Taperssection.com
Gear / Technical Help => Post-Processing, Computer / Streaming / Internet Devices & Related Activity => Topic started by: noam on March 28, 2015, 12:41:07 PM
-
A WAV file of a staged opera with an orchestra pit, recorded from the audience - the orchestra covers the singers. Is there a software that can be taught to identify the voices and boost them alone, making them louder than the orchestra? I don't mean dynamic compression or use of a dynamic equalizer that works on identifying frequency bands - is there a software that can learn to identify sounds (in this case the human voice) based on variables other than frequency range and loudness?
Noam
-
To my knowledge there's nothing that would do this.
-
I'm going to say "no" as well. Would be cool, though. :)
-
If this could be fixed by software, it would do a lot to solve mic placement problems, but fact remains that the loudest source at the mics still wins.
-
The short answer is no, not really. The human voice is just way too complex to simply boost like that, especially with and orchestra covering most likely all those same frequencies. There are people that do that type of work for a living, but it is as much a scientific experiment as anything else. Definitely not a "click of a button" type situation
-
This can be useful for spying, to tease apart the voice of someone under surveillance from background noise. I think I've seen it in movies.
-
iZotope's response:
Hello,
Thank you for contacting us and for your interest in iZotope!
We do not really have any kind of tool that can do this. Unfortunately, the best way to do this would be to get the original multitrack recording (if available) and remix it. To my knowledge, there are no tools on the market that can do this very well currently.
While we do have products that can remove noise from music or vice versa, separating singers from orchestra is much more difficulty as they are both "tonal" sounds and take up similar frequencies.
Let me know if you have any further questions!
Sincerely,
Michael
iZotope, Inc.
http://www.izotope.com/
-
Your best bet is careful equalization to subtly emphasize the formants of the vocals, making the enunciation clearer. Try mid-Q boosts centered around 1 to 4 kHz. You will be limited in how much you can do by the tonal effect this will also have on the music, but there is probably room there for a bit of emphasis. Sometimes a sort of "W" shaped EQ curve that also boosts the extreme highs and lows as well as the midrange will help the music parts sound more tonally correct by partly off-setting the midrange vocal emphasis.
Or enlist the old-skool analog vocal extraction and enhancement services of Gene Hackman-
(http://www.graffitiwithpunctuation.net/wp-content/uploads/2012/10/The-Conversation-1.jpeg)
From The Conversation, an excellent film by Francis Ford Coppola, 1974.
-
From The Conversation, an excellent film by Francis Ford Coppola, 1974.
Excellent flick, indeed. I probably haven't watched it in a decade or two. This just might be a fine time to break it out! Thanks for the reminder, Gut.
-
Despite your email from iZotope, I think it might be worth playing around with the Spectral Repair tool in RX. It's pretty amazing what it can do, and while I don't think it will get you what you are looking for exactly, it could possibly allow you to select part of the vocal, extract that as a new item and process it and then play that separate item alongside the original. Kind of like the idea behind parallel compression. This is likely to be highly tedious for anything but a short excerpt though.
-
From The Conversation, an excellent film by Francis Ford Coppola, 1974.
Excellent flick, indeed. I probably haven't watched it in a decade or two. This just might be a fine time to break it out! Thanks for the reminder, Gut.
It's on Netflix, or at least was as of a couple weeks ago.