Does the magic color need to be within the frequency range that the camera and video format can record?
It seems possible to produce a four-channel camera sensor specifically for this use case, if it was popular enough. Given how common green screen is in Hollywood I'm surprised I haven't heard of anyone doing it. Maybe Hollywood just doesn't care because they can do regular green screen and just hire someone to fix it in post.
I am guessing that that is dangerous to human health because infrared fails to elicit the flinch reaction that too-strong visible light does.
That's one of the things that makes it suitable for thermal nano cancer therapies. You
(1) Create a nanoparticle with a gold core of the right size that it excites when exposed to near-infrared radiation (the far end of "near" is ideal),
(2) Coat it with something that's hopefully less toxic than the cancer because the gold nanoparticles are pretty reactive (silica was popular last I checked, though I'm not convinced that's actually safe, and there hasn't been much testing on it),
(3) Coat that in something that binds to the right antigens,
(4) Inject that into the patient so that you'll eventually have a tumor rich in these nanoparticles while the rest of the body has a very low concentration,
(5) Shine an intense near-infrared beam at the nanoparticles. It safely penetrates body and causes minimal heating, instead depositing all its energy at the tumor, selectively cooking just the bits of tissue you don't want anymore.
It's yet another way to make mice immortal, but human testing is a long ways off. The hard parts are (2) and (3), along with the (4a) I didn't mention where most of these things don't want to stay "nano" in most chemical environments, and the solution they're suspended in is usually also not ideally suited to being injected into living mammals who would like to retain that "living" property.
Glowing-hot steel produces enough IR that the workers that work with it have 7 times the rate of cataracts as the general population.
NIR is much safer than the visible spectrum at the same power levels. The threshold for a few minutes of high-intensity NIR increasing your chance of eventual cataracts is staring at an NIR source 10x brighter (w/cm^-1) than the sun integrated across all its wavelengths. The threshold for sun-colored wavelengths causing eventual cataracts above the baseline from a few minutes of staring is under 1x.
Also worth noting, very very near IR (basically red, though we can't see it) doesn't quite enjoy those same properties, and during their time working with hot metal you'd expect a lot of energy in that band.
Also worth noting, many steel workers are exposed to dangerous amounts of visible light too. I absolutely believe that they can get enough NIR to cause problems, but if I wanted to try to prove that NIR specifically causes their problems to somebody else then I'd want to try to account for that fact (and for incidental welder exposure, ...).
Another alternative I was thinking about is just to use one of those monochrome astrophotography cameras, something along the lines of this[3]. The monochrome version doesn't have the IR-cut filter and sensitivity is pretty ok down into near-IR.
edit: The IMX492 sensor in that one has a 40% response at 850nm, and 850nm IR diodes are plentiful and shouldn't emit[4] much at all below the 700nm of the IR-cut filter the color camera should have.
Not sure how mixing sensor sizes and such would affect things, or if it's better/easier to just run two of the same camera.
[1]: https://s1-dl.theimagingsource.com/api/2.5/packages/publicat...
[2]: https://www.lifepixel.com/
[3]: https://www.zwoastro.com/product/asi294mm-mc/
[4]: https://lumileds.com/wp-content/uploads/files/DS191-luxeon-i...
Building 2 nigh identical cameras, but one with a sensor outside of the frequency range of the other, is probably a very expensive custom job.
It seems like the cube with mounts is the most expensive part, and there's no way it can cost more than one of your cameras, let alone the lens, right?