I believe we can do a fabulous job of bringing the music from the 78rpm era back to vibrant life if we really understand wear and if we could model the instruments and voices.

In other words, I believe we could reconstruct a performance by semantically modeling the noise and distortion we want to get rid of, as well as modeling the performer’s instruments.

To follow this reasoning—what if we knew we were examining a piano piece and knew what notes were being played on what kind of piano and exactly when and how hard for each note—we could take that information to make a reconstruction by playing it again and recording that version. This would be similar to what optical character recognition (OCR) does with images of pages with text—it knows the language and it figures out the words on the page and then makes a new page in a perfect font. In fact, with the OCR’ed text, you can change the font, make it bigger, and reflow the page to fit on a different device.

What if we OCR’ed the music? This might work well for the instrumental accompaniment, because then we would handle a voice, if any, differently. We could have a model of the singer’s voice based on not only this recording and other recordings of this song, but also all other recordings of that singer. With those models we could reconstruct the voice without any noise or distortion at all.

We would balance the reconstructed and the raw signals to maintain the subtle variations that make great performances.   This could also be done for context as sometimes digital filmmakers add in some scratched film effects.

So, there can be a wide variety of restoration tools if we make the jump into semantics and big data analysis.

The Great 78 Project will collect and digitize over 400,000 digitized 78rpm recordings to make them publicly available, creating a rich data set to do large scale analysis. These transfers are being done with four different styli shapes and sizes at the same time, and all recorded at 96KHz/24bit lossless samples, and in stereo (even though the records are in mono, this provides more information about the contours of the groove). This means each groove has 8 different high-resolution representations of every 11 microns. Furthermore, there are often multiple copies of the same recording that would have been stamped and used differently. So, modeling the wear on the record and using that to reconstruct what would have been on the master may be possible.

Many important records from the 20th century, such as jazz, blues, and ragtime, have only a few performers on each, so modeling those performers, instruments, and performances is quite possible.  Analyzing whole corpuses is now easier with modern computers, which can provide insights beyond restoration as well as understand playing techniques that are not commonly understood.

If we build full semantic models of instruments, performers, and pieces of music, we could even create virtual performances that never existed.  Imagine a jazz performer virtually playing a song that had not been written in their lifetime. We could have different musician combinations, or singers performing with different cadences. Areas for experimentation abound once we cross the threshold of full corpus analysis and semantic modeling.

We hope the technical work done on this project will have a far-reaching effect on a full media type since the Great 78 Project will digitize and hold a large percentage of all 78rpm records ever produced from 1908 to 1950.  Therefore, any techniques that are built upon these recordings can be used to restore many many records.

Please dive in and have fun with a great era of music and sound.

(we get a sample every 11microns when digitizing the outer rim of a 78rpm record at 96KHz.   And given we now have 8 different readings of that, with 24bit resolution, we hopefully can get a good idea of the groove.   There are optical techniques that are very cool, but those have their own issues, I am told

10″ * 3.14 = 31.4″ circumference = 80cm/revolution

@ 78rpm:  60 seconds/min / 78revolutions/minute = .77 seconds / revolution

80cm/rev   / (.77sec/rev)  = 104cm/sec

96Ksampes/sec

104cm/sec / (96ksamples/sec) = 11microns )

By Jessica Thompson, Coast Mastering

The Great 78 Project
A few times a year, I join B. George in the Internet Archives’ warehouses to help sort and pack 78rpm discs to ship to George Blood L.P. for digitization. As a music fan and a professional mastering and restoration engineer, I get a thrill from handling the heavy, grooved discs, admiring the fonts and graphic designs on the labels, and chuckling at amusing song titles. Now digitized, these recordings offer a wealth of musicological, discographic and technical information, documenting and contextualizing music and recording history in the first half of the 20th century.

The sheer scale of this digitization project is unprecedented. At over 15,000 recordings and counting, the value strictly in terms of preservation is clear, especially given the Internet Archive’s focus on digitizing music less commonly available to researchers. Music fans can take a deep dive into early blues, Hawaiian, hillbilly, comedy and bluegrass. I even found several early Novachord synthesizer recordings from 1941.

As a researcher and audio restoration engineer, the real goldmine is in the aggregation of discographic and technical metadata accompanying these recordings. Historians can search for and cross reference recordings based on label, artist, song title, year of release, personnel, genre, and, importantly, collection. (The Internet Archive documents the provenance of the 78rpm discs so that donated collections remain digitally intact and maintain their contextual significance.) General users can submit reviews with notes to amend or add to metadata, and the content of those reviews is searchable, so metadata collection is active. No doubt it will continue to improve as dedicated and educated users fill in the blanks.

