I have 2 stereo-recordings of a speech/discussion event. Since the room was a circular dome, there is a lot of echo in one of the recordings (much less in the other), making the voices difficult to understand. So I want to identify the frequencies of the echo to dampen them. But I can’t figure out how to analyse and compare the spectrums of the two recordings in order to identify the echo-frequencies. Both recordings are loaded into Audacity and synchronized in time. I have tried studying Plot Spectrum - http://manual.audacityteam.org/man/plot_spectrum.html - but I can’t see how I need to use it.
(Next question would be which Effect to use to dampen the specific frequencies. )
Thank you!
(I have already improved the sound significantly with compression and noise reduction.)
(I use Audacity 2.1.3 on Windows 7 Home Premium 64-bit SP1)
The better recording is only better for the main speaker, not for the rest of the participants. So I need to switch back and forth between the 2 recordings.
I couldn’t see ANY difference. So I am wondering if I am using that analysis correctly.
Yes. That’s why I would only dampen them, not remove them completely.
Now I have installed the “VST Transient” plugin in Audacity and I have tried using it. But I am a bit puzzled as to how it can help me dampen the echo of the circular dome? Can anyone explain how to use it for that? Thank you!
I tried the SPL [u]De-Verb plug-in[/u]. It didn’t have any noticeable effect (and I had to install Reaper to make it work…) So the company advised me to try a professional tool called iZotope RX6 from https://www.izotope.com/en/products/repair-and-edit/rx/features.html . BUT why hasn’t anyone pointed out that Audacity has a built-in effect called Reverb?!? I am going to try that now! It looks quite more advanced than the SPL De-Verb plug-in(!)
I didn’t want to ‘start all over’ with a huge professional tool like iZotope RX6, so I marked about 0.15 sec. of reverb at the end of a man with a deep voice saying a word that didn’t end with ‘s’ or ‘t’ or the like, and then I used that as Noise-profile for a normal Noise Reduction with the following parameters: 12,9,3. It’s not perfect, but better than nothing. The only strange thing is that it seems to work OK until somewhere near the end of the 1h:20m recording…? Maybe someone opened a door and thus changed the resonance frequencies of the room(?)