Note this video is not showing near IR effects.. Either they learned from what I said, or they are using mid-IR. It would help if the white balance compensation was set on the camera for that weird greenish light they were using to film with.
Either way, for a laser professional like me, that was somewhat painful to watch, its working, but its not quite there. Machine stretched and aligned fibers are the way to go. Even the source alignment when they couple the fiber to the laser module and
bond it needs to be automated... Which is done all the time and is available off the shelf.... How to automate the alignment has been published in great detail over the years.
There are some scientific uses for stretched fibers. Commercial fiber tapering machines are off the shelf items.
As for knowing the fiber diameter needed, the math is readily available if you know the index of refraction of the hair.
The electronic analogy is knowing guide cutoff wavelength, and S Parameters for your transmission line, ie S11, S12, S21,S22. Then you break open braid open on the Coax, introducing a loss, and measuring again.... The modeling software would be on the order of 10K$ for the basic package. Optical modeling software is not cheap...
It should be theoretically possible to support the fiber in a way that TIR is not spoiled, but that is going to need phase II funding to engineer. That mount would let it be cleaned...
That clearly is not something like stock 62/125 u fiber....
Steve (edited 12:46:48 on 9/30/2015, because I screwed up on S22)