Add bidirectional audio support for HomeKit cameras

[andersbaekgaard]

Summary

• Add bidirectional audio support for HomeKit cameras (speaker/talkback functionality)
• Audio path: WebRTC/Browser mic → FFmpeg transcoding (OPUS→AAC-ELD) → HomeKit speaker via SRTP
• Implements connection tracking to prevent duplicate consumer-track pairs
• Adds accept-audio parameter to FFmpeg for waiting on external audio sources
• Thread-safe SRTP writes with mutex protection

Requirements

• FFmpeg must be compiled with libfdk_aac for AAC-ELD encoding (required for HomeKit speaker audio)

Test plan

• Tested with Aqara G4 doorbell - microphone works in Home app
• Verified audio latency is optimal for the device (~3s jitter buffer is device firmware, not go2rtc)
• Test with other HomeKit cameras that support bidirectional audio

---

Implemented by Anders Baekgaard in cooperation with Anthropic Claude Opus 4.5
https://claude.ai/

[AlexxIT]

Thanks! The implementation is quite complex and introduces a lot of confusion into the core.
I won't accept it in this form, but I will try to understand the idea and adapt it when I have time.

And I really don't like the idea of ​​using custom ffmpeg. This is one of the block factors for implementing two-way for HomeKit cameras.

[andersbaekgaard]

Thank you for the prompt feedback.

I understand your concern about complexity, the PR got much more complex than I expected when first looking into it. I could not get streams matching to work for the aac-eld transcoder producer, due to archtiectural limitation. With your insight, you could maybe make it simpler.

Note that there is also a simple bug fix.

[kombihacke]

Would really love to see this integrated. Cannot get my Aqara G410 running with 2 way audio. Thankful for any advice.