Does training with Multi Speakers improve model quality to generalize better and handle variety inputs?

[chigkim]

Does training with Multi Speakers improve model quality to generalize better handle variety of unseen input sources? Or, it just lets you inference to different speaker?

[Majboor]

recognize and generalize differences between voices.

[chigkim]

Thanks. Does it have impact on quality of output when processing variety of unseen input sources?

[Majboor]

multiple set of speakers = better learning to recognize and generate this way the model is more representative of natural language use.