Hi, I did some preliminary tests with mod_com_amd (Answering Machine Detection app voice_start) and got an accuracy of about 66% (well it was just for 27 calls so far but more will come). So I think I will need to adjust the parameters. So I am planning to either use brute force or a genetic algorithm to find adequate configuration values. I can automate a system to make such calls, play tagged call recording files over them and get the AMD result from freeswitch then check the fitness of the configuration. No problems here. However, setup and processing of calls take time and I was thinking if there would be some way to use mod_com_amd so that I could process hundreds of call recording files in a few seconds. It is a long shot but maybe someone has some hint.