The Mode
parameter sets the algorithm mode for the speech-to-text process. You can set it to:
fixed
–to use a fixed beam width for the speech-to-text process.relative
–to use a model that performs calculations relative to a clock for the speech-to-text process.live
–to use recognition that keeps pace with the rate at which data goes to the server. The rate can change during processing. To enable live mode for streamed speech-to-text conversions, you must set the Mode
parameter in the stt
and the stream
module configuration sections.In live or relative mode, it is crucial to ensure that sufficient time is allowed for the recognition process to be robust and to produce transcriptions with good accuracy. To ensure this:
GetStatus
action. IDOL Speech Server repeats the warning after each hour of audio processing. If data streams too fast for the system's computational resources, recognition accuracy might be impaired. If you use live or relative mode with a DNN acoustic model, certain DNN file size and CPU processor limitations are in place to ensure that the target can be met.
Type: | String |
Default: | Fixed
|
Range: | Fixed , Relative , or Live |
Required: | No |
Configuration Section: | stt and stream modules |
Example: | Mode=relative
|
See Also: | Mode (action parameter) ModeValue |
|