Microphone Noise Suppression DSP: Spectral Subtraction vs Deep Learning Filters: Competitive Benchmark Analysis and Hardware Interactions

Between Nvidio Voice Isolation, RNNoise, and the newer deep learning stacks like NVIDIA Broadcast, I’m chasing that sweet spot where my callouts stay crisp but my mechanical keyboard clatter gets nuked without turning me into a robot. For competitive play specifically, latency is everything—I’m seeing anywhere from 20ms to 100ms+ depending on the filter and whether it’s processing on the CPU or GPU. So what are you actually running in tournament settings, and what measured latency deltas are you seeing per solution?