OTR Cleanup

Intriguing! Sounds really interesting. I never considered separating the voice from the track. I don't know why, it seems the reasonable thing to do now that you've said it, but it never occurred to me before! Yeah, pulling the vocals out as a separate piece would result in having more control in restoring by cleaning the voice separately, specifically. Same goes for the backing track.

What made you think of doing it that way? - or is it a common method I should have been aware when I had been trying to do it?

Which episode did you work with? How bout posting the resulting MP3?

I'm not sure it's a common method.  I was just searching around the internet while pondering how to clean up some of the OTR I have.  I was thinking, wouldn't it be nice if I could somehow just extract the noise from the rest of the track, and then read about vocal extraction.

It'll post a vocal extracation from Episode 1 of Quiet Please.  There are a number of processing methods in the program, each with a bunch of models and settings.  I'm still working my way through them all, but I did notice that the results vary widely between different shows, even with identical processing.  Add in all the variations, and you can play with this for a long time.

You'll notice that the organ music is still there, but muted, in the track I posted.  In Episode 2, it was pretty much gone, but found in one of the other tracks it extracted (actuall, the 'Other' track).  However, in Episode 2, the vocal isn't nearly as clean.

Again, lots of playing around to do.  But I think with the right settings and models, this has a lot of promise.

Here is the original file.

And here is the vocals track from that file.  All default settings, just had to choose the model, one of the first I produced.

@artisan-radio That's quite good! But to edit like that there has to be a compromise like audio quality as you can't separate the noise from the audio you want to keep. You may as you are finding making the audio more muffled as you are removing some parts of the audio you don't want to.
That's my problem with these(muffled audio) especially CBS RMT. I spent time listening to 10 seconds of all I downloaded and discarded the bad ones but I kept about 300 or so that were good or good enough. All the other shows I have are not bad. I could just download the batch and even with 900 episodes of Suspense none are that bad.

Yeah I listened again and the second one is way too muffled compared to the first. That with me would be not acceptable.

The whole point of UVR is to separate the vocal track from the audio, and have one or more tracks with various instruments (and noise).

You can apply various filters in that program, or others, to clean up the audio in all tracks separately, and the process works much better.

From what I can hear, the separate vocal track created by UVR sounds pretty much the same as the initial audio - if the original is muffled, then the created vocal will be muffled.  UVR may have added a bit more muffling, but not much.  And that was with all defaults - I'm sure with different models and settings, it could be made much better.  From what I read afterwards, these parameters can have a huge effect on the quality of the output and amount of separation.

The bonus is that you can now listen to the audio comfortably, whereas the initial audio was difficult, if not impossible, to listen to, with all the noise present.

No one is saying that you're going to get hifi out of all this.  You're just going to end up with audio that is considerably better than the noise-riddled recordings that unfortunately make up Quiet Please (and other OTR).  A lot of the surviving OTR transcription disks were rescued from the trash, and are in dubious shape.

@artisan-radio I still think the method your using is intriguing, it seems you'll have more control to work on the voice without having to worry about altering it's underlying track at the same time. But no, it's still not a magic bullet, nor is there such thing. But... And this is simply ponderations:

The existing copies of the QP series are notorious for their poor quality, the audio of which being essentially irretrievable. --But not all, a few episodes are in quite excellent shape. - So, just maybe, by training AI on Chappells voice and delivery in the good episodes it can then be utilized perhaps to "restore" the worst portions of an episode and perhaps even re-create the missing sections of the partial episodes -- and even create a reasonable representation of what the missing in its entirety original episodes (utilizing the scripts) might sound. 

AI has no problem emulating real artist to the extent that they're practically indistinguishable from the real person, so it's really not so far fetched an idea at all. 

That might be kind of what CapCaps ai capabilities are doing already.

Another way I was thinking was to train the AI to separate organ, voice and 'other'.  The other being all the noise (and hopefully not too much other stuff).  Most OTR (except the really big budget ones) only used organ for background music, and, of course, had the vocals.  That would cover the majority of OTR, including Quiet Please.

I'm not sure what you would train the AI with on the good episodes.  The aforementioned way would just be an extension of what that AI can already do.  They've already extended it to piano, organ wouldn't be much of a stretch.

And you're quite right.  There is no magic bullet.

Posted by: @artisan-radio

↑

Another way I was thinking was to train the AI to separate organ, voice and 'other'.  The other being all the noise ..

Yeah, that's actually probably in line of softwares like CapCap are beginning to do. I suspect the Izotropes software (which is supposed to be amongst the top dog of restoration programs) is probably already began implementing.

Playfully I gave it a try with Suno, needless to say it didn't work too well, but then again it's not made for that:

Ok, I completely spelt it wrong, it's not Izotropes - it's Izotope RX. And I asked Google AI if such software is beginning to use Ai for restoration, and here's it's reply:

• iZotope RX 11

  : The newest version uses "state-of-the-art neural networks" to power its Dialogue Isolate and Music Rebalance modules. Its Repair Assistant is an AI tool that automatically analyzes your audio and builds a custom processing chain to fix it in seconds.

