Sorry I was unclear so let me rephrase it : Room Shaper is not a Linear Time Invariant (LTI) process that can be characterized by its impulse response and therefore implemented by the mean of convolution (which is what standard HAF filter and Audiolense do).
There is some sort of “pre-processing” of the input signal that is looking for the problematic/resonant frequencies of the room (as measured by the user and uploaded in the interface). When the input signal contains a “note” at a problematic frequency, Room Shaper is sort of “shortening” that note with a form of wavelet transform. The Room Shaper shortened note is then played by the audio system in the room, and the room “expands” the short note because of the resonance, which therefore sounds closer to the “original”, thanks to the resonance. This precompensation is tricky to achieve in order to sound as natural as possible, while keeping computing ressources reasonable. From what I’ve heard, Thierry did a good job (I am a customer and user of Room Shaper).
A proper implementation in Roon would require to integrate fully with Roon DSP, for example through APIs. This is not possible today, but I agree it would be extremely beneficial so that third parties like HAF or hobbyists can provide more DSP options, for example:
- headphone correction for all models (there are huge databases out there of headphone measurements on GitHub)
- alternative and better upsampling algorithms à la HQPlayer. HQPlayer is definitely outperforming Roon’s standard upsampling algorithms
- automated room correction modules, based on system measturement à la Dirac
- cross talk reduction filters
- non linear room mode correction like Room Shaper etc…
A full VST integration seems unlikely as it raises GUI challenges, but an opening of Roon’s DSP pipeline would bring many benefits.
There’s been no real improvement to Roon DSP since Roon 1.3 three years ago except UI refresh and bug corrections, I hope we’ll see a new step soon!