|
|
|
10-25-2016, 01:32 PM
|
#1
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Speed correction restoration of analog tape source
Does anyone doing this kind of stuff know of a way to tie pitch/speed expansion/compression to a reference pitch from the recording in a closed loop? Maybe the bias frequency (if captured) or possibly a tuned instrument in the audio range for the reference?
I remember reading about an app called Capstan that was meant to do this. It's price tag was shocking and the reviews on its performance were extremely poor.
Is there a rocket scientist around here with a magical script or something?
I can point to a tone in the spectral view in iZotope and say "Computer, adjust the speed/pitch so that line is flat." But of course this isn't Star Trek...
|
|
|
10-25-2016, 01:48 PM
|
#2
|
Human being with feelings
Join Date: Aug 2011
Location: Great Lakes, N. A.
Posts: 1,872
|
Have you tried asking Siri?
|
|
|
10-25-2016, 02:53 PM
|
#3
|
Human being with feelings
Join Date: May 2016
Posts: 369
|
melodyne can do it , capstan is made for it, both from the same company "celemony"
|
|
|
10-25-2016, 03:31 PM
|
#4
|
Human being with feelings
Join Date: Jan 2010
Location: Kalispell
Posts: 14,745
|
Humm, there might be some software to do this. In the old tape days it took two sources (tape decks) to do this and it worked very well.
But that worked with the phase between the two sources, which you adjusted by varying the speed of the capstan of one or both tape decks. Most professional tape decks had servo motors, so the varying had to be speed adjustments of the servo motors.
Of course in this case you'd have to do this through out the duration of the song.
wouldn't you be better of just using tempo mapping?
Quote:
Does anyone doing this kind of stuff know of a way to tie pitch/speed expansion/compression to a reference pitch from the recording in a closed loop? Maybe the bias frequency (if captured) or possibly a tuned instrument in the audio range for the reference?
|
Aah, as I reread your post, are you trying to correct the varying speed of the tape to keep the instrument(s) in pitch?
Heh heh, I'll be extremely interested in what you come up with.
|
|
|
10-25-2016, 03:35 PM
|
#5
|
Human being with feelings
Join Date: Nov 2011
Posts: 3,409
|
|
|
|
10-25-2016, 03:41 PM
|
#6
|
Human being with feelings
Join Date: Nov 2011
Posts: 3,409
|
Capstan is $4458. *cough cough*. IIRC Plagent is in the "if you have to ask you can't afford it" department? Cedar doesn't list a price on their site...
Seems like a dedicated/clever developer could come up with something in the <$500 range and make a killing. (I'm sure it's no simple thing to implement.)
|
|
|
10-25-2016, 03:42 PM
|
#7
|
Human being with feelings
Join Date: Nov 2011
Posts: 3,409
|
Maybe Xenakios could make a new version of Hourglass that does this. :-)
|
|
|
10-25-2016, 03:59 PM
|
#8
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by Tod
Humm, there might be some software to do this. In the old tape days it took two sources (tape decks) to do this and it worked very well.
But that worked with the phase between the two sources, which you adjusted by varying the speed of the capstan of one or both tape decks. Most professional tape decks had servo motors, so the varying had to be speed adjustments of the servo motors.
Of course in this case you'd have to do this through out the duration of the song.
wouldn't you be better of just using tempo mapping?
Aah, as I reread your post, are you trying to correct the varying speed of the tape to keep the instrument(s) in pitch?
Heh heh, I'll be extremely interested in what you come up with.
|
Correcting wow/flutter transport fail issues in a closed loop using a reference from the recording, yes.
Not necessarily keeping an instrument in tune. Restoring the recording to the point before transport related issues altered bits of it. Especially moments where the tape snagged/dragged for a bit creating a moment of gross fluctuation.
Long stretches of incorrect speed are a no-brainer to correct by ear with Elastique Pro manually in Reaper. One or two trouble spots are still no problem. EP being integrated into Reaper makes this stunningly quick actually. Dozens of them... makes me think of better uses for my time!
Thanks for the responses so far everyone.
Looks like the same big price players as a few years ago.
Quote:
Originally Posted by clepsydrae
Capstan is $4458. *cough cough*. IIRC Plagent is in the "if you have to ask you can't afford it" department? Cedar doesn't list a price on their site...
|
Yep... More than I've paid for all the Macs I've ever owned put together!
