Thanks — you’re seeing LiveKit’s AudioMixer drop streams that miss their frame deadlines. Below are targeted, practical steps you can take on low-end devices to prevent those timeout warnings. Which part do you want to try first: reducing processing, increasing buffering, or detecting CPU constraints so you can adapt behavior? Recommended fixes (pick one or combine): • Reduce audio processing ◦ Disable or simplify monitoring/analysis that

_MonitoringAudioStreamIterator

does (VAD, frequent logging, heavy DSP). ◦ Use mono (

num_channels=1

) and lower sample rate (e.g., 16 kHz) when high fidelity isn’t required. • Increase buffering ◦ Raise

queue_size_ms

on the

AudioSource

to give the mixer more leeway for stalls (example:

queue_size_ms=2000

). • Make processing non-blocking and regular ◦ Ensure the monitoring iterator emits frames at steady intervals (10–20 ms) and never blocks on I/O or heavy computation. ◦ Handle backpressure gracefully instead of performing long synchronous work on the audio thread. • Detect and adapt to CPU pressure ◦ On clients, detect CPU constraints (e.g., via the LocalTrack CPU-constrained event in recent SDKs) and disable non-essential features (monitoring, visualizations) when constrained. • Use more stable compute (if running server-side) ◦ Avoid burstable cloud instance types (t3/t4); use dedicated CPU instances for predictable performance. If you want, tell me: • Which SDK or component is producing

_MonitoringAudioStreamIterator

(server-side Python, JS client, custom engine), and • Current sample rate,

queue_size_ms

, and whether that iterator does VAD or other heavy processing. With that I can give exact config/snippet suggestions. Sources: Agent session | LiveKit docs | Resolving "InvalidState - failed to capture frame" errors in AudioSource | Optimizing Video Calls for Low-Powered Devices in LiveKit (JS SDK ≥ v2.14.0... | Troubleshooting Latency and Timeout Errors with Turn Detection on AWS