Cobalt wrote:
Air is transparent to near-infrared, and the density of air is – order of magnitude – 1/1000th of the density of a solid.
The thermal radiation picked up by the camera is actually mid-infrared. Specifically, FLIR Lepton camera covers a spectral band of 8-14 µm, just in the range where most substances have at least some absorption peaks due to molecular vibrational modes. The air is indeed too thin to be detected; however, it may be possible to detect other matter entrained by it, including water vapour in large amounts, various organics, and dust. I have just looked at the boiling kettle, and it produced a barely noticeable swirl of steam, discernible in the video but not in the photo.
I find that IR camera attachment incredibly handy for measuring temperatures of components on circuit boards
It’s quite revealing what gets hot and what doesn’t, and with today’s tiny components you can easily have some component running hot without it being obvious. For example I have a chip which apart from measuring voltages (an ADC) it also has a temp sensor in it, and I mounted it close to a terminal block, so I can measure the temperature of the terminal block (for thermocouple CJC). I was getting readings several degrees off and could not understand it. The thermal camera revealed the culprit: a resistor dissipating a few tens of mW and heating up the PCB, just enough 
The better IR sensors are ITAR restricted and in the current news one can see why: lots of military applications. Right now, if one was in Ukraine, they would be dead handy for putting on the front of little bombs which you drop from a DJI drone and which recognise the shape of a tank…
