3 ms·
If the audio of a phone call was pre-recorded on a good microphone it could sound awesome even over low bandwidth "voice" connections. Audio for voice calling i
by giantrobot 3y ago
If the audio of a phone call was pre-recorded on a good microphone it could sound awesome even over low bandwidth "voice" connections. Audio for voice calling is optimized almost entirely for latency. This means super short buffers and relatively simple encoding.
Streaming video/audio has a distinct advantage of not having the same latency sensitivity. It's not a two way conversation so if the stream is a second or two behind the wall clock as long as it's in sync you're never going to know unless you're watching a stream of an event you can see live, then you'll see the encode and transfer delay.
Besides latency, the microphone/audio input pipeline on a phone is in one of the shittiest recording environments it could be in. It's right next to the users mouth, sometimes far too close, so it needs automatic gain control to keep from being overdriven. It also needs to filter out the audio coming from the speaker lest it generates feedback. The microphone itself also needs to be very small and is attacked to the body of the phone so it's difficult to isolate from body sounds like the user's hand rubbing over the body.
The fact a cell phone can even transmit reasonable sounding voice is a minor miracle involving many thousands of engineer-hours. The extra processing power and bandwidth available today to handle wideband voice helps a little with quality but it's still very latency sensitive and still suffers vs less latency sensitive codecs or one-way streaming.