Bugfix: Audio Distortion & PipeWire Hang (Linux)¶

This document describes how to resolve audio distortion (“klirren”), robotic voice artifacts, and system audio hangs that can occur when using SL5-aura-service alongside other media applications like OBS, AnyDesk, or high-frequency TTS streams.

Symptoms¶

Voice output/input sounds distorted, metallic, or “clirring.”
System audio hangs with the message: Establishing connection to PulseAudio or Please wait.
Total audio loss after high CPU load or concurrent stream usage.
Journal logs show: spa.alsa: hw:X: snd_pcm_status error: No such device.

Root Cause¶

On Manjaro and other modern Linux distros, PipeWire manages audio. Distortion usually stems from:

Buffer Underruns: Conflict between concurrent streams (e.g., AnyDesk capturing audio while TTS/OBS is running).
Sample Rate Mismatch: Frequent switching between 44.1kHz and 48kHz.
USB Timing Issues: High bus load causing USB headsets (like Plantronics/Poly) to temporarily disconnect.

Solutions¶

1. Immediate Recovery (The “Nuclear” Reset)¶

If the audio stack is frozen or distorted, kill all audio-related processes. They will auto-restart immediately.

# Force kill PipeWire and its PulseAudio compatibility layer
killall -9 pulseaudio pipewire pipewire-pulse

2. Prevention & Stability Settings¶

Disable AnyDesk Audio¶

AnyDesk often tries to hook into the audio device, causing hardware conflicts.

Action: Open AnyDesk Settings -> Audio -> Disable “Transmit audio” and “Reproduce audio”.

Fix PipeWire Sample Rate (Recommended)¶

Force PipeWire to stay at 48kHz to avoid resampling artifacts during TTS playback.

Create the config directory: mkdir -p ~/.config/pipewire
Copy the default config: cp /usr/share/pipewire/pipewire.conf ~/.config/pipewire/
Edit ~/.config/pipewire/pipewire.conf and set:
```
default.clock.rate = 48000
```

Restart services:

systemctl --user restart pipewire pipewire-pulse wireplumber

3. Post-Production Recovery (FFmpeg)¶

If you recorded a session and the audio is distorted (“clirring”), use the following ffmpeg filter chain to repair the file.

Recommended Repair Command¶

This command applies a de-clipper, noise reduction, and a low-pass filter to remove high-frequency digital artifacts without re-encoding the video.

tested and very good:

ffmpeg -i input_video.mp4 
    -af "adeclip, afftdn, lowpass=f=5000, volume=1.5" \
    -c:v copy reparatur_output.mp4  

may a better with (not tested):

ffmpeg -i input_video.mp4 \
  -af "adeclip, afftdn, lowpass=f=3500, volume=1.8" \
  -c:v copy repaired_audio_output.mp4

not tested:

ffmpeg -i input_video.mp4 -af "arnndn=m=cb.rnnn, lowpass=f=5000" -c:v copy ki_reparatur.mp4

Filter Breakdown:

adeclip: Rounds off digital clipping spikes.
afftdn: Reduces FFT-based digital noise.
lowpass=f=3500: Cuts off frequencies above 3.5kHz where most “clirring” occurs (makes the voice clearer/warmer).
volume=1.8: Compensates for volume loss during filtering.
-c:v copy: Keeps original video quality (extremely fast).

Debugging Tools¶

To monitor audio health in real-time during development:

pw-top: Shows real-time errors (ERR column) and buffer status.
journalctl --user -u pipewire: Checks for hardware disconnects.