AI Noise Reduction (Studio Sound)
evmux includes an AI-powered audio enhancement layer that runs on your microphone input before your voice is sent to other participants or the broadcast. It removes background noise (traffic, fans, crowds, keyboard clicks) and applies studio-grade processing (equalizer, compressor, noise gate, de-esser, auto-gain) so basic microphones sound professional.
In settings the option is labelled “AI noise reduction” with a small “New” badge. If it is not available on your account, you see the browser’s built-in noise suppression instead, labelled “Background noise reduction”.
What it does
When enabled, evmux:
- Takes your microphone input.
- Runs it through an AI noise reduction model.
- Applies a chain of studio processors (EQ, compressor, noise gate, de-esser, auto-gain).
- Sends the processed audio on in place of the raw microphone feed.
The processed audio is what other participants hear and what gets broadcast. There is no separate “studio sound off for broadcast” toggle.
When it’s off
Turning the option off is genuinely off, not just a lowered setting: your raw microphone audio passes through untouched, with no extra processing in the path.
If the noise reduction module fails to load (for example, because of a network problem), evmux disables it automatically and your audio falls back to the browser’s normal behavior. You do not get an error popup.
Interaction with other audio settings
AI noise reduction has known conflicts with two other audio controls. The audio settings panel shows a warning in real time when both are on.
Conflict 1 — AI noise reduction + Echo cancellation
When both AI noise reduction and Echo cancellation are enabled, a yellow warning appears under the echo-cancellation checkbox:
⚠ Echo cancellation may cause audio scrambling when used with AI noise reduction.
The two filters compete for the same audio. Leave echo cancellation off when AI noise reduction is on (which is also the recommendation if you use headphones).
Conflict 2 — AI noise reduction + Stereo audio input
When both AI noise reduction and Stereo audio input are enabled, a yellow warning appears under the stereo checkbox:
⚠ Stereo is not supported with AI noise reduction. Audio will be processed as mono.
The noise reduction is built for voice and works in mono. Your signal is mixed down to mono before processing and stays mono through the broadcast, so a stereo source reaches listeners as a single channel.
Interaction with virtual background
Virtual background processes only your video and AI noise reduction processes only your audio, so the two never interfere with each other. You can use both at the same time.
How to enable / disable
In the editor: Settings → Audio → “AI noise reduction” checkbox. See audio-management and settings.
The setting is per user, not per session: it is saved with your account and stays on across sessions until you change it.
Quick answers
| Question | Answer |
|---|---|
| Where is it? | Settings → Audio → “AI noise reduction” checkbox |
| Does turning it off really turn it off? | Yes — your raw microphone audio passes through unprocessed |
| Why does my stereo sound mono? | With AI noise reduction on, audio is mixed down to mono. The settings panel warns about this |
| Why do I hear scrambled audio with echo cancellation? | Both filters compete; the settings panel warns you to turn one off |
| Does the broadcast hear the AI-processed audio? | Yes — the processed audio replaces the raw microphone feed for everyone |
| What if it fails to load? | It switches itself off and your audio falls back to normal, with no error popup |