You can use AudioDeviceInput to control microphone input, and AudioAnalyzer to analyze volume levels
AudioAnalyzer:GetSpectrum returns an empty array if any of the analyzer’s inputs come from an AudioDeviceInput – but for deciding whether AI can hear players, I think .PeakLevel or .RmsLevel should work, if all you need is overall volume!
I doubt that Roblox will give us access to what the player says.
@SOTR654 we are working on speech-to-text transcription (announced here) which could help