Access to the technical metadata offers a valuable teaching tool to those of us who practice audio preservation. For audio professionals new to 78s and curious about how much difference a few tenths of a millimeter of stylus can make, the Internet Archive offers 15,000+ examples of this. Play through the different styli options, and it quickly becomes apparent that particular labels, years and even discs do respond better to specific styli sizes and shapes. This is something audio preservationists are taught, but rarely are we presented with comprehensive audio examples. To be able to listen to and analyze the sonic and technical differences in these versions marries the hard science with the aesthetic.

Playback speeds were not standardized until the late 1920s or early 1930s, and most discs were originally cut at speeds ranging from 76-80rpm (and some well beyond). The discs in the George Blood Collection were all digitized at a playback speed of 78rpm. Preservationists and collectors debate extensively about the “correct” speed at which discs ought to be played back, and whether one ought to pitch discs individually. However, performance, recording and manufacturing practices varied so widely that even if a base speed could generally be agreed upon, there will always be exceptions. (For more on this, please check out George Blood’s forthcoming paper Stylus Size And Speed Selection In Pre-1923 Acoustic Recordings in Sustainable audiovisual collections through collaboration: Proceedings of the 2016 Joint Technical Symposium. Bloomington, IN: Indiana University Press.)

Every step of making a recording involves so many aesthetic decisions – choices of instrumentation, methods of sound amplification, microphone placement, the materials used in the disc itself, deliberate pitching of the instruments and slowing or speeding of the recording – that playback speed simply become one of many aesthetic choices in the chain. As preservationists, we are preserving the disc as an historic record, not attempting to restore or recreate a performance. (Furthermore, speed correction is possible in the digital realm, should anyone want to modify these digital files for their own personal enjoyment).

How do they sound? Each 78rpm disc has an inherent noise fingerprint based on the frequency and dynamic range the format can replicate (limited, compared to contemporary digital playback formats) and the addition of surface noise from dust, dirt and stylus wear in the grooves. As expected, the sound quality in this collection varies. Some of these discs were professionally recorded, minimally played, stored well, and play back with a tolerable, even ignorable level of surface noise relative to the musical content. Others were recorded under less professional circumstances, and/or were much loved, frequently played, stored without sleeves in basements and attics, and therefore suffer from significant surface noise that can interfere with enjoyment (and study) of the music.

Yet, a compelling recording can cut through noise. Take this 1944 recording of Josh White performing St. James Infirmary, Asch 358-2A. This side has been released commercially several times, so if you look it up on a streaming service like Spotify, you can listen to different versions sourced from the same recording (though almost certainly not from the same 78rpm disc). They play at different speeds, some barely perceptibly faster or slower but at least one nearly a half-step faster than the preservation copy digitized by George Blood L.P. They also have a range of noise reduction and remastering aesthetics, some subtle and some downright ugly and riddled with digital artifacts. The version on the Internet Archive offers a benchmark. This is what the recording sounded like on the original 78rpm disc. Listen to the bend in the opening guitar notes. That technique cuts through the surface noise and should be preserved and highlighted in any restored version (which is another way of saying that any noise reduction should absolutely not interfere with the attack and decay of those luscious guitar notes).

McGill University professor of Culture and Technology Jonathan Sterne wrote a book – The Audible Past: Cultural Origins of Sound Reproduction – that is worth reading for anyone interested in a cultural history of early recording formats, including 78s. As Sterne says, sound fidelity is “ultimately about deciding the values of competing and contending sounds.” So, in listening to digital versions of 78s on the Internet Archive, music fans, researchers, and audio professionals alike engage in a process of renegotiating concepts of acceptable thresholds of noise and what that noise communicates about the circumstances of the recording and its life on a physical disc.

Fortunately, our brains are very good at calibrating to accept different ratios of signal to noise, and, I found, the more I listened to 78rpm recordings on the Internet Archive, the less I was bothered by the inherent noise. Those of us who grew up on CDs or digitally recorded and distributed music are not used to the intrusions of surface noise. However, when listening to historic recordings, we are able to adjust our expectations and process a level of noise that would be ridiculous in contemporary music formats. (Imagine this week’s Billboard Top 100 chart topper, Bruno Mars’s “That’s What I Like,” with the high and low end rolled off, covered in a sheen of crackles and pops). The fact that these 78rpm recordings sound, to us, like they were made in the 1920s, 1930s, 1940s lets them get away with a different scale of fidelity. The very nature of their historicity gets them off the hook.

