Post by: aip on November 24, 2007, 05:20:30 AM
I tried the search function to find any ideas about diminishing crowd noise on some recordings I did of live groups, and found nothing specific.
Sorry to ask here if it's inappropriate.
The venue is an Irish pub, the room approximately 22ft x 12ft, with a ceiling about 10ft, and the groups primarily play unamplified acoustic instruments, and voice. The recordings are 24/44.1, using a stereo mic and the M-Audio Microtrack, and the playing and feeling are great, hence the public is excited, and a bit too present. The goal is to slightly diminish their presence, so that with multiple listening sessions, the music remains more important than the ambience.
Am trying parallel processing with EQ/Comp/Noise removal,etc, and things are slowly progressing.
If anyone could steer me to a thread or any other source with info about this problem, I would be deeply grateful.
aip
Post by: Thomas W. Bethel on November 24, 2007, 09:41:01 AM
The human voice in right in the same spectra as the instruments and "crowd noise" is mostly people conversing, glasses clinking and a general rumble of bodies moving around. The audience conversing covers exactly the same frequencies as someone singing since they are both the human voice.
A couple of things you could try.
First use a program like Izotope's RX or some other spectrum editor and try and get out the most annoying "crowd noise" secondly if possible get a sample of just the crowd noise loop it and make it as long as the recording and reverse the phase and add it back in using a multi-track editing program like WLs Montage function. The problem is that the crowd noise is constantly changing and this is probably not going to be optimum for what you want.
I guess my real suggestion would be to use microphones that are more cardioid in nature and will block out a lot of the crowd noise or hang up a large sign saying "QUIET - RECORDING IN PROGRESS"
There is a great record album (now a CD)that was done in the 70's and recorded on a Revox Tape recorder called "Jazz in the Pawnshop" and it has a lot of crowd noise throughout the whole album but the music is what is important that that is what the ear focuses on. I personally do not find "crowd noise" objectionable unless the crowd noise is louder than the music.
I am sure others here will have other GREAT suggestions.
Best of luck!
Post by: aip on November 24, 2007, 01:01:16 PM
Thanks a lot for answering!
I am using Izotope's Rx, and it is a big help, allowing clean removal of low frequency thumps etc., and, through auto training and homeopathic (slight) noise removal, I am getting fair results. I am painfully aware that most of the "human" noise is right "smack dab in the middle" of the musical frequencies. Using Rx on a copy of the tracks, and Neodynium, and other treatments, I am then lining up the tracks in PT7, and doing a mix that will use the treated tracks to slightly boost those frequencies most pertinent to the music. The takes were a series of about 15 different nights, concerning 7 different groups, and were not intended to be a true professional recording. The ambiance is definitely part of the project, I'm just trying to tame down the crowd a bit so that the CD can be enjoyed for several listenings.
I hadn't thought of the crowd phase switch trick, but will try - although it too will mess with the music.
Not looking for miracles, just trying to give a sweet gift back to the musicians who played!
All the best, and have a pint on me!
aip
Post by: cerberus on November 24, 2007, 01:44:52 PM
which could be a strong difference factor between crowd noise and the music.
i would suggest making separate noise profiles for each song; using both the head
and tail (just before any applause); and any mid-song pauses that
are "in the clear". not the automatic training mode.
jeff dinces
Post by: bruno putzeys on November 24, 2007, 03:02:47 PM
| Thomas W. Bethel wrote on Sat, 24 November 2007 15:41 |
secondly if possible get a sample of just the crowd noise loop it and make it as long as the recording and reverse the phase and add it back in using a multi-track editing program like WLs Montage function. The problem is that the crowd noise is constantly changing and this is probably not going to be optimum for what you want. |
That's not going to work - it works with a perfectly periodic signal (e.g. hum) but crowd noise is random so the noise you're adding won't even partly cancel. It will simply add up (power wise).
Personally I'd leave the noise in. Some of the most-loved live recordings of acoustic gigs have constant crowd chatter and imho this usually adds to the atmosphere (unless the crowd isn't sympathetic with the band of course). Plus, whatever you'll take away trying to remove crowd noise will also cost you in information from the actual music. It's best to let the listeners' ears do the job.
Post by: Thomas W. Bethel on November 24, 2007, 05:48:22 PM
| Bruno Putzeys wrote on Sat, 24 November 2007 15:02 | ||
That's not going to work - it works with a perfectly periodic signal (e.g. hum) but crowd noise is random so the noise you're adding won't even partly cancel. It will simply add up (power wise). Personally I'd leave the noise in. Some of the most-loved live recordings of acoustic gigs have constant crowd chatter and imho this usually adds to the atmosphere (unless the crowd isn't sympathetic with the band of course). Plus, whatever you'll take away trying to remove crowd noise will also cost you in information from the actual music. It's best to let the listeners' ears do the job. |
GREAT SUGGESTIONS.......
The phase reversal suggestion was only that - a suggestion. I know from experience that it usually does not work but it is worth a try especially since this is a labor of love. As I said in my reply I really like crowd noise unless it is so loud that you cannot hear the band or the singer because of it. The CD of Jazz at the Pawnshop is a good example of nicely blended crowd noise and good listenable music. Listening to the CD makes you think you are there.
Post by: aip on November 25, 2007, 05:47:32 AM
Thanks for the reply.
I am using separate noise profiles on each song, usually testing several regions, and trying auto, and comparing the B and C versions of Rx treatments. I always consider each recording as a new case to explore, and very rarely superimpose a solution coming from a different song. In the case of the near 20 hours of recordings, reduced to about 75 minutes, some songs were chosen by the musicians (all friends of mine), and the performances are really good. In this stage of pre-mastering, I'm just trying to reduce "a little" the sometimes over enthusiastic crowd's presence.
It's a stimulating challenge, and anything that I learn with this project will of course help in future projects.
Izotrope's Rx is a good soft (I also have Spark XL) and further processing in Neodynium helps as well. I have the originals in the session, and when mixing in the treated tracks, I do mute/solo comparisons to stay as close as possible to the original timbre and feeling, then bringing up the treated tracks to give a little more presence to the music. EQ alone doesn't allow this, but it's a tightrope as far as avoiding denaturing the overall sound.
As I said, this is not a pro project, although these musicians are top knotch.
Thanks again to all for youd advice!!!!
aip
P.S. - I am thinking of getting a DSP card for my G5 2x2G OS10.4 that is compatible with PT7.4, and am leaning towards Waves APA, or TC Powercore. Any other ideas?