HDR is kind of cool – a nice way to get past the limitations of solid-state image sensors and recover some of the latitude of film, even improving on it.
The problem is that solid state image sensors tend to have very linear responses to light – an underexposed image vanishes into the noise floor of the sensor while an overexposed image clips off to pure white. Film exposure response is commonly called an “s-curve” and basically means there’s some data in the random conversion of a light sensitive molecule here or there even in the most underexposed image, and a few that resist converting under the harshest blast of light such that there is perceivable data in both.
This film is a pretty impressive example of HDR video. But there’s something a bit odd about such a technical achievement in cinematography mixing up “underexposed” and “overexposed.” The funny thing is, they’re using the terms as in making a print (e.g. printing on photo paper) or an x-ray where more light darkens the print: the paper starts out white and turns black with more light vs. a film or digital exposure where the media yields a black result that increases in representational lightness with with increasing light exposure.