Ask a hundred viewers why they left a stream and picture quality will barely feature. Audio will: too quiet, echoing, one speaker inaudible, a remote guest sounding like they're in a tunnel. Audio is most of the perceived quality of a stream, and it's usually the least-planned part of the production.
The stream needs its own mix
The mix that sounds right in a room — through a PA, with the room's own acoustics — is wrong for a stream. Room PA feeds are equalised for the space and carry every reverberation. The stream needs a dedicated mix: direct microphone feeds, VT audio, remote guests and any music, balanced for headphones and laptop speakers. On bigger shows that's a separate sound engineer with their own desk listening on the stream, not in the room.
Audio splits from the venue
When a venue's AV team runs the room, the stream takes an audio split: a copy of each microphone (or at minimum a clean pre-fade mix) delivered on a stage box or via Dante to the stream desk. Agree it in writing before the day: which sources, what connector, who provides the cable, and whether it's pre- or post-fader. "We'll give you a feed from the desk" is not specific enough.
Mix-minus: why remote guests hear themselves
A remote speaker needs to hear the show to take part in it. If you send them the full programme mix, that mix contains their own voice, delayed by the round trip — and they'll hear themselves echo a second later, which makes speaking almost impossible. Mix-minus is a return feed of everything except that person's own microphone. Every remote contributor needs their own mix-minus. Broadcast calling tools handle it automatically; DIY setups routing a Teams call into a mixer usually don't, and that's the echo you've heard on so many webinars.
Levels and loudness
Aim for consistent loudness across every source: the presenter, the pre-recorded video, the remote guest and the music bed should all sit at the same perceived level so viewers aren't reaching for the volume control. A limiter on the stream output prevents the sudden peak that distorts. Where we can, we normalise pre-recorded content in advance to a target loudness rather than fixing it live.
Microphones: the right type in the right place
Lavalier (clip-on) mics for presenters who move; headsets for anyone who turns their head a lot; handhelds for audience questions and as an emergency backup on stage; a boundary or gooseneck for a seated panel if lavs aren't possible. Fresh batteries at every break. And a rule: nobody on the stream is ever on a laptop microphone.
The three mistakes that account for most complaints
- Feeding the room PA mix to the stream — reverberant, uneven, unusable.
- No mix-minus for remote guests — echo, and a speaker who stops mid-sentence.
- Nobody listening on the stream — problems that are obvious to a viewer go unnoticed for twenty minutes in a gallery monitoring the room.
Put a competent person on headphones, listening only to the stream, for the whole show. It's the cheapest quality improvement available.