Last month I cut an episode recorded in front of about ninety people in a rented event space. The conversation was the best that show had ever produced. The audio was the worst. Somewhere around minute forty, a phone rang two rows back, the host lost his thread, and the guest said the single most quotable line of the whole hour directly over the top of a laugh from the crowd. I spent longer on that one sentence than on the rest of the episode combined.
That is the trade. A live audience makes people better and makes the file worse. I think it is usually worth it. But hosts keep walking into live recordings expecting the edit to be the same job it was in a quiet room, and it is not the same job at all. So here is what I actually do differently, and what I wish hosts did differently before they hand me the card.
An audience changes the guest more than it changes the host
Hosts get nervous. Guests get good.
I have watched this enough times now that I treat it as a rule. Put a guest in a room with fifty strangers and they stop giving you the media-trained version of their answer. They tell the story with the timing they use at dinner. They pause for effect because a pause in front of a crowd is a real thing rather than dead air. They land jokes they would never risk into a microphone alone in a hotel room.
The same guest recorded remotely gives me a flat, careful, slightly overlong answer that I have to trim down to something with a shape. There is a whole post about that problem in why your guest answers are too long, and I stand by all of it, but I will add this: an audience fixes about half of it for free. People self-edit when they can see boredom.
Hosts go the other way. Hosts start performing. They talk over the guest to keep the energy up, they ask two questions in one breath, and they start narrating the room back to the room. Most of my cuts on a live episode are host cuts.
The audience is not the problem in a live recording. The host trying to entertain the audience is the problem in a live recording.
If you are hosting one of these, my honest advice is to pretend the crowd is not there for the first ten minutes. Do the interview you would have done anyway. The room will come to you.
The audience track is a separate instrument, not noise
The biggest mistake I see is treating crowd sound as contamination to be minimized. Then the episode goes out sounding like it was recorded in a closet with occasional ghost laughter bleeding through, which is the worst of both worlds.
Room sound is the entire reason to record live. So capture it on purpose. One dedicated audience microphone, placed in the middle of the room facing the crowd rather than the stage, gives me something I can ride up under a laugh and pull back down under a question. Without it I am stuck with whatever bled into the lapel mics, which is thin, phasey, and impossible to control independently.
What I want from a live shoot, in order of how much it saves me:
- A clean lavalier or boom on every speaker, recorded to separate tracks.
- One dedicated room or audience mic, positioned away from the speakers.
- A board feed if there is a PA, recorded as its own file, not mixed into anything else.
- A clap or a countdown at the top so I have a sync point that is not a guess.
- The house music turned off during the actual conversation, whatever the venue says.
That last one is not a technical preference. Music in the room is music in your episode forever, and neither of us can remove it after the fact.
Sync goes wrong more often on live days
Live shoots involve more devices, longer runtimes, and more people who touched record at slightly different moments. Recorders drift. Cameras stop themselves at file size limits nobody warned you about. A phone recording as backup is running on a completely different clock than everything else.
I do not fix this with vibes. I find the sharpest transient in the room, usually a clap or a door, and I align to it, then I check again at the end of the file because a match at minute one means nothing if the two files drift apart across ninety minutes. I wrote up the whole recovery process in what to do when audio desyncs, and about a third of the examples in it came from live recordings.
If you are shooting with several angles, and most live setups end up with at least a wide and two closeups, this is also where multi-cam editing stops being a luxury. Cutting a live conversation on a single locked-off wide is technically fine and emotionally flat. You want the reaction. The audience laughing is good. The guest noticing the audience laughing is better.
Prep matters more, not less
There is a real argument about whether to talk to a guest before you record at all. I have gone back and forth on it and mostly landed on the side laid out in the pre-interview call debate, which is that a short call about structure beats a long call about content.
For live, I move further in one direction. Do the structure call. Skip the content call entirely. A live audience punishes rehearsed answers harder than a microphone does. You can hear when a guest is retelling something they told the host on Tuesday, and so can two hundred people.
What the pre-call should cover for a live show is logistics that affect the edit. How long are we going? Is there a Q and A section, and if so, will you repeat the audience question before answering it? Because if you do not repeat it, I cannot use that section. A handheld mic passed into a crowd of ninety people picks up roughly nothing usable, and the answer to a question nobody can hear is not content.
Q and A is the part everyone gets wrong
I would guess two thirds of the live episodes that come across my desk have an unusable final fifteen minutes.
The pattern is always the same. Great conversation, energy high, host opens the floor, someone in the back asks a long rambling question at conversational volume from six meters away, the guest answers brilliantly, and I am left with four minutes of great audio responding to something inaudible.
Three fixes, in order of how much I like them:
- The host repeats every question, in full, into their own microphone. Free, reliable, works every time.
- A runner carries a proper handheld mic to each questioner, recorded to its own track.
- Questions are collected on cards during the break and read out by the host.
Option one costs nothing and it is the one people forget under pressure. Tape a note to the podium.
Cleaning it up without sanding it flat
Here is where I get opinionated. Automated audio processing is very good at making a live recording sound like it was not live, and that is a failure, not a success.
Run aggressive noise reduction on a room track and you get that underwater artifact where the laughter turns to hiss and then vanishes. Run heavy leveling and the crowd swells up between every sentence like the sea. The tools are not stupid, they are just optimizing for a target that is wrong for this material. I have written more about where the line sits in Auphonic vs human audio mixing, and live audio is the clearest case I know for the human side of that argument.
What actually works is boring and manual. Gentle broadband reduction on the speaker tracks only. The room mic left mostly alone. Manual rides under the moments that matter. Someone listening to the whole thing and deciding, sentence by sentence, whether a laugh is warmth or clutter. That is why human editors watch the entire episode here and nothing gets cut by silence detection. A silence detector hears a two second pause before a punchline and deletes the punchline.
Video is where live pays you back
The audio is harder. The video is a gift.
A room full of people is production value you did not have to build. Wide shots have depth. There are faces to cut to. The lighting is usually already handled because the venue handled it. Compare that to two people on a black backdrop with a key light each, which is the format I cut most weeks and which looks exactly like every other show.
Live footage also clips beautifully. A short with visible audience reaction outperforms the same joke on a static talking head, and I do not think that is close. If you are pulling verticals off a live episode, cut them so the laugh lands inside the clip rather than after it, and let the wide shot breathe for a beat at the end.
Same logic applies to packaging. A live episode gives you a genuinely different thumbnail frame, a stage, a crowd, real depth of field, instead of another headshot on a gradient. My sister studio ClickTheory does that side of it, and the live episodes are always the easy ones because there is something actually happening in the image.
What I would tell you before you book the room
If you are considering your first live recording, do not treat it as a normal episode with an audience bolted on. Treat it as a different format that happens to use the same two chairs.
Record it slightly longer than you need. Assume the first ten minutes are warmup you will cut. Get the room mic. Repeat the questions. Tell your guest nothing about what you are going to ask beyond the shape of it, because the room will do the rest of the work for you.
And remember that remote guests never get any of this. The eye contact problem I described in remote guests and the eye contact problem simply does not exist when two people are sitting three feet apart with a crowd watching. They look at each other because there is nowhere else to look.
That is the real reason I like these episodes. Not the applause. The fact that two people finally talk to each other instead of to a lens.
If you have a live episode sitting on a drive and you are not sure what to do with the room track, send it over. I would rather look at the files than guess.