I was hoping something new had come along. I've seen some positive reviews for some of these. Plangent Process and Cedar Cambridge look pretty serious. But I've also seen scathingly bad reviews (especially for Capstan) that make me pause.
I could see getting into this kind of work more professionally. It's strangely rewarding for me. Not sure there's enough work out there to justify investing in some of the above mentioned tools though.
|
|
|
10-25-2016, 04:03 PM
|
#9
|
Human being with feelings
Join Date: Nov 2011
Posts: 3,409
|
Quote:
Originally Posted by serr
I could see getting into this kind of work more professionally. It's strangely rewarding for me.
|
More generally, I love doing audio restoration. (Especially now that I have RX 5. :-) ) It's like the early days of photoshop when you could so easily amaze your friends.
|
|
|
10-25-2016, 04:09 PM
|
#10
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by clepsydrae
More generally, I love doing audio restoration. (Especially now that I have RX 5. :-) ) It's like the early days of photoshop when you could so easily amaze your friends.
|
Spectral editing with RX is truly a magic trick!
Their broadband NR is pretty useful too (whereas no one elses is even close). At least when you split the program up into different frequency range bands and go after each band separately.
|
|
|
10-25-2016, 04:27 PM
|
#11
|
Human being with feelings
Join Date: Nov 2011
Posts: 3,409
|
Yeah -- there are a few UI things that are really irritating that i'm growing to grumble more about as i use the program... e.g. you can select a frequency band you want to work on (e.g. a constant ring throughout the recording, say a fan noise or something), but this only works on shorter recordings. Meaning, if I have an hour-long piece of audio and want to remove the ring from it, i have to chop it up into pieces or something, which is a pain in the ass, and avoiding that pain is the whole reason someone pays for something like RX 5 in the first place. And it just doesn't seem like there's any reason for it (in that use case). I understand that the under-the-hood details are complicated, but still.
Or: you can adjust the volume of the signal going out to your interface, but you have to open up the settings dialog every time in order to do it... seems like a really weird oversight.
Quote:
Originally Posted by serr
Their broadband NR is pretty useful too (whereas no one elses is even close). At least when you split the program up into different frequency range bands and go after each band separately.
|
Interesting -- what method do you use to do the decompose/recompose? RX 5 itself? I'd be worried about artifacts in that process outweighing the benefits vs. NR on the audio as a whole -- but no, it's better?
|
|
|
10-25-2016, 04:43 PM
|
#12
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by clepsydrae
Yeah -- there are a few UI things that are really irritating that i'm growing to grumble more about as i use the program... e.g. you can select a frequency band you want to work on (e.g. a constant ring throughout the recording, say a fan noise or something), but this only works on shorter recordings. Meaning, if I have an hour-long piece of audio and want to remove the ring from it, i have to chop it up into pieces or something, which is a pain in the ass, and avoiding that pain is the whole reason someone pays for something like RX 5 in the first place. And it just doesn't seem like there's any reason for it (in that use case). I understand that the under-the-hood details are complicated, but still.
Or: you can adjust the volume of the signal going out to your interface, but you have to open up the settings dialog every time in order to do it... seems like a really weird oversight.
Interesting -- what method do you use to do the decompose/recompose? RX 5 itself? I'd be worried about artifacts in that process outweighing the benefits vs. NR on the audio as a whole -- but no, it's better?
|
I use the NR plugin in Reaper.
The only thing I use the RX app for is the spectral editor as it's the only choice.
I usually use the Waves linear phase eq's or linear phase crossover eq's in their multiband comp for band splitting.
Then wrestle with it until I'm only removing noise and not adding any via artifacts. No one wants wind chimes added to the recording! Well, maybe Enya...
|
|
|
10-25-2016, 04:50 PM
|
#13
|
Human being with feelings
Join Date: Nov 2011
Posts: 3,409
|
Are there are any linear-phase crossover-oriented JS or VSTs, or must one always do that by ear? It'd be nice to have a close-to-mathematically-perfect decompose/recompose flow. (I assume the band splitter JS is not linear-phase? I suppose i don't have to worry about that, though, as long as the bands recompose properly.)
I also wonder a lot about the phase implications of NR plugins and RX 5.
|
|
|
10-25-2016, 05:02 PM
|
#14
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
I demo'd the JS FFT band splitter and it was artifact city. Nowhere near nulling with the original when just split and recombined. I've had the Waves plugins forever now so I keep using them.
I do null tests often to keep an eye on things as it comes up.
