OBS is the one app that tells you its numbers
Conferencing apps give you a bar that wiggles. OBS gives you decibels, and its knowledge base publishes exactly where the coloured zones change. That makes it the only app in this cluster where a browser-based mic test can be made to agree with the application precisely, which is why this page adopts OBS’s thresholds outright instead of inventing its own.
| OBS indicator colour | Range OBS publishes | What MicTester calls it |
|---|---|---|
| dark green | less than −50 dBFS | Nothing arriving |
| light green | between −50 and −20 dBFS | Too quiet |
| yellow | between −20 and −9 dBFS | On target |
| red | between −9 and −0.5 dBFS | Hot |
| white | larger than −0.5 dBFS | Clipping risk |
Two of those boundaries have names in the same article. −20 dBFS is the Alignment Level, and OBS explains the choice directly: the value was chosen to be near the average audio level for speech, and traditionally you play a sine tone through a whole equipment chain and set each device so it reads exactly there. −9 dBFS is the Permitted Maximum Level, above which OBS says there is a small but real chance the sound clips before it reaches the viewer.
The article also explains why the ceiling was set nine decibels below full scale rather than at it, and the reasoning is three margins of three: three decibels because a sample peak meter can under-read the true peak by that much, three more because peaks are hard to read off a moving bar, and three for alignment error across a chain of equipment. OBS notes that its peak-hold feature eliminates the middle one for anyone who uses it — which is why the meter on this page has a hold reading and a reset button. The eleven decibels between the alignment level and the permitted maximum are something else entirely: they are the room speech is expected to occupy above its average, which is why a crest factor of about nine decibels fits inside them.
Gain staging, in the order OBS insists on
The single most useful sentence in the OBS audio guide is that you always start at the device. The fader is the last adjustment, not the first, because everything before it has already decided how much signal and how much noise you have.
- The microphone or interface. OBS’s guide says to check for a physical knob, usually labelled Gain. This is where you win or lose.
- The operating system. OBS gives a specific Windows trick here: in the Sound control panel’s Recording tab, open the device’s Properties, go to the Levels tab, right-click the percentage value and choose decibels — then aim for around 0.0 dB for the least clipping. The setting persists when you switch the display back to percent.
- This page. Read the peak hold. If speech peaks land between the two marks, the first two steps are right and you can stop adjusting.
- The OBS fader. Only now, and only for balance against your other sources. OBS notes you can enter percentages above 100% in Advanced Audio Properties if you truly need more.
- Record a session and listen back before going live — OBS’s own final instruction, and the reason the recorder above exists.
Every OBS audio control worth knowing
| Control | Where | What it does | Browser equivalent |
|---|---|---|---|
| Mic/Auxiliary Audio device | Settings → Audio | The Quick Start Guide’s step 3 says OBS captures desktop audio and microphone by default, and to set the devices manually here if the mixer meters are not moving. | deviceId |
| Fader | Audio Mixer dock | Post-capture volume, in dB. Adjust it last — OBS’s guide says to start at the device. | none |
| Monitor button | Audio Mixer dock | Plays a source back through the monitoring device set in Settings. | none |
| Noise Suppression filter | Filters on the audio source | Three methods are documented: RNNoise (the default, higher quality and more CPU), Speex (configurable, default suppression level −30 dB) and NVIDIA Noise Removal, which needs the NVIDIA Broadcast SDK installed. | noiseSuppression |
| Downmix to Mono | Advanced Audio section of the Audio Mixer | Forces one-channel audio to both sides. OBS’s guide flags the symptom it fixes: only the first meter lighting means viewers hear you in the left channel only. | channelCount |
| Audio Meter Decay Rate | Settings → Audio | How fast the meter bar falls back after a peak. | none |
Why a browser test is worth running at all when you have OBS
Because OBS shows you the signal after its own filter chain, and this page shows you the signal before anything. When a stream sounds wrong, the useful question is whether the raw capture was already wrong, and the fastest way to answer it is to look at the same microphone somewhere that has no filters at all. If the peak hold above sits at −38 dBFS with every switch off, no OBS filter will rescue you: you are amplifying a weak signal and the noise floor comes up with it. Fix the gain, then go back.
The second reason is speed. Loading OBS to check a microphone before a stream is a minute you may not have; this page takes ten seconds and gives you the same number. And if the microphone turns out to be dead, the Windows 11 page maps the browser’s exact rejection to the thing that is broken, which is more specific than “OBS shows no meter”.