Plugin/AI solution to eliminating crosstalk in podcast recordings

[JLeigh2024]

For years, I've been muting and fading crosstalk on multi-track podcast recordings...eliminating voices that recorded are on other tracks to reduce room noise/reverb etc. I've often experimented with gates, but just hate what happens with attack and decay. Also, nothing seems to be able to find those quiet "uh-huh" and "yeah" that happens when another voice is talking on another track. Are there any new plug-ins that use AI to analyze a particular voice on a track, and then eliminate any other voice on that track, while also providing a smooth attack and decay fade?

[tope d]

Originally Posted by JLeigh2024

For years, I've been muting and fading crosstalk on multi-track podcast recordings...eliminating voices that recorded are on other tracks to reduce room noise/reverb etc. I've often experimented with gates, but just hate what happens with attack and decay. Also, nothing seems to be able to find those quiet "uh-huh" and "yeah" that happens when another voice is talking on another track. Are there any new plug-ins that use AI to analyze a particular voice on a track, and then eliminate any other voice on that track, while also providing a smooth attack and decay fade?

Interesting!
I've used moises and spitter to remove drum mic bleeds but haven't thought of it on dialogue. Will follow this thread and see if there are already existing tools online :)

Also wondering how your mic is placed..

[JLeigh2024]

I'm just a post-production guy, usually with little to no say about how the audio is captured. Which is why I'm always spending so much time isolating voices.

[The Weed]

Are these remote people, in separate places or all together, but on different mics?

[Sardi]

Originally Posted by The Weed

Are these remote people, in separate places or all together, but on different mics?

They’d be in the same room, otherwise there would be no spill being picked up by the other mics which is what the OP is describing.

What you want is an auto mixer like the Dugan system. There are a couple out there and they work quite well. I did a bunch of tests comparing a Dugan system with these plugins and also against the automixer built into my Sound Devices kit and some of my digital mixers. They all were pretty close to the Dugan.

Depending on how low the umms/ahhs/oks are, it does a pretty good job of getting them. Far better than a gate as it’s not muting, but rebalancing and pushing levels down. But if it’s very quiet, you can miss it.

The best workflow, for me anyway, is to DeNoise and de-reverb each ISO track before feeding it into an automixer as it helps a lot, particularly if it was recorded in a less than stellar room - ie: no acoustic treatment.


Sent from my iPhone using Tapatalk

[JLeigh2024]

Exactly my situation. Appreciate the thoughtful response.

[peapodpost]

Spectralayers has an "Unmix - Multiple Voices" module [1] that is worth a shot.

[1] https://download.steinberg.net/downl...le_voices.html