Mix-Minus: Why Your Remote Guest Hears an Echo
Every participant in a broadcast must hear everything except themselves. What an audio bus really is, how to build one feed per listening position, the three failures that account for almost every echo, and why headphones are part of the design.
On this page
Your remote guest hears themselves a moment after they speak, and it derails them. The fix has a name — mix-minus — and it is one idea, not a technique: every participant must hear everything except themselves. Everything below is that sentence, applied.
The one table you need
Write down who is in your broadcast, and what each of them should hear. For a host, a remote guest and the audience, the answer looks like this:
| Who is listening | What they must hear | What they must NOT hear |
|---|---|---|
| Host (in your headphones) | Remote guest, playback, room mics | Your own microphone, live in your ears |
| Remote guest | Host, playback, room mics | Their own voice coming back |
| Room / PA speakers | Remote guest, playback | Any microphone that is in the same room |
| Viewers watching the stream | Everything | Nothing — the viewers' mix is the only complete one |
Notice that no two rows are the same. That is the whole problem. A single "output" that you send to everyone is wrong by construction, because every participant needs a different subtraction. If your setup has one mix, someone is hearing themselves.
Why hearing yourself is different from an echo
There are two separate faults that both sound like "echo", and they are fixed in different places, so it is worth telling them apart in the first minute.
A loop. Your guest's voice reaches you, you send your whole programme mix back to them, and their voice is inside that mix. They hear themselves delayed by the round trip. If the return path is loud enough, the loop can also feed itself and build into howling. This is a routing fault. Turning down volumes hides it; only subtraction removes it.
Room spill. A microphone in your room picks up the loudspeakers in the same room. The guest hears themselves because your speakers are playing them and your microphone is capturing your speakers. This is an acoustic fault, and the fix is headphones, not routing.
If your guest hears themselves and you are on headphones, it is a loop. If your guest hears themselves and there are speakers on in the room with an open microphone, it is spill — and no amount of clever routing will fix it while those speakers are on.
What a bus actually is
A bus is a labelled destination that sources can be sent to. That is all. It has no physical existence and no cost; you can have as many as you have things to send audio to.
The reason a bus feels abstract is that the simplest possible setup has exactly one, and when there is only one destination nobody needs to name it. The moment there is a second listener with different needs — a remote guest, a room PA, a translator, a recording — you need a second destination, and you need to choose independently what goes into each one.
So the mental model that makes all of this easy is: one bus per listening position, not one bus per microphone. Count your listening positions from the table above. That is your bus count. Then, for each bus, tick which sources go into it, leaving out whatever that listener is the source of.
Three failures and the sign that names each one
These are the three that account for almost everything, and each has a tell that identifies it without any measurement.
The guest hears themselves, on a delay. They pause, stumble, or start speaking more slowly and carefully. The tell is that it only happens to the remote participant, never to anyone in the room. The cause is that the return you send them contains their own signal. The fix is a bus for the guest that excludes the guest.
The audio is in the recording but not in the stream, or the reverse. The tell is that everything sounds right in the room and to the host, and only the output is wrong. The cause is that the recording and the broadcast are being fed from different buses, and a source was added to one and not the other. The fix is not to add the missing source in a hurry mid-show but to check every bus against your table before you go live — the fault is always that two destinations drifted apart.
Howling that builds after a few seconds. The tell is the build-up: it starts as a hum or ring and grows. The cause is a loop with enough gain to sustain itself — a microphone hearing a speaker that is playing that microphone. The fix is to break the loop, not to reduce it. Mute the path, then rebuild it as a subtraction.
Why "minus" and not "mute"
People often try to solve this with a mute button, pressing it when the guest speaks. It works, for about a minute, and then it fails in the two ways it always fails: the operator is late, so the first syllable of every sentence is missing, and nobody can interrupt anyone, so the conversation stops being a conversation.
Mix-minus is not the fast version of muting. It is a different thing: the guest's channel is simply never routed into the guest's own return, permanently. Nobody has to do anything during the show. The guest can interrupt, the host can interrupt, both are heard by the audience, and neither hears themselves. Once you have set it up you can forget it exists, which is the actual point.
