SplitSize

To process a task across multiple cores, the audio is divided into chunks. The chunks are then distributed between the task managers. The SplitSize parameter specifies the size of the chunks, in seconds.

The optimal chunk size depends on the number of task managers available and the length of the audio to process. For maximum speed, aim to use all available task managers, which requires producing as many chunks as there are task managers. For example, to process a 10 minute audio file across 10 task managers, you would specify a SplitSize of 60 seconds. However, producing more chunks can decrease accuracy because errors can occur at audio divisions. For most tasks, HPE recommends chunk sizes of between 60 and 300 seconds.

Action: AddTask
Task: All tasks containing only supported modules. For a list of the modules that support multicore processing, see the HPE IDOL Speech Server Administration Guide.
Type: Integer
Default: 120
Example: SplitSize=120
See Also:

Overlap

TaskManagers


_HP_HTML5_bannerTitle.htm