Reaper's editing abilities let me "get my hands right on the audio". Best audio editor I've ever used. You can zoom all the way from 12 hours across the screen to a 96k sample being 1/4" long all in the same screen without the hassle of having to open up alternate editor windows at the sample level (which is a hard requirement for a DAW for me at this point).
I'm addicted to parallel processing. It's like cheating! I'll build things in layers. So I'm pretty on top of phase and alignment.
I often have a layer of the original audio in the mix (even if buried low) with various NR'd copies and there are no phase issues introduced.
|
|
|
10-25-2016, 05:02 PM
|
#15
|
Human being with feelings
Join Date: Mar 2007
Location: I'm in a barn
Posts: 4,467
|
can you recapture it? Likely not, but if you have the tape, and a machine...
(take all of this with a grain of salt as I'm not a "pro" and I've only worked with cassette and 1/4in 2 tracks of dubious quality - but actually I guess eeking percentages of quality out of such equipment is tangible experience that "pros" may not have)
Capture it at the highest speed available on the machine - and repitch it down digitally. (this should be a lossless process or at least minimally so) If this changes the bass response in a bad way for the material, maybe skip this step.
Capture half of it backwards, and flip it in the DAW. (completely lossless) The wow and flutter is at it's worst with a full load of tape on the motor. Start in the middle and work your way out.
Good tape machines may not suffer these issues. So maybe not worth the effort to recapture.
|
|
|
10-25-2016, 05:16 PM
|
#16
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by James HE
can you recapture it? Likely not, but if you have the tape, and a machine...
(take all of this with a grain of salt as I'm not a "pro" and I've only worked with cassette and 1/4in 2 tracks of dubious quality - but actually I guess eeking percentages of quality out of such equipment is tangible experience that "pros" may not have)
Capture it at the highest speed available on the machine - and repitch it down digitally. (this should be a lossless process or at least minimally so) If this changes the bass response in a bad way for the material, maybe skip this step.
Capture half of it backwards, and flip it in the DAW. (completely lossless) The wow and flutter is at it's worst with a full load of tape on the motor. Start in the middle and work your way out.
Good tape machines may not suffer these issues. So maybe not worth the effort to recapture.
|
If this was a case of having an analog tape source, I wouldn't use a malfunctioning deck. I'm talking about recordings that may have multiple analog generations (each introducing unique mayhem). The best surviving tape is then transferred to digital properly with a functioning and calibrated deck.
I at least dial in the speed and azimuth on the deck for the transfer. Or as close to the average as possible if everything is already a moving target from a previous generation.
Speed/pitch expansion/compression linked (ie varispeed mode) is fully lossless now with Elastique Pro, so digital speed correction is less of an issue.
But the moving target stuff...
The original signal is right there! I can see it. I can hear it. I can look at different frequency components in spectral view and see plenty of landmarks. I want to just be able to draw an outline around a target frequency line and tell it "make that flat"!
But... "Buy our software to do that for only the same price as 14 Mac Pro machines!" Um... NO!
|
|
|
10-25-2016, 05:44 PM
|
#17
|
Human being with feelings
Join Date: Jun 2011
Location: Belgium
Posts: 5,246
|
Is the wow and flutter on the tape? Or is it your tape machine?
I suppose it is on the tape already, otherwise it would be fairly easy to fix.
I also think this is a real niche market. How many tapes still get transferred and when do the producers care enough about wow and flutter to have it professionally corrected?
__________________
In a time of deceit telling the truth is a revolutionary act.
George Orwell
|
|
|
10-25-2016, 05:52 PM
|
#18
|
Human being with feelings
Join Date: Dec 2011
Posts: 2,161
|
For a restoration project that required pitch, wow & flutter correction ... we rented the CAPSTAN software for this assignment.
CAPSTAN is designed specifically for this. I have RX [great program], but would not want to attempt these type of corrections.
We only had a few days with CAPSTAN. I spent the first 3/4 of time learning and experimenting on how best to use it.
As powerful as CAPSTAN is, there is plenty of User decisions that go along with it.
PITCH correcting a 2-mix is auditory fatiguing. I had a pitch pipe handy to help me keep my pitch reference. Remember ... with WOW/Flutter, the entire track is shifting around, which means your internal reference can get blurred.
I found I had to listen to the track playing in the background at lower volume levels [often in AURATONES]. Then when I heard a problem, walk over and make a correction.
Sonic results. Yes ... any form of pitch/timing changes will affect the sound quality and/or the spectral output. Of course, the CAPSTAN track was still needed EQ corrections in this restoration.
