Multi-speaker mixture

[walidsst]

is it possible to generate a multi-speaker mixture, as I noticed in your package it mixed all folders audio's in the foreground folder, ex: my folder will be like follow:

foreground:
       |__> spk1
                 |__> audio1.wav
                 |__> audio2.wav
       |__> spk2
                 |__> audio1.wav
                 |__> audio2.wav
                 |__> audio3.wav
       |__> spk3
                 |__> audio1.wav
                 |__> audio2.wav

the output of the mixture will be as follow:

--> output_wav = spk1/audio1 + spk2/audio3 + spk3/audio1

but in my case I only want to random mix two spekears, if my foreground folder has 1000 speakers the output wav will contain all the 1000 spkears, (correct me if I'm worng)

also another point I'm interest on is it possible for example to have a custom output data, I need another file to act as a reference (ref most be different from the one used in the mix) of one of the mixed speakers ex:

output:
       |__> mix1
                 |__> mix.wav = spk1/audio1 + spk2/audio3
                 |__> ref.wav = spk1/audio2
       |__> mix2
                 |__> mix.wav = spk2/audio1 + spk3/audio3
                 |__> ref.wav = spk2/audio2

• the package only expect .wav files in my case I only have .flac should I transfer all my files from .flac to wav before using this package (this is additional work and space)