How to isolate voice in a 2-mic setup

[pallen]

Hey folks,

I'm producing an interview podcast. The interviews will be face-to-face in the same location. We did a test interview, using two microphones and a standalone recorder, recording on separate tracks. When I import them into Audacity, I can hear both voices on each track. (On one track, the interviewer is loud and clear, but I can still hear the interviewee faintly in the background. Vice versa for the other track.) How do I go about cleaning up each track so it only has one voice on each?

Thanks so much!

[steve]

If you want "only" one voice per channel, you need the two people in separate rooms when you record, or record each person separately (not real-time).

If "mostly" one voice per channel, physically separate the two people as much as possible, and get the mics as close as possible to the people talking. Directional microphones will help.

"Cleaning up" the recording after the event is a non-starter (as I assume you are discovering).

[Trebor]

How do I go about cleaning up each track so it only has one voice on each?

You could try squelching what should be silence, but where you can hear the other person talking ...

Audacity Envelope Tool demo.gif (113.57 KiB) Viewed 988 times

[kozikowski]

This is how Charlie Rose does it.

The "Famous Table" is not an accident. That's how the sound people keep the two performers apart. That and yes, limbo lighting is dramatic and something of a trademark for him, but the real point of the WNYC studio is it's dead quiet with no echoes. They're wearing bright, close-talking lavalier microphones and that's about as good as it gets. I assume there's still a little cross-leakage, but it's so low that it can be ignored.

So yes, the two microphones and two-channel recorder was a good idea, but I bet you shot in a live room with the actors too close and the microphones are picking up each other's wall echoes in addition to direct leakage.

You can get directional lavaliers and I enjoyed a show shot wild in New York City with them. It came off very well with the performers announcing perfectly, literally in the middle of Grand Central Station, but one of the performers tried to do something a little out of the ordinary and the track went straight into the toilet. Directional lavaliers are rough to use.

Isn't post production fun when you have cross-talk like that?

Koz

[steve]

You could try squelching what should be silence, but where you can hear the other person talking ...

and you can also automate that using the "Auto Duck" effect (http://manual.audacityteam.org/man/auto_duck.html), but whether manually or with Auto Duck you are relying on only one person talking at a time. In real life there is almost always overlap in places.

[kozikowski]

All these tricks only work when the voices are separate during the exchange. If they like to interrupt and step on each other, that's the end of the world. Audacity cannot reliably separate instruments, voices and sounds in a mixed performance.

This is another way to shoot these things.



That's a long-distance shotgun microphone used in close proximity for interviews. Ira Glass was one of the first people to do that on This American Life. Everybody looked in horror and said it would never work. Now everybody does it that way.

There's probably a Sennheiser MKH 416 inside that wind sock.

Koz

[kozikowski]

A friend of mine did a voice job on one of these in Biscayne Bay. He was on the floor of the boat with a shotgun pointing up at the performer.

Successfully.

Screen Shot 2017-05-17 at 5.55.28.png (89.03 KiB) Viewed 977 times

Koz

[pallen]

Thanks for the assistance, everyone. I still have a lot to learn.

[kozikowski]

Thanks for the assistance, everyone. I still have a lot to learn.

Field interviews have all the problems of exterior theatrical production plus multiple microphones. People who can routinely pull that off are production heroes.

Remember if you record in a restaurant you can't use anything with copyright music in the background.

About a year ago I talked to someone doing an interview here.

http://kozco.com/pix/Cathay-Pacific.jpg

I thought he was nuts. He said it was intentional and he did it for the obvious signature of the plane noise, but he was using one blast-proofed microphone in this instance (it's also windy there). Newsies use one, long-handled microphone and swap it back and forth, but they're not doing long-form interview. I'm with you. Two mics are a good idea, they just have to be managed carefully.

Koz

[The_Ratfiles]

Ok, here is how we do it at The Ratfiles Podcast, and it works as well as any studio could hope for. We have our recording studio/room fairly soundproofed. Using Behringer mics (XLR), we join the twin microphones with an XLR 'Y-cable' and then run them into a top-end laptop's USB input (using Ubuntu 14.04...works best, for some unknown reason, better than 16.04) and record using Audacity 2.1.2.

Our end result is a very nice, clean quality recording of two hosts sitting in a room 3-feet apart, without ANY reverb or cross-talk (although, we do sometimes try to talk over each other, that's etiquette, not bad recording properties). Clean-up is easy in post-processing. The only 'downside' is that since we only have one audio track (one per session, instead of per-mic, per-person), we have to get the gain fairly synchronized between the mics before we start recording - and luckily (or unluckily) we do this by mouth-to-microphone distance, which is fairly easy to do using this specific 8500 model Behringer microphone. I can get us within say, 2-3 db's of matching gain by pre-show sound-checks...and then if I need to 'tweak' post-production, I can simply go into effects, do a 'noise reduction', then some 'on timeline' gains or reductions if I have any peaks or spikes.

Now, knowing this, go check out our podcast and see what you think.

theratfiles.buzzsprout.com
theratfiles.com (homepage)

for the sake of a couple hundred dollars in mics, cables, and a decent 3.5ghz laptop with 8gb RAM and Audacity. I'm telling you, the real magic is in the Ubuntu/Linux, not the other gear...I can't get this same quality with my Windows-based server, and it is running a quad-core Intel chip and Win7Pro...it just doesn't sound the same in the final MP3 file.