• Acon Digital: Their Extract:Dialogue 2 plugin is built entirely on deep learning to separate speech from background noise and reverb in real-time. Their DeVerberate 3 also uses a neural network to automatically identify and reduce room echoes.
• CEDAR Audio: Long known for precision, CEDAR recently introduced VoicEX 2, a dedicated "AI-based voice extractor". It simplifies restoration into just two controls: one for background noise and one for the voice, letting the AI handle the complex separation.
• Waves Clarity Vx: A newer competitor that gained popularity specifically because it is a "one-knob" AI-only plugin designed to remove even the most extreme background noise from vocals instantly

I went to youtube and found a playlist of 91 episodes of the show Quiet Please in discussion here and sampled about 15 of them at random out of the list and the audio quality on what seems to be all of them would be too bad for me to consider. There's many other series like the 5 or 6 shows I run that are acceptable to good audio quality. Very good on two of them. I don't know if it is worth all the trouble of trying to weed out the noise as the audio itself is so muffled on the majority of them that listening quietly you wouldn't understand what was said in many cases and add to that the audio dropouts. Separating the audio from the other interference if that could be done is in my mind futile as the audio itself is horrible. You can try cranking up the treble and volume levelling or enchancing but you can't get what is not there. 

@mark The Quiet Please series is well known as amongst the poorest quality of copies. That, in part, is why QP is of such interest to collectors. Most other OTR programs usually had multiple sources of the original audio to pick the best and compile into a complete series. But there is only one single known source for the Quiet Please episodes, and those were the ones Chappell's wife donated to the college and someone had hurriedly (and haphazardly made cassette tape copies of it before Paley Museum got a hold of them ( the original platters) and stored them away - they still have them in possession now but they've never made any attempts at restoring them and refuse to let anyone else (like Radio Spirits for example) attempt restoration either.

That's why no successful restoration of Quiet Please has ever occured, because no one has been able to get access to the original disc - there are methods of restoring the audio directly from the disc, but not from the taped copies. - but people try anyway because it's the only option there is.

You know, my fascination with the QP series, in part - a lot of it has to do with the story behind it and the close friendship that arose between Cooper and Chappel because of it, and how neither of them even ever got paid anything for doing it every week for almost 3 years, They never had a sponsor, it was basically their own labor of love, (although some of their guest cast members probably were paid -presumably out of Coopers and Chappell's pocket).  Every week Cooper would write a new episode and every week Chappel would act the script, but it wasn't like their "job":perse, not really, but those couple years were the very foundation of their lifelong friendship.. 

To be honest, if not for the backstory of the series, I probably would not of ever had any real interest in those productions at all. It's the rich combination of all that surrounded it.

If interested, here is a copy of the website I built around 10 years ago after extensive research and had the site hosted on a free server (until it wasn't free anymore). Anyway, its still more of an indepth look at Quiet Please than anything you'll find anywhere.

Anyway, if your interested in the Quiet Please history the best single source of info is my own  archived website from years ago, (it's not on the internet archive, I checked), but here is the archive of my QP site in a zip file: Just extract the zip file above to a folder and then click on the index.html file and the site will load in your browser and you can browse the whole site offline.

By the way, I am the one that managed to locate a copy of the missing QP episodes (The Venetian Blind Man) though it is missing about 10 minutes from the middle.. No one had ever heard it before. I had found it buried in a Universities Archives, I wasn't even certain what it was, it was filed only as: "Willis Cooper, Spoken Word, misc" or something like that, so on a gamble I paid about $'20 for them to make me a digital copy as well as photos of the disc itself. I had to sign something that said I would not distribute copies of it --- but I broke the law and made it public anyway (shame on me I guess). Also discovered that Paley does have in its possession two of the other "missing episodes"' that's never been available anywhere else at all in any condition and no one has ever even heard (at least not in the last 50 or 60 years).

Anyway, if your interested in the Quiet Please history the best single source of info is my own  archived website (link above) from years ago, (it's not on the internet archive, I checked), but  the archive of my QP site is still around in a zip file: Just extract the zip file above to a folder and then click on the index.html file and the site will load in your browser and you can browse the whole site offline.

I don't particularly care about pristine audio quality.  I just want to be able to hear the vocals, and therefore, the story.

I'd never heard of the Paley Museum before, and looked it up.  I fail to see what good it is to preserve media when it's just hidden away.   They do say they're digitizing their collection; perhaps they intend to attempt restoration themselves.

I'm kind of glad Radio Spirits was refused access to restore OTR, including Quiet Please.  I don't support companies that issue dubious copyright claims to material, and effectively refuse access to that material in a different way, i.e., charge a ton for it.

I'm sure there are lots of OTR missing episodes buried within collections, either private or public (as in the case of Daley).

Way back when, I started a collection of transcription disks, I had maybe 2 boxes full.  Even then, not particularly trying hard, I found, recorded and donated multiple 'missing' episodes of Guest Star to OTRR.  Most of the disks I obtained off of e-bay, believe it or not, in large lots (shipping was much cheaper then).

Although I got rid of most of my stuff when I downsized, I can understand the desire to own something physical from that period.  What I don't understand is the hoarding - keeping the content away from everyone else.  Record it yourself, get someone else to record it - but just put it out there.

History is being lost, not only from the rewriting of it by current politicians, but also from collectors.