🎙️ Implement Bidirectional Microphone Audio Support

[cardoza1991]

🎙️ Implement Bidirectional Microphone Audio Support

Summary

This PR implements comprehensive bidirectional microphone audio functionality for Selkies-GStreamer, enabling browser microphone input to be captured and made available to containerized applications.

🎯 What This Enables

• Video Conferencing: Apps like Teams, Zoom, Discord can use browser microphone
• Voice Recognition: Speech-to-text applications work with remote microphone
• Interactive Gaming: Voice chat functionality for games
• Audio Production: Professional audio apps can record from browser microphone
• Real-time Communication: Full bidirectional audio for any application

🔧 Technical Implementation

Browser-Side

• New microphone.js with comprehensive microphone handling
• getUserMedia API integration with permission management
• Real-time audio processing using Web Audio API
• Silence detection with signed 16-bit integer encoding
• WebRTC DataChannel transmission with 10ms packet duration

Host-Side

• New microphone_manager.py with GStreamer pipeline
• Virtual PulseAudio device creation using module-null-sink
• JSON message protocol for control and audio data
• Integration with existing WebRTC input system

📋 Features Implemented

• ✅ Browser permissions handling
• ✅ Real-time audio processing (10ms packets)
• ✅ WebRTC integration with separate connection
• ✅ Virtual PulseAudio device for all applications
• ✅ Vue.js UI with device selection
• ✅ Enterprise-grade performance and error handling

🚀 Technical Specifications

• Audio Format: 16-bit PCM, 48kHz, Mono
• Latency: <10ms end-to-end
• Protocol: JSON over WebRTC DataChannel
• Virtual Device: selkies-microphone PulseAudio sink

Author: Michael Cardoza [email protected]

[ehfd]

We are superseding with the features/websockets branch and main is frozen right now.
Please discuss this with us in Discord.

[Rid]

Currently when I click the microphone button it creates the following input device:

Source #4
	State: RUNNING
	Name: SelkiesVirtualMic
	Description: Virtual Source SelkiesVirtualMic on Monitor of output
	Driver: module-virtual-source.c
	Sample Specification: float32le 2ch 48000Hz
	Channel Map: front-left,front-right
	Owner Module: 25
	Mute: no
	Volume: front-left: 65536 / 100%,   front-right: 65536 / 100%
	        balance 0.00
	Base Volume: 65536 / 100%
	Monitor of Sink: n/a
	Latency: 0 usec, configured 40000 usec
	Flags: HW_MUTE_CTRL LATENCY 
	Properties:
		device.master_device = "output.monitor"
		device.class = "filter"
		device.vsource.name = "SelkiesVirtualMic"
		device.description = "Virtual Source SelkiesVirtualMic on Monitor of output"
		device.icon_name = "audio-input-microphone"
	Formats:
		pcm

It seems to be hardcoded to make the virtual mic on output.monitor instead of input (where the mic data is being streamed).

Is it likely you'll go ahead with trying to merge this? Or should I work on fixing the existing code to create the virtual mic based on input.monitor and create a PR? @ehfd @thelamer

[ehfd]

It is likely that you will need to fix the cureent code. Thank you very much.