Recommended action: Update today to ensure the best performance during your next video call.
The microphone effect used a temporal RNN trained on a dataset of 100,000+ noise types (keyboard clicks, fans, traffic). V1.0.0.25 introduced a slider for noise suppression strength: to 100% (aggressive) . At 100%, the model successfully removed vacuum cleaner sounds but introduced “digital warbling” on human sibilance (‘s’ and ‘sh’ sounds). Nvidia Broadcast V1.0.0.25