Doing a cleanup on the girlysound recordings

[TrampolineFrSpace]

I was hoping that we'd get some more cleaned up tracks.  I guess that I'll have to look for some audio tools.

[harkrider]

I think that Ken is the one with the grandiose plans to clean up Girlysound, being the keeper of the lossless virgins--er versions.   What news, Ken?

[DoctorM]

I love the idea of these cleaned up... but if they're not from an uncompressed source it seems to me to be a bit of a time waster.

Heck the tracks are CBR 192kbps for crying out loud.  I'd suspend disbelief if they were at least encoded with lame using --alt-preset standard or extreme.

Care was not given to the original mp3 compression.
Please make a lossless version available.  If not to the general community then at least to someone capable and willing to do a clean up.

[avj]

Just as I said I wouldn't do, I went and disappeared into the ether.

I was just discouraged after spending so much time and effort and having a crappy end result. I echo DoctorM's sentiments, and I am eagerly waiting the opportunity to work on some non-lossy tracks.

[TrampolineFrSpace]

What's the chance that we could get flac encoded versions and get them uploaded to the top site?

[TrampolineFrSpace]

If and when the original wav files created from the casette tapes are made available, the girlysound site can host the original wav files or possibly the FLAC encoded (lossless) equivalents.

There isn't anything from this side preventing them from being made available.

I confirmed that this week with the man.

[avj]

Excellent. The waiting shall continue!

[harkrider]

I confirmed that this week with the man.

You talked to Ken?

[TrampolineFrSpace]

I confirmed that this week with the man.

You talked to Ken?

No, you're the man. 

I did post something in the girlysound thread over at LPF, but there was no response.

I'd love for Ken to make it available.

[jackstpaul]

What sort of software are you using? What sort of noise-reduction do you use, i.e. plugins? Method?

I use cooleditpro (now Adobe Audition) and have it's noise reduction and separate hiss reduction processes, but use Sonicfoundry's noise-reduction plugin--generally regarded as the best.

Are you sampling some of the true silent space and then using that as the noise filter of what to remove, or are you just using some pre-set filter? The former is far better to use as you can isolate actual noise like machine, ambient, etc. rather than pre-set parameters.

No matter what you do with any editing using effects, as NR is, you end up slightly distorting the sound--beyond what the effect is supposed to do. So you need to be very judicious with using noise/hiss reduction. You cannot remove it entirely--or don't want to--because you'll be removing some of the performance sound itself. The more you use, the greater the artifacts that you end up with. It's a balancing act requiring a lot of close monitoring.

The stand-alone hiss reduction that I have and have seen elsewhere isn’t based on sampling, though sonicfoundry has hiss reduction as part of its noise reduction function.

You can apply click/pop elimination and clipping restoration separately and for what I’ve done with my cooleditpro using the click/pop eliminator hasn't made any problems, but the Sounicfoundry type has caused problems if used too much of a degree.

What resolution are you using?--contra sampling kps? You won’t get an original source higher than 16k from almost any format. Cooleditpro can't save things at the 32 bit level, but it performs edits/effects at the 32 bit level, so you want to convert the source you're using to 32 level when you do edits lest you lose some sound or create some more artifacts because of the applying 32 bit edits to 16 source. You need to convert back to 16 bit in order to save because the program can't save at 32 and you'll end up losing some sound or + artifacts form the loss incurred by going from 32 to 16 in such a forced rather than natural manner. I'm not sure that better programs like protools home or Qbase or other top-shelf programs will get you beyond the 16 to conversion issue, and I’m sure they have a conversion resolution function.

At any rate (no pun intended) you'll end up with some variation by converting to and back from mp3--or other format, hence greater are needed in performing any noise or hiss reduction.

A good test of the outcome of the revisions you make is to listen to an isolated voice speaking (or singing) rather than to a musical portion where there’s far more going on re: frequencies, etc. and it’s harder to isolate the after-effects. Of course, you want to listen to the other parts, too.

One major problem you’ll get by doing on a mix—as is an mp3—is that the individual tracks that go into the mix will each have their own noise profile, but you can’t get at them individually without having the individual, original track rather than a mixdown, so that can limit the ability to reduce noise/hiss reasonably, e.g. with 4 original tracks you can get 4 different types of machines noise, varied hiss, different ambient noise, etc. Different noise from other electrical objects nearby. Of course, if one of the 4 tracks is a mix itself, the problem is all the greater.

