KrispNet – Voice Playback with Deep Learning

Artificial Bandwidth Expansion (we call it HD Voice Playback) refers to the idea of upsampling a lowband audio to wideband audio in a way that it improves voice quality. If the Conferencing Service enriched lowband audio (8kHz) before sending to Laptop users they would hear higher quality audio (16kHz) instead of a voice coming from tunnel. original lowband – 8kHz lowband audio from a regular call

ffmpeg wideband – 8kHz lowband audio converted to 16kHz via

krispNet wideband – 8kHz lowband audio converted to 16kHz via krispNet (2Hz) You can’t hear the difference in audio samples on a Mobile browser since it downsamples audio before playing it.

Source: 2hz.ai