In analog form, crackles and pops can be mesmerizing, almost like the sound of a crackling fire. However, once digitized, those previously random pops become fixed in time. What may have been enjoyable in analog form becomes a permanent annoyance in digital form. The threshold of acceptable noise levels moves again.

This means that noise associated with recording carriers such as 78rpm discs is almost always preferably to noises introduced in the digital realm through the process of attempted noise reduction. Sound restorationists understand that their job is to follow a sonic Hippocratic oath: do no harm. Though noise reduction tools are widely available, they range in quality (and accordingly in cost), and are merely tools to be used with a light or heavy touch, by experienced or amateur restorationists.

The question of whether noise reduction of the Internet Archive’s 78rpm recordings could be partially automated makes my heart palpitate. Though I know from experience that, for example, auto-declickers exist that could theoretically remove a layer of noise from these recordings with minimal interference with the musical signal, I don’t believe the results would be uniformly satisfactory. It is so easy to destroy the aura of a recording with overzealous, heavy-handed, cheap, or simply unnecessary noise reduction. Even a gentle touch of an auto-declicker or de-crackler will have widely varying results on different recordings.

I tried this with a sampling of selections from the /georgeblood/ collection. I chose eleven songs from different genres and years and ran two different, high quality auto-declickers (the iZotope RX6 Advanced multiband declicker and CEDAR Audio’s declick) on the 24bit FLAC files. The results were uneven. Some of the objectively noisier songs, such as Blind Blake’s Tampa Bound, Paramount 12442-B, benefited from having the most egregious surface noises gently scrubbed.

Tampa Bound Flat Transfer vs Tampa Bound Declicked, Dehissed and Denoised
that’s a lot of noise!

However, a song with a strong musical presence and mild surface noise such as Trio Schmeed’s Yodel Cha Cha, ABC-Paramount 9660, actually suffered more from light auto-declicking because the content of the horns and percussive elements registered to the auto-delicker as aberrations from the meat of the signal and were dulled. A pop presents as an aberration across all frequencies. Mapped visually across frequency, time and intensity, it looks like a spike cutting through the waveform. A snare hit looks similar and is therefore likely to be misinterpreted by an auto-declicker unless the threshold at which the declicker deploys is set very carefully. This difference is why good restorationists earn their pay.

Yodel Cha Cha flat transfer and denoised. Notice the “clicks and pops” have been scrubbed,
but so has wanted high end content in the music.

 I am approaching this collection as a listener and music fan, as a researcher, and as an audio professional, three very different modes of listening and interacting with music. In all cases, the Internet Archive 78rpm collection offers massive amounts of music and data to be explored, discovered, enjoyed, studied and utilized. Whether you want to listen to early Bill Monroe tunes, crackles, pops and all, or explore hundreds of recordings of pre-war polkas, or analyze the effects of stylus size on 1930s Victor discs, the Internet Archive provides the raw materials in digital form and, not to be underestimated, preserves the original discs too.

Written by B. George, the Director of ARChive of Contemporary Music in NYC, and Curator of Sound Collections at the Internet Archive in San Francisco.

While audio CDs whiz by at about 500 revolutions per minute, the earliest flat disks offering music whirled at 78rpm. They were mostly made from shellac, i.e., beetle (the bug, not The Beatles) resin and were the brittle predecessors to the LP (microgroove) era. The format is obsolete, and the surface noise is often unbearable and just picking them up can break your heart as they break apart in your hands. So why does the Internet Archive have more than 200,000 in our physical possession?

A little over a year ago New York’s ARChive of Contemporary Music (ARC) partnered with the Internet Archive to focus on preserving and digitizing audio-visual materials. ARC is the largest independent collection of popular music in the world. When we began in 1985 our mandate was microgroove recordings – meaning vinyl – LPs and forty-fives. CDs were pretty much rumors then, and we thought that other major institutions were doing a swell job of collecting earlier formats, mainly 78rpm discs. But donations and major research projects like making scans for The Grammy Museum and The Ertegun Jazz Hall of Fame placed about 12,000 78s in our collection.

For years we had been getting calls offering 78 collections that we were unable to accept. But when space and shipping became available through the Internet Archive, it was now possible to begin preserving 78s. Here’s a short history of how in only a few years ARC and the Internet Archive have created one of the largest collections in America.