Headphones are part of the design, not an accessory
Any microphone in a room where loudspeakers are playing the remote guest will send that guest back to themselves through the air. No routing solves this, because the loop is closing outside your equipment.
This is why every broadcast setup that involves a remote participant puts the local people on headphones. It is not a preference or a professional affectation. It is the only way to keep the loudspeaker path out of the microphone path. If you must have speakers on — a room audience that needs to hear the guest — then the microphones in that room have to be part of the subtraction for the guest's bus too, and you should expect to spend real time on speaker placement and microphone directionality. Headphones are dramatically easier.
Why the guest's own echo cancellation is not the answer
Every video-call application has echo cancellation built in, and it is genuinely good. So a reasonable question is why any of this is necessary — why not let the guest's software deal with it.
Because echo cancellation is a repair, and it works by removing something it recognises. It listens to what is being played out of the speakers, looks for that same signal arriving at the microphone, and subtracts an estimate of it. When the estimate is good the result is clean. When it is not — because the room is reverberant, because the volume changed, because two people spoke at once, because the signal was processed on its way through your equipment and no longer matches what the canceller remembers — the residue comes through. And crucially, the canceller has to hold back the microphone slightly while it works, which is why call software clips the first syllable when two people start talking together.
Mix-minus does not repair anything, because there is nothing to repair. The guest's signal is not in the guest's return at all, so there is no echo to cancel and no reason for anything to duck. This is why professional setups do the routing properly even when every participant is on a platform with excellent cancellation: the repair is a fallback for situations you cannot control, not a substitute for a correct signal path in a situation you can.
The practical consequence is that you should still ask your guest to wear headphones even though their software would probably cope. It costs them nothing and it removes the last uncontrolled loop in the chain.
Setting it up, in the order that works
Start from the table, not from the equipment. Write down every listening position and what each must hear, before you touch anything.
Then build one bus at a time and test each one before moving on. Send the host bus, put on the headphones, and confirm you hear the guest and not yourself. Then build the guest bus, and have the guest confirm the same thing — ask them directly, because they are the only person who can hear their own return.
Test with two people talking at once. Almost every routing fault is invisible when only one person speaks at a time, and obvious the moment two do. If your test is a polite alternating conversation, you have not tested it.
Test the recording and the stream output separately and last. They are two more listening positions, and they are the two nobody checks until afterwards, when it is too late to fix.
What to check when it goes wrong mid-show
Ask the guest what they hear, in those words. Not "is the audio ok" — ask what they hear. The answer distinguishes every fault above in one sentence, and it is the fastest diagnostic available to you.
Then work backwards through exactly one bus. The temptation under pressure is to start changing levels everywhere; resist it, because a mix-minus fault is always a routing fault in one specific destination, and the other destinations were fine a minute ago.
If you cannot find it in a minute, mute the guest's return entirely and continue. A guest who hears nothing but can still be heard is workable for a few minutes. A guest who hears themselves is not.
Frequently Asked Questions
What is mix-minus, in one sentence?
A mix-minus is an audio feed that contains everything except the signal of the person it is being sent to. Each remote participant gets their own, so nobody ever hears their own voice returned to them.
Why does my remote guest hear themselves?
Because the audio you send back to them contains their own microphone. You are almost certainly sending them the same complete mix that your viewers get. Build them a separate feed with their own channel removed.
Is muting the guest while I speak the same thing?
No. Muting depends on an operator reacting in time and it prevents anyone from interrupting anyone. Mix-minus removes the guest's channel from the guest's own return permanently, so the conversation stays natural and no one has to do anything during the show.
Why does everyone in broadcast wear headphones?
Because a loudspeaker playing the remote guest, in a room with an open microphone, sends the guest back to themselves through the air. That loop closes outside your equipment, so no routing can break it. Headphones remove the loudspeaker from the path.
The sound is in my recording but missing from the stream. Is that a mix-minus problem?
It is the same class of problem: the recording and the stream are two different listening positions fed from two different buses, and a source was added to one and not the other. Check every destination against a written list of what it should contain, before going live.
dcast Team
Professional video streaming experts helping creators succeed.
Related Articles
Start Your Video Business Today
Join thousands of creators monetizing their content with DCAST.
Get Started Free



