The (acoustic) echo cancellation (AEC) feature works great, but there are two bugs connected to it.
Bug 1: The voice activity detection (VAD) is fed the unprocessed audio, if set to automatic. That means if you play something on the speakers, that is later removed by the AEC, the VAD will still activate.
Bug 2: When using the “Record Test” feature in the audio tab and listening to it afterwards (by clicking on “Listen”), the speaker signal that gets subtracted from the microphone signal by the AEC is the speaker signal during listening, not the speaker signal during recording (which it should be). That might create the impression, that the feature is actually not working and worsens the sound (it sounds muffled).
Feature request: Teamspeak Processed Input
Problem:
It would be great, if you could use the processed input audio in recording software, but currently there is no standalone software, which really provides AEC functionality, at least I could not find one. The AEC in the software Krisp does not really work (especially with voices being played back by the speakers) and sounds awful compared to the implementation in teamspeak. The realtek audio drivers technically support AEC, but I have not tested that. Also, that only works when the input and the output connected to one realtek device.
Solution:
Adding an audio input in Windows/Linux/MacOS that gets fed the processed input audio (like the software Krisp does). That might be useful specifically for AEC.
