The NMix
parameter sets the target size of the speaker model being trained in terms of the number of components. Increasing the NMix
value produces larger models. Training larger models might increase the speaker ID accuracy but also increases matching times.
Action: | AddTask, CheckResources |
---|---|
Task: | StreamSidTrain, WavSidTrain |
Type: | Integer |
Default: | 256 |
Range: |
64–512 |
Example: | NMix=256
|
See Also: | NMix (configuration parameter) |
|