It cannot be done today, because Windows is taking over the device once you connect it, and the only way around that is a programmable EDID emulator (not the $20 cheapos, the $200+ ones). For that price, you’d better invest towards a new headset.
The way WMR is implemented, there are parts inside the OS, so for everyone claiming “just open source it”, they don’t realize that it will just give you a bunch of unusable code (because some of the most critical code like the compositor is inside DWM, and that’s not getting open sourced XD).
There are solutions like Monado reimplementing some of this code, but 1) they do not work on Windows (contrary to beliefs, the Monado support for Windows does not work with devices, only emulation), 2) even if you implemented the missing code on Windows, it is subject to the EDID limitation I listed above and 3) would not provide a user-friendly system (if you hated dealing with WMR, trust me you don’t want to deal with developer-targetted products).
Also, IMO, this effort is vain - you guys are way way overestimating how many people actually use these devices… The math above claiming $5m revenue is off by a ridiculous factor. I can’t share the numbers, but trust me, you are way way way off. Nobody is going to make real money building this.
By the time there can be a solution free of the extra investment (EDID emulator) or functional-enough (eg: Monado support), the device will be easily out-performed by other headsets (I mean, it already is today) and probably even cheaper headsets by then (something like Quest 3 is already there).