Slices audio clips into chunks of a specific number of seconds at the closest moment of silence (decided by silero-vad). Uses onnx which avoids the pytorch dependency.
