NAM vs NNM vs ANN comparison of neural network profiles of tube preamp

In this video i tried to compare the quality of todays famous free amp modeling neural platforms: NAM (Neural Amp Modeler), Amperium NNM, ANN (Tank-G) For comparison i made profiles of the CRAB SATAN tube preamp (this preamp inspired by Randall Satan / Fortin Natas amp). Tried to achieve the highest possible quality of the models on all 3 profiling platforms: NAM - a standard architecture, approximately 10,000 epochs, minor customisation of loss function. NNM - training was in several stages, took almost 16 hours. Can be played as VST plugin and as amp in processor Amperium Live. ANN - neural network for M-Vave Tank G processor, the network was trained for about 15 hours. Recorded from Tank G, connected to send/return Ampero Stomp (ampero uses only as reamp interface to send WAV into tankG )