I have time now so I rewrote the code in that historical AI denoise project to accept stereo files, process each channel separately and output stereo at 24bit instead of 16 bit. Def save some time if you got those files. Paste that shit in colab if you need it.