Our first major donation came from the Batavia Public Library in Illinois, part of the Barrie H.Thorp Collection of 48,000 78s.

We’re always a tad suspicious of large collections like these. First thought is, “Must be junk.” Secondly, “It’s been cherrypicked.” But the Thorp Collection was screened by former ARC Board member Tom Cvikota, who found the donor, helped negotiate the gift and stored it. That was in 2007. Between then and our 2015 pickup Tom arranged for some of the recordings to be part of an exhibition at the Greengrassi Gallery, London, (UK, Mar-Apr, 2014) by artist Allen Ruppersberg, titled, For Collectors Only (Everyone is a Collector).

What makes the Thorp collection unique is the obsessive typewritten card catalog featured in a short film hosted on the exhibition’s webpage. Understanding why you collect and how you give your interests meaning is a part of Allen’s work – artworks that focus on the collector’s mentality. One nice quote by Allen referenced in Greil Marcus’ book, The History of Rock n’ Roll in Ten Songs is, “In some cases, if you live long enough, you begin to see the endings of things in which you saw the beginnings.”

Philosophical musings aside, there are 48,000 discs to deal with. That meant taking poorly packed boxes — many of them open for 20 years — and re-boxing them for proper storage. The picture below shows an example of how they arrived (on the right), and how they were palletized (on the left.)

The trick to repacking in a timely fashion is to not look at the records. It’s a trick that is never performed successfully. Handling fragile 78s requires grabbing one or just a few at a time. So we’re endlessly reading the labels, sleeving and resleeving, all the time checking for rarities, breakage and dirt.

Now we didn’t do all this work on our own. Working another part of the warehouse was two-and-a-half month old Zinnia Dupler — the youngest volunteer ever to give us a hand. Mom also helped a bit.

A few minutes after the snap I found this gem in the Thorp collection. Coincidence? I don’t think so…

“Burpin” is a country novelty tune from out of Texas by Austin broadcaster and humorist Richard “Cactus” Pryor (1923 – 2011). It came from a box jam-packed with country and hillbilly discs. This was a pleasant surprise, as we expected the collection to be like most we encounter – big band and bland pop. But here was box-after-box of hillbilly, country, and Western swing records. Now, I use’ta think I knew a bit about music. But with this collection, it was back to school for me. Just so many artists I’ve never heard of or held a record by. As we did a bit of sorting, in the ‘G’s alone there’s Curly Gribbs, Lonnie Glosson and the Georgians. Geeez! Did you know that Hank Snow had a recordin’ kid, Jimmy, and he cut “Rocky Mountain Boogie” on 4 Star records, or that Cass Daley, star of stage and screen, was the ‘Queen of Musical Mayhem?” Me neither.  The Davis Sisters, turns out, included a young Skeeter Davis(!) and not to be confused with the Davis Sister Gospel group, also in this collection. Then there’s them Koen Kobblers, Bill Mooney and his Cactus Twisters, and Ozie Waters and the Colorado Hillbillies. No matter they should be named the Colorado Mountaineers, they’re new to me.

For us this donation is a dream: it allows us to preserve material that was otherwise going to be thrown away; it has a larger cultural value beyond the music; and it contained a mountain of unfamiliar music, much of it quite rare. And most of it is not available online.

It was a second large donation that prompted the Internet Archive to move toward the idea that we should digitize all of our 78s. The Joe Terino Collection came to us through a cold call, the collection professionally appraised at $500,000. The 70,000 plus 78s were stored in a warehouse for more than 40 years, originally deposited by a distributor. Here’s the kicker: they said that we could have it all, but we had to move it – NOW! Internet Archive did and it came in on 72 pallets, in three semis, from Rhode Island to San Francisco, looking like this…

So Fred Patterson and the crackerjack staff out in our Richmond warehouses (Marc Wendt, Mark Graves, Sean Fagan, Lotu Tii, Tracey Gutierrez, Kelly Ransom, and Matthew Soper) pulled everything off the ramshackle pallets and carefully reboxed this valuable material.

How valuable? Well, we’re really not so sure yet, despite the appraisal, as just receiving and reboxing was such a chore. One hint is this sweet blues 78 that we managed to skim off the top of a pile.

The next step is curating this material, acquiring more collections and moving towards preservation through digitization. Already we have a pilot project in the works with master preservationist George Blood to develop workflow and best digitization practices.

We’re doing all this because there’s just no way to predict if the digital will outlast the physical, so preserving both will ensure the survival of cultural materials for future generations to study and enjoy. And, it’s fun.