The Universal Speech Model to use.
By default, the sidtrain
module uses the USM that is included in the base speaker ID pack. If you are training a new speaker model which you later wish to combine into a system with existing models trained without using the USM, you should set USMEnabled
to False
during training.
Action: | AddTask |
---|---|
Task: | StreamSidTrain, WavSidTrain |
Type: | String |
Default: | None (the default is derived from the base speaker ID pack) |
Example: | USM=baseModel.usm
|
See Also: | USMFile (configuration parameter) |
|