I’ve been working on GoCaller for over six months, a bot that gives you TeamSpeak-style shot calling in Discord: one person talks and every squad hears it live in their own channel, even across servers. It has several modes (one-way broadcast, hub where the leader hears everyone, and full conference), and each one needed its own audio routing and mixing with mix-minus so nobody hears their own echo. Most of my time went into CPU and latency optimization: skipping decode and re-encode whenever possible, keeping the mixer loop lock-free, reusing buffers to avoid allocations, and pausing work for empty channels.

Repo: https://github.com/sealbro/go-discord-caller
Try it: https://discord.com/discovery/applications/1484911601210495038

Would love feedback on the design, or ideas for squeezing out more performance!