Audio Quality Demystified

An interactive look at how the `rtcscore` library uses the E-Model to estimate the Mean Opinion Score (MOS) for WebRTC audio.

The E-Model: A "Penalty Box" Approach

The E-Model calculates audio quality by starting with a perfect score (the R-Factor) and then subtracting penalties for various network and equipment impairments. This infographic lets you explore how those penalties add up.

Râ‚€
Base Score (93.2)
I
Impairments
R
Final R-Factor
→
MOS
Final 1-5 Score

Interactive Impairment Explorer

Adjust the sliders and toggles to see how different factors impact the R-Factor and the final estimated MOS. These are the core inputs for the `rtcscore` audio model.

Simulates data packets failing to arrive. Even small amounts can severely degrade quality.

The delay for a signal to go and come back. High RTT makes conversation feel unnatural.

Delay added to smooth out inconsistent packet arrival (jitter). It contributes to the total delay.

Sends redundant data to recover lost packets, reducing the penalty from packet loss.

Final Estimated MOS

4.2

From R-Factor to MOS

The final step is converting the calculated R-Factor into the familiar 1-to-5 MOS. The relationship is not linear; the impact of changes in the R-Factor is more pronounced at the lower and higher ends of the quality scale.