I’ve recorded a series of podcast episodes in which six microphones were spaced around a large table, one microphone for each participant recording onto an individual mono channel. There’s a slight echo effect from the way that everyone’s voice gets picked up slightly in other microphones, and the echo goes away completely when I solo a speaker’s track.
Is there any way to automatically solo a track when someone is speaking in it? (I.e., always mute the tracks that aren’t the loudest?) There are a couple of moments of cross-talk, laughter, etc., that I’ll need to maintain in multiple tracks, but long periods of an individual talking sound much better when I silence the other tracks.
I’m using Audacity 2.0.3 on OSX 10.7.5. I used the .dmg to install Audacity.
Thanks, Damien–that’s a good idea. I’m running all of these straight into a six-channel digital recorder, so I’ll look for some way to simulate a noise gate. The Nyquist Noise Gate plugin, perhaps? I’m sort of a noob, so I didn’t know that process was called noise gate at all! This will be very helpful.
For the gating idea to work properly you need to gate each microphone individually rather than gating the mix.
If it is a big table, can you get the microphones closer to the people speaking?
If it is a small table, you could try recording with one PZM microphone in the middle of the table.
Or perhaps give each person a lapel mic.
Rachel Maddow is a good class on audio. She has a noise gate on her live microphone because it seems they’re broadcasting from a loading dock instead of a proper studio. You can hear the gate pumping and everything she says has a little tail on it particularly obvious when somebody is talking or making noises. They have to stop processing as effectively when they have guests. One day they had two top government administration officials on the show. This was a valuable “get” and I really wanted to hear what they were saying.
I wrote a really stern letter after the show wanting to know who thought it was a good idea to order Domino’s Pizza take-away and assemble an Ikea credenza during the interview? Where was the producer during all that or was she the one ordering pizza and dropping the tools?
I was not pleased. You could tell the administrators weren’t, either. Each time there was a loud noise or odd voice they couldn’t tell where to look.
So noise gates aren’t the answer, but they and lavaliers can help a lot. Nobody ever wants to hear this, but a completely dead recording studio helps a lot, too. The absence of echoes make post-production a lot easier. Also, and I have pictures buried here, somewhere, the NBC radio studios in Washington are arranged with directional microphones around a table such that each microphone’s dead zone (cardioid or supercardioid) points to the other guests. Between that, a dead room and the headphones, there is little or no interaction.
There’s a “rule” about keeping the distance between the voice and the microphone three times shorter than the distance between the microphone and the other voice. This rule was written for miking multiple musical instruments, but the rule is valid.
Thanks, Steve and Kozikowski. Lav mics seem to be a good approach to this. One flaw in my setup is that the mics aren’t unidirectional; they’re Shure SM58s that are doing a remarkably good job of excluding off-axis sound, all things considered. The problem with both lav and strict unidirectional mics is that the guests on my podcast are inexperienced microphone users. I might remind them of good technique at the beginning of the show, but ten minutes in they’re either crossing their arms and causing clothing noise in their lav mics, or swaying back and forth and moving in and out of axis on unidirectional mics. I’m at the point where I might be done with trying to train one-off podcast guests on microphone technique and I’m looking for a way to use a forgiving setup and then exclude its flaws in post-processing.
I’m looking for a way to use a forgiving setup and then exclude its flaws in post-processing.
Which, we admonish people every fifteen minutes, you can’t actually do. There is no “clean it up later,” and forum posts that start like that are almost universally doomed.
You have to actually shoot it. If you could clean it up later, then the newsies wouldn’t be doing this at major interviews.
The “roundtable” technique is not dreadful. You put a nice omni-directional microphone in the middle of the table and make up the differences in volume later with volume compression. Not optimal, but workable, and fails if you have a loudmouth and a lamb in the same interview.
This is a construction project I did a while ago for a videoconferencing job. That’s a Shure SM58 and it’s placed under the screen pointing back at the audience. The mic is sitting in a little hole so it doesn’t roll around.
In actual use, it’s painted black and vanishes into the dark room.
In this instance, it’s sitting on a table which in turn is sitting on a concrete floor, so there is no desk or table sound interference. 30 inches on a side means it’s large enough not to have environment effects. I have also placed one of these things with an omni-directional microphone in the middle of a conference room table and placed it on several layers of sound absorbing material like towel or polyfiber quilt filling or batting (consult your local fabric store).
This is pressure zone configuration and it doubles the sensitivity of the microphone with few if any other bad effects. Do not use a desk microphone stand. Any short distance between the microphone and the table produces odd comb effect tone problems in the show. Either suspend the microphones a good distance from the desk – and then pad the desk…
Mobile phones have forced the introduction of several special “gates”.
The algorithms can generally be summarized as VAD (Voice Activity Detection).
Its purpose is to gate all that isn’t voice. This includes background sounds, echoes (“Echo Cancellation”) and cross-talk.
The simplest VAD methods are based on the energy level of the incoming signal (similar to what you’ve proposed above).
The difference to a normal noise gate is that the parameters are set automatically (i.e. the threshold and noise profile must be estimated on the fly, based on non-voice frames).
Unfortunately, those algorithms are mostly developed for low sample rates and they give those hollow, tunnel-like kind of result.
Another possibility is to remove all common audio to two of those recordings. https://forum.audacityteam.org/t/karaoke-rotation-panning-more/30112/1
This means, you’ll combine two tracks to a stereo track and remove the center (if they are aligned properly). Then, you can split them back to mono.
Alas, this is not feasible. with 6 different microphones because this gives 15 different pairs (5+4+3+2+1), whereas we only have 1 combination for two mics and 3 for three of them.
Each week have a new guest/ person to be interviewed. It is the bane of my existence for a few reasons, many being poor mic use or in more than one situation, having NEVER used skype or a microphone before.
Having a gate patched in as an insert allows me to apply “more gate” as an effect using the Aux knob on my mic and less on the incoming audio etc. It’s a fine balance but once I got it set to be acceptable, I let it be and don’t touch the knobs!