The bottom line. The UN-corrected track was totally UN-usable. The corrected track was a major, major improvement, and was approved for release.
Personally, I wish I had more time with the software. Every correction made, does expose problems that were more subtle ... it is a process of elimination/correction.
This one track was extremely damaged in the transfer, with major pitch drops and drifting, and basically a total loss. With CAPSTAN, we pulled off something just slightly under a miracle.
The financial investment in CAPSTAN is significant. The software is not as user friendly for a first time user as one would hope. Learning the work flow is one thing. With severe corrections, the 'automatic correction' won't get it right and requires user intervention. Adjusting the amount of correction is the Art part of the Science.
Lastly ... I do get tape transfers to digital for Restoration/Mastering. I did have a short time to run several tracks of other taped projects that played without noticeable pitch errors.
CAPSTAN did identify slight Wow/Flutter errors, and the automatic mode did tighten this up.
5 day rental for $200, or outright purchase $4500. Alex Tribet.
|
|
|
10-25-2016, 05:54 PM
|
#19
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by cyrano
Is the wow and flutter on the tape? Or is it your tape machine?
|
Already recorded. Original tapes lost.
Quote:
Originally Posted by cyrano
I suppose it is on the tape already, otherwise it would be fairly easy to fix.
|
Yep. I'd repair the deck in that case first.
Quote:
Originally Posted by cyrano
I also think this is a real niche market. How many tapes still get transferred and when do the producers care enough about wow and flutter to have it professionally corrected?
|
Yeah, there's no way I'd ever be able to recover the investment of any of the above mentioned products.
|
|
|
10-26-2016, 07:15 AM
|
#20
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
Unsure of an easy way to do the analysis, but for purposes of correction, you could use REAPER's playspeed envelope (intensively manually tweaking it until everything is in tune).
|
|
|
10-26-2016, 08:06 AM
|
#21
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by Justin
Unsure of an easy way to do the analysis, but for purposes of correction, you could use REAPER's playspeed envelope (intensively manually tweaking it until everything is in tune).
|
I hadn't considered that. Thanks Justin!
Yeah, this one's a labor of love for sure.
I still need to finish the learning curve with the new features for stretch markers and variable rates...
Do I understand correctly that either approach is a lossless veri-speed move? (Using Elastique Pro in linked pitch/time "veri-speed" mode.)
Hmmm...
Slice and dice and work stretch markers or automate the playspeed...
Probably playspeed right?
Last edited by serr; 10-26-2016 at 09:34 AM.
|
|
|
10-26-2016, 09:12 AM
|
#22
|
Human being with feelings
Join Date: Jan 2011
Location: Ottawa, Canada
Posts: 293
|
As far as automatic analysis is concerned, with wow/flutter (tape transport speed variations), you would see a global change in pitches. So, if you could identify a bunch of fundamentals, and they all move together by the same amount, that could be indicative of a speed change. Only the common mode effect can be caused by flutter. Narrow the range of interest to a band around 4 Hz, and that might be a place to start.
I haven't read it (yet), but this AES paper is probably of interest: http://prestospace.org/training/images/118AES_paper.pdf
Cheers
Kris
|
|
|
10-26-2016, 09:24 AM
|
#23
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by DrFrankencopter
As far as automatic analysis is concerned, with wow/flutter (tape transport speed variations), you would see a global change in pitches. So, if you could identify a bunch of fundamentals, and they all move together by the same amount, that could be indicative of a speed change. Only the common mode effect can be caused by flutter. Narrow the range of interest to a band around 4 Hz, and that might be a place to start.
I haven't read it (yet), but this AES paper is probably of interest: http://prestospace.org/training/images/118AES_paper.pdf
Cheers
Kris
|
Yeah, that's what I'm saying! I can look at that very thing with the spectral view in iZotope and it's just RIGHT THERE staring me in the face!
Yikes! I didn't get enough sleep last night to even approach reading that.
|
|
|
10-26-2016, 09:42 AM
|
#24
|
Human being with feelings
Join Date: Jan 2011
Location: Ottawa, Canada
Posts: 293
|
Quote:
Originally Posted by serr
Yeah, that's what I'm saying! I can look at that very thing with the spectral view in iZotope and it's just RIGHT THERE staring me in the face!
|
The challenge will be with solo instruments...detecting the difference between vibrato and flutter will be a challenge. Though I think the paper mentions looking at the power line frequency (50 or 60 Hz) for variations in that. Of course the line frequency is never really constant either.
