Preset for the speakers

[rasenganai]

instead of downloading the speaker clips from google drive for the wav clips. save the GST outputs as npy files which can be loaded and used for speaker information.