EQ effects are a tricking thing, and should be done after using noise reduction, lest you possibly make some of the noise worse and more difficult to remove. But using the equipment for recording as I’ve heard for the GS/Juvenalia will really distort sound re: frequencies, almost always involving a failure to capture higher frequencies in a truly representative way. What I’ve done in general with EQ’ing taped stuff-cassette and 4track, is to boost the -mid-to-low of high ranges a bit 1400-2800—since that’s where the voice and rhythm guitar ranges are most prominent, and slightly reducing the lowest of the lows. As long as there aren’t higher freq things like highnote lead guitar or piano—or something very different like a harmonica or flute that’s the focus. More to the higher of that range and maybe a bit higher to boost the voice. But Liz doesn’t sing very high so you don’t need go a whoel lot higher than that.

There are other specific types of EQ to use, e.g. parametric etc. But that gets very fancy. To boost a particular part—like the beginning without doing it in isolation, you can use envelope filters set to boost the given section as part of the flow, with smooth transition, rather than do one part and then let the rest stay as is. The transition period is really important to have.

That said, I think enough cleaning can be done without altering the sound that is desired to be left, the performance themselves, but it needs to be done very carefully, not too greatly, and using sampled noise from the track itself rather than a preset noise profile.

[harkrider]

There was an attempt to produce a less noisy version of Hello Sailor about a year ago.  I'm not sure about the software or method used, but the info is buried in one of these threads, as well as a link to the file.  The guy who did it logs in periodically.  He'll probably respond to your technical questions. 

I think it's a worthwhile project, and will help to expand GS's appeal.  Some people have a hard time getting past the tape hiss and whatnot, which is too bad, because the same people (if forced to listen, in a rental car that I'm driving, for instance) will often warm up to the music.  But I don't know that the youtube generation would wait that long.  Preserving the silences before songs is also a hotly debated topic.  Listen to the opening of White Babies, for instance.  Dyed in the wool GS fans would object to cutting the "dead air" sections because they contain such artifacts.

My suggestion would be to focus on a particular song.  Stratford-on-guy (my version hosted on this site at least) is in particularly bad shape with respect to noise.  The tape 1 and 2 material, though, came from 1st generation DAT, courtesy of Ken Lee & Gevaudan, and sounds pretty good already.  Still, removal of tape hiss would be interesting.  I would propose hosting the "post-processed" versions on the front page, alongside the "originals" that are already available.  I'll be happy to give you FTP access for easy upload.  It would also be nice to reduce the file sizes.  Those MP3 files are pretty big.

From my personal perspective, your interest in post-production GS is quite welcome.  Like I said, I think that it will only broaden its appeal to people who would otherwise not listen at all.  The idea that portions of the tracks might be re-used in creative ways?  Well, that's also pretty interesting, and would only prove what most of us believe -- that the stuff is culturally influential, and somewhat viral in nature.   In other words, I say go for it, and let's see how it sounds.

[jackstpaul]

acidmouse, how big are the files?
 
I have done a ton of this in the last 1.5 years on my own stuff. I had a musical partner adn we started working on stuff and recording on a boombox just so we could listen and gauge where we were at. Then we started working on a 4-track--just barely--and then he died. We did just 2 vocal tracks and 1 acoustic gtr track of one song of mine. Then another of 1 acoustic track and a scratch vocal--he f'd up the words. We did a basic guitar and 1 or 2 vocal tracks for 3 of his songs.

Left with hours of cassette stuff, I years later went back to it to clean it up. I did noise cleaning, hiss cleaning, some clicks, a handful of clips, and I did some EQ'ing. Everyone says you shouldn't try to make a stereo out of mono, but I tried anyway. Boosted lows on one side and highs on another side. Tricky stuff but it added to it. Putting a very, very tiny delay between the two sides made for a much better, open sound.

Then I took the one 4-track song of mine where I had 2 different vocal tracks and started music from scratch b/c I wanted a full-blown sound for the song--as intended. I made 13 music tracks, gtr, keys, synth, drum, bass and imported the 2 vocals from the 4 track. From those 2 tracks I made 10 vocal tracks; I do separate versions of dry, chorus, and chorus+reverb for each, and made 2 backup vocals of the two originals by screwing around with it. Did my other song using new instrument tracks using his chorus and bridge vocals (he messed up the verses), sang the verses myself, and also did the bridge along with him. I was able to see to fruition--minimally--what we intended to do together. He was a great singer; I'm a Neanderthal singer.

