The Internet Was Stable. The Live Interview Still Sounded Broken.

At 8:55 AM in a Nairobi newsroom, the internet test is green. The presenter hears the remote guest clearly. Then the interview goes live and viewers report an echo, a delayed voice, and words disappearing under the music bed.

The first instinct is to blame bandwidth. Often, that is the wrong diagnosis. Hybrid audio fails when the same signal is sent down two paths. The guest hears the programme mix returning from the studio, then hears their own voice again through the call delay. Or the streaming computer captures the full monitor mix instead of a clean broadcast mix, so music, talkback, and headphone spill all go out together.

The reliable fix is to separate the mixes. The presenter and producer need a monitor mix. The remote guest needs a mix-minus feed, which contains the programme but leaves out the guest's own return. The stream needs its own controlled output, with the microphone, playback, and remote caller balanced for the audience rather than for the control room. Headphones should be used for every remote contributor, and the full chain should be tested with a private or unlisted stream before the red light comes on.

Do not just ask, “Can we hear the guest?” Ask three different questions: Can the guest hear only what they need? Is the stream receiving one clean programme mix? Can the producer cut a source without changing the presenter's headphones? Those answers reveal the routing problem before the audience does.

Your five-minute newsroom check today: send a test call to the usual remote guest, record thirty seconds, and listen for an echo after the guest stops speaking. If the echo arrives late, you have found a return-path problem, not a weak internet connection.

Good live audio is not rescued by guessing at the last minute. It is built by teaching every signal where to go before the broadcast begins.