six eight five, and blue twenty-nine point zero seven. Bars at two pixels per luminance unit measure about one hundred fifty-two point four nine, two hundred ninety-nine point three seven and fifty-eight point one four. These are image statistics, not viewer responses. A two-pixel black-and-white image has luminances zero and two hundred fifty-five, mean one hundred twenty-seven point five. Both deviations have magnitude one hundred twenty-seven point five, so population standard deviation is also one hundred twenty-seven point five. A uniform opaque gray image at one hundred twenty-eight has contrast zero. Contrast bars at two pixels per unit therefore have widths two hundred fifty-five and zero. The source rounds contrast to one decimal before scoring. It uses original pixel dimensions, so equal display size is not a claim of equal image content. A title of twenty code units gets a two-point bonus for both images. Twenty-one gets five for both. With contrasts one hundred twenty-seven point five and zero, scores change from one hundred twenty-nine point five and two to one hundred thirty-two point five and five. Common two-pixel-per-score bars measure two hundred fifty-nine versus four, then two hundred sixty-five versus ten. The score difference remains one hundred twenty-seven point five. A strict greater-than test selects A only when its score is higher; ties select B. No experiment, click-through measurement, account connection or statistically calibrated prediction happens here.