From all of that I got a ton of experience with editing, recording, and particularly cleaning, so it just struck me as natural to try it with GS or any other very lo-fi stuff.

Merely cleaning hiss is very easy to do and beneficial and if done to right degree, won't harm the original source in any noticeable way.

For what I was talking about on the LPF is best thought of--though I didn't use the examples--of Tom's Diner. The original is great, and so is they did with the remake using the original vocals I just always have wondered what the early stuff could sound like if it was enhanced in some ways; what would she have done if she'd had the ability or had people to play the extra stuff for her. Or to have more vocal tracks.

It's not about destruction of what exists, just extrapolation (and cleaning) in new creative ways. Since I've done it so much already, it's not something radical for me to conceive.

Let me know about the file size and where to find. I can whip a simple test version off pretty quickly.

[harkrider]

The MP3 files average about 7MB each.  While this is not too big for the server, I'm sure that it discourages downloads to some extent.  I'm sure that the files could average 3MB each without loss of quality.  I don't have the non-lossy versions.  I would make them available if I had them, along with any other source materials.  The girlysound server can support several terabytes of bandwidth per month, no problem.  But compact versions would make downloading easier for the masses.

Tapes one and two -- http://girlysound.com
Tape three -- http://girlysound.com/tape3.html

The "cleaner" versions of hello sailor is here -- http://girlysound.com/new

Now let's talk about the "sampling" issue.  I've personally got no problem with people getting creative.  Phair borrowed from other artists when recording GS.  Some was done in parody, some as homage.  Regardless, she's never been shy about "fair use" of other's art.  Your Tom's Diner example is a good one, and what about the Beastie Boys?  I get the feeling that Liz wouldn't mind being sampled by those freaks, or anyone, for that matter.

Go for it, man.  That's what I say.

[jackstpaul]

 Just started checking into the girlysounds audio issues. I decided to start with Fuck and Run b/c Hello Sailor is too long. Some questions and help needed

My plan is to download the mp3's and also to record the audio as it streams and comaprign the efforts. The latter for the purposes of seeing if capturing the sound initially as a wav on my end might/might not be better than converting from mp3 to wav. I'll screw around with both. But though I have dial up, it's easier to download since I have a downloader program, but capturing the streaming audio is a drag because it takes so long to fully load,

QUESTION: Can someone tell me the lengths of the songs themselves so that I can get going on shorter songs where I can work all through a song contra working on a longer song taking all the much longer to get all the way to test example. Appreciate it.

 Along that same line, I didn't listen to the two test versions of Hello Sailor yet, but I’ll download those.

QUESTION: I’m not really familiar with usign the mp3 format—Any recommendations on conversion programs? I can’t open mp3’s in my audio software. I’ll convert to wav (or capture as wav), edit, and then….we’ll see. Maybe any converting should be done by someone with good conversion programs. Advice needed.

QUESTION Fuck and Run: Where was this recorded and how? The left and right channel wavs have slight visual differences, but if it's mono split in two there can be little weird differences in appearance. Sounds like a mono original track recorded live.

Was this a tape deck or 4 track?

QUESTION: Where was it recorded? It doesn't sound like a normal, furnished room. It sounds like a bigger, empty space--a distinct echo/reverb. Did Liz record in the bathroom?--as people sometimes used to do more so in the days of tape--I did

QUESTION: What sort of commonalities are there to the recording processes of the GS songs? I’ve heard of cassettes and 4-tracks. Sometimes people record on a 4 track but on one track with a microphone in the middle of everything.  I’m curious because the noise profiles for songs recorded with the same equipment under similar circumstances can be very similar or in partially so in some very specific way.

Anything anyone can tell me about the recording conditions for the songs—as best known—can help.

For Fuck and Run:
There's some very ugly electric static noise--which can be hard to get rid of, but one of my two noise tools does better than expected on this sort of noise.

Interestingly--and I didn't already have Girlysound so close listening to it is new to me--Fuck and Run has the standard ugly noise, then something gets turned on or plugged in and the noise gets far worse as another kind is added to the mix, then there's a tap-tap that seems non-deliberate, yet synchronous like something swaying back and forth--like a button on a piece of clothing as the person moves (against a guitar, for instance). It's two taps, then a sniffle, then a light tap. Maybe a count-in with the sniffle at the time of the 3rd tap?

Otherwise there’s about 8 seconds of noise until she starts—she’s remarkably quiet aside from the tap-tap thing. No moving around, etc.

 

[Surflord]

Jesabell can fly high hi.