Echo Reduction and Optimizing

Hey, I’m not new to Audacity but I’m certainly not a power user. I have been recording a podcast for a few years and have only a few audio hiccups happen from time to time. I recorded an interview today with a local VR game developer at their office and had to use their conference room. Unfortunately it wasn’t the best sounding room for echos but I had no other choice. This is the first big interview I’ve done and want to make sure I can make it sounds as good as possible. I’ve attached a raw sample from the beginning of the recording that has the voices of all 3 participants from today’s interview and as well as a sample from a previous episode of my show (after post) to give a baseline of the audio quality I generally shoot for. I use a Blue Yeti USB mic when I record more than 2 people (like in this situation). I’m not asking for miracles, I know echo is a big no no but I have to work with what I have. Any help improving this audio would be huge. Thanks in advance.

There is a free plugin called DtBlkFx which has a “contrast” effect. That turns-up the contrast on the spectrogram, so the quieter sounds, like the reverb, get quieter, and the louder sounds , like the voices, get louder…
Spectrogram of ''I am joined by'' before-after DtBlkFx (contrast increased).png
Now the bad news : DtBlkFx adds conspicuous computery processing-artifacts …

There is a weird trick you can do with Noise Reduction that might help. The two clips are Normalized and Normalized with Noise Reduction.

I normalized to -3.5 to avoid having tiny sound values to work with. Note my raw sample is louder than yours. Normalize is a pure volume change.

Select the whole clip. Do not try to “find” the echoes.
Effect > Noise Reduction: Profile.

Leave the whole clip selected and apply Noise Reduction with the values 12, 6, 6.
Normalize to bring the volume back up.

If you reduce the first value lower than 12, it will just slowly stop working. If you go up, it will start sounding like a bad cellphone.

Someone on the forum discovered you could do that to get a little relief.

That’s all I got.


How are you going to shoot something like that next time?


Thanks for the help. As for next time I’m still working out a plan for that. Maybe you could give me some pointers?

What values would you use to bring the volume back up?

As far as next time, I’m still considering different options. Any tips would be appreciated. Obviously insisting on a better room would be a good start.

I use a Normalize value of -3.5dB for almost everything. It’s good for me to pick a value and stick with it so I always use the tool the same way.

In production-speak, -3dB is the maximum peak sound value that ACX AudioBook will accept for a voice submission. It is recommended that you do all production in WAV (Microsoft) until you’re ready to submit. Unfortunately, ACX requires submissions in MP3 and not WAV. MP3 conversion can cause minor volume changes, so I settled on -3.5. Close to -3dB with little or no risk of going over.

Another common value is to Normalize everything to -1dB. That assumes you’re going to post music on-line or other entertainment work such as podcast. The -1dB value usually keeps you out of trouble when you convert to MP3 for uploading. It’s never a good idea to hit or exceed 0dB. That causes instant, obvious, permanent sound distortion.

When you said what the job was, I had to resist sinking into a corner sobbing. You hit the jackpot. An interviewer and two guests in an empty, bare room. There’s no win.

I can do better than you did with individual lavalier microphones and a sound mixer.

It will be much clearer, but the show is still going to sound like I recorded in a bathroom, and that setup will cause multiple hundreds of dollars to vanish into the microphone store.

Can you get the guests to climb into a headset microphone like they use on TED Talks? That would actually work, but it would be stunningly expensive—and you still need the sound mixer.


Like that.
Screen Shot 2016-09-13 at 11.23.32 PM.png

The more screwed up your room is, the closer you have to get the microphone to the performance. And the bigger the check you have to write. If you kill enough of these shows, you get so you can hear problems ahead of time.

It’s a lot cheaper to say, “can we record somewhere else?” Guests can be remarkably agreeable if you make it clear you want them to sound as good as possible.