Barber Half Gainer - Anti-Aliasing + Better Accuracy with New Archi

In this pack, some captures of my Barber Half Gainer have been retrained with a larger neural network. For the anti-aliasing treatment, I used a "super input" file and a new architecture that could possibly be called "REVzHI"(?). I defined it in continuity with the REVxSTD, REVySTD and REVyHI architectures developed by thegearpage38 and presented on TGP at the end of 2025. This "REVzHI" gives results very close to REVyHI. I just noticed very slightly deeper bass for REVyHI and slightly more open treble (more clarity) for "REVzHI". The reason I chose to use and present it here is because "REVzHI" has a very interesting advantage over REVyHI. For me, it reduces training time by at least half locally with NAM version 12.2 (4h.. VS 2h..). And for those who don't have a high-end graphics card, that's a significant point. I have also defined this new architecture in a standard version ("REVzSTD") which already gives excellent results and with a training time similar to that of xSTD. You'll find in the pack 6 profiles in HD mode ("REVzHI") and 2 profiles in STD mode ("REVzSTD"). NAM files with calibration metadata included. Technical information: Interface: Motu M2 Reamp Box: Radial Pro-RMP Reamp level: 12 dBu Captures normalized to: -3 dB Epochs per capture: 1000 ESR on HD mode: 0.0001034 to 0.0001252 ESR on STD mode: 0.0001783 to 0.0001788 If you like my work, you can supporting me via paypal. https://paypal.me/kosmicaudio