APRIL 22, 2025
Live Stream Audio Problems, and How to Fix Them Mid-Show
Audio is the failure people actually leave over. A soft-focus shot or a dropped frame gets forgiven within a few seconds. Sound that crackles, clips, disappears, or is simply too quiet empties a stream faster than anything else on the technical list.
Too quiet is the one that gets missed, because it never announces itself. Nothing is obviously broken, nothing alarms, and everyone in the control room is listening on good headphones at a level they set themselves. The show just quietly underperforms for everybody watching.
So listen the way the audience will. Put the stream on a laptop and on a phone, both on their built-in speakers, with no headphones. If you cannot hear it clearly on a cheap laptop, the levels are too low. It takes thirty seconds and it is the single check most likely to catch a problem nobody else in the room can hear.
It is also the failure with the least room to think. Video degrades gradually and visibly. Audio tends to go from fine to gone, in front of an audience, with the show still running.
What it actually looks like when it goes
We were backstage at a conference, running the show from a control table set up beside a walkway that VIPs and guests were using to reach the stage. Somebody walked past, caught the corner of the table, and knocked a piece of gear out of its connection. The master feed lost audio instantly.
We knew before the room did. We run a monitoring tool we built ourselves that watches the outbound stream and alarms the moment audio on the master goes flat. Our sound engineer checked his end and confirmed audio was still leaving the mixer, which told us the fault was downstream of him. That single confirmation collapsed the search from somewhere in the chain to one device, and we re-patched around it.
Nothing about that recovery was clever. It was fast because the signal path was known, the monitoring was already running, and there was somewhere to re-patch to.
The order to check things in
When audio goes and the show cannot stop, the sequence matters more than the diagnosis. Ours:
- Trace the signal path from source to output, confirming audio is present at each step. You are not looking for the cause yet. You are looking for the last point where it was still fine.
- Power cycle anything that looks wrong the moment you reach it. A surprising share of live audio faults are a device that has quietly stopped passing signal and will start again.
- Swap the failing device if you have a spare within reach.
- If you cannot swap it, reroute around it.
That last step decides how bad the outage is, and it is decided before the show starts. A secondary cable run you can re-patch to in seconds is the difference between a few seconds of dead air and a broadcast that never comes back. Running that second path costs almost nothing on load-in day.
Most audio problems are bought weeks before the show
The failures hardest to fix live are the ones decided long before anyone arrives, when someone chose microphones for a room they had never stood in. The core mistake is treating that as a microphone-selection problem. It is a room-and-distance problem.
Buying for the voice instead of the acoustics
In an unfamiliar room, directivity is the only insurance you have. Pattern choice should be driven by where the PA is hung, what the reverb time is, and how far the mic ends up from someone’s mouth. You know none of those. That argues for tighter patterns and shorter working distances than you would otherwise choose.
The polar plot is not telling you the truth
Most cardioids are effectively omnidirectional below 200–300 Hz. Everything that actually ruins an unfamiliar room (HVAC rumble, street noise, a subway underneath, footfall on a wood floor over a basement) walks straight in regardless of how carefully you aim. And a hypercardioid’s rear lobe is a real liability when you do not know where the mains are hung.
Critical distance decides more than the capsule does
Every doubling of distance costs 6 dB of direct sound. The reverberant field does not move. A microphone that sounds excellent at eight inches is a different product entirely at thirty.
That is why lavaliers are the highest-risk remote purchase. You start eight to ten inches off-axis, lose another 6 dB on every head turn, and you are doing it in a room whose RT60 you are guessing at.
More open mics than the room can carry
The math here is unforgiving: every doubling of open microphones costs roughly 3 dB of gain before feedback. A lectern gooseneck, two handhelds, and a couple of ambience mics in a low-ceilinged room with the mains downstage of the talent, and you simply cannot get loud enough. People find this out at soundcheck, with the doors about to open.
No spectrum plan
Wireless is where never having been in the building bites hardest. You do not know the local DTV occupancy. You do not know whether the hotel has forty other systems running in the ballroom next door. You do not know whether the walls are plaster over metal lath, or whether the LED wall is about to throw hash across your entire band. Buying a system without checking band availability for that specific market is a good way to own equipment you cannot legally or practically use on the day.
Assuming somewhere to put it
These sink more shows than frequency response does:
- No stand position that is not in a camera shot
- A glass conference table that turns a boundary mic into a resonator
- A landmarked interior where you are not permitted to attach anything to anything
- A lectern with no gooseneck hole
Specs the room noise floor makes irrelevant
If the room’s NC rating is 40, the difference between 7 dBA and 12 dBA of self-noise does not exist. You are paying for a number you will never hear in that space.
For anything with a camera on it, wardrobe decides lavalier placement more than any other single factor, and wardrobe is unknowable remotely. A guest in a lapel-less dress, heavy jewelry, or a stiff synthetic makes half a lav kit useless.
What to ask for when you cannot get in the room
If a site visit is not possible, this is the short list that de-risks it most:
- A phone video of someone clapping and talking in the space, with the HVAC running
- A photo of the ceiling and any wall treatments
- Rough dimensions
- Where the PA lives
- What is already installed: house DSP you may or may not get access to, and whether there is a tie-line you can reach
Then buy the boring, redundant, tighter-pattern option, with more spares than feels necessary. The premium on an unfamiliar room is paid in redundancy, not in a nicer capsule.
Some of this you can run yourself. Some of it you shouldn’t.
If the room is big, the audience matters, or there is no second chance at the moment you’re capturing, that is the point where this stops being a settings problem and starts being a production one. Our sound engineering work covers exactly this.
Where production time really goes on a livestream, and which parts are worth handing off. An honest breakdown of what a studio does for you.
Four people, one conversation, and nowhere to hide. Camera assignments, the mic plan, and the cutting decisions that make a panel watchable on stream.