It's a pretty interesting problem. Wish I had more time to look at it.
Cheers
Kris
|
|
|
10-26-2016, 09:49 AM
|
#25
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Something that would let you "select" the audio element to key off of would be just the thing. I'm imagining marquee selecting one of the frequency 'lines' in the spectral display.
|
|
|
10-26-2016, 09:57 AM
|
#26
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
Quote:
Originally Posted by serr
I hadn't considered that. Thanks Justin!
Yeah, this one's a labor of love for sure.
I still need to finish the learning curve with the new features for stretch markers and variable rates...
Do I understand correctly that either approach is a lossless veri-speed move? (Using Elastique Pro in linked pitch/time "veri-speed" mode.)
Hmmm...
Slice and dice and work stretch markers or automate the playspeed...
Probably playspeed right?
|
Play speed envelope (adjusting speed while not preserving pitch) is exactly analogous to changing the speed of tape... so in theory adjusting the speed using that should give you artifact-free correction (aside from changes due to antialiasing filters etc)
|
|
|
10-26-2016, 10:15 AM
|
#27
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by Justin
Play speed envelope (adjusting speed while not preserving pitch) is exactly analogous to changing the speed of tape... so in theory adjusting the speed using that should give you artifact-free correction (aside from changes due to antialiasing filters etc)
|
Perfect! Thanks for the confirmation.
|
|
|
10-26-2016, 10:53 AM
|
#28
|
Human being with feelings
Join Date: Jan 2014
Posts: 5,207
|
It would be rad if ReaTune could output a 0-1 normalized decimal output representing the change in pitch occuring from the automatic correction. Then you could print this to an automation lane, copy it to the master playrate, remove ReaTune, and correct errors.
|
|
|
10-26-2016, 11:07 AM
|
#29
|
Human being with feelings
Join Date: Sep 2010
Posts: 12,562
|
Quote:
Originally Posted by Fergler
It would be rad if ReaTune could output a 0-1 normalized decimal output representing the change in pitch occuring from the automatic correction. Then you could print this to an automation lane, copy it to the master playrate, remove ReaTune, and correct errors.
|
Yes. Yes it would.
|
|
|
10-26-2016, 11:15 AM
|
#30
|
Human being with feelings
Join Date: Nov 2011
Posts: 3,409
|
Quote:
Originally Posted by DrFrankencopter
Of course the line frequency is never really constant either.
|
Read a thing somewhere about forensic audio work that uses a centrally-recorded history of the mains frequency variation to validate recordings; e.g. someone says in a court of law that "this recording was taken on Jan 4th 2012 at 2:03pm" and they can analyze the mains hum in the recording and correlate with the known mains frequencies when the recording supposedly took place. They can also use that to detect if surreptitious edits happened. (Not sure how commonly this is actually done.)
|
|
|
10-26-2016, 01:31 PM
|
#31
|
Human being with feelings
Join Date: Jun 2011
Location: Belgium
Posts: 5,246
|
For edit detection, that will usually work without a hitch.
For timing/dating, it'll only work if anomalies are present. And over here, the net frequency is so stable that you only get about one correction a month on average. Not really useful for timing/dating unless you're extremely lucky.
__________________
In a time of deceit telling the truth is a revolutionary act.
George Orwell
|
|
|
10-26-2016, 04:08 PM
|
#32
|
Human being with feelings
Join Date: Feb 2012
Location: Sweden
Posts: 1,206
|
Quote:
Originally Posted by James HE
can you recapture it? Likely not, but if you have the tape, and a machine...
|
^Yes.
I wouldn't know any other tips or ideas than what has been forth already. But if I found myself in this situation, I would do everything possible to get it as good as possible at the source. Have some expert clean the tape, and make sure the tape machine is in impeccable shape, even get a couple of alternative tape machines too, and adjust the tape heads, bias, angles to get the right sound out of the tape, try recording off the tape at all possible speeds, even focusing the toneheads to playing back one channel at a time, just make sure one would get everything out of the tape itself, before opening the digital toolbox. That should minimize the operations needed via digital tools.
Maybe obvious but .. it's the surest tips I got.
__________________
There are only two kinds of people in the world: those who entertain, and those who are absurd.
- Britney Spears
Last edited by Colox; 10-26-2016 at 04:19 PM.
|
|
|
Thread Tools |
|
Display Modes |
Linear Mode
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT -7. The time now is 11:13 PM.
|