Release notes

August 2024

• To improve the efficiency of Enhanced Speech to Text Built on Whisper, we have implemented a method of possible change of the Beam hyperparameter. By alternating the applied level of the Beam, the customer is able to increase the processing speed while slightly decreasing the model's accuracy.

• For optimal utilization of the Graphics Processing Unit, we have implemented the possibility to perform more jobs concurrently. For now, this is possible only for Speaker Identification Extraction, but in soon should be applied for other technologies