Audio Restoration and Repair Takeaways from AES San Francisco 2017

by Jonathan Wyner, Director of Education, iZotope

August 22, 2017

Michael Romanowski, Jonathan Wyner, and Jessica Thompson

From left to right: Michael Romanowski, owner and chief mastering engineer at Coast Mastering; Jonathan Wyner, Director of Education at iZotope; and Jessica Thompson, archivist and mastering engineer

In July I had the chance to travel to the Bay Area to visit with a great community of audio professionals at the Audio Engineering Society and speak on a couple of topics. The first talk was on the subject of audio restoration techniques, during which I collaborated with Jessica Thompson, archivist and mastering engineer. We discussed approaches to restoring and repairing audio in the context of music production, restoring historic/archival recordings, and post production. The talk was moderated by Michael Romanowski of Coast Mastering.

(Jessica has co-authored a guide entitled Save Your Stuff! Beginner’s Archiving for Musicians; for more information, check out her site.)

I was mindful that I was in the region that gave birth to the SoundDroid that was developed to help restore the soundtrack of the first re-issue of the original Star Wars movie. To my knowledge, it was the first single-ended noise reduction system (it didn’t require “pre-processing” the audio to be effective), which gave birth to the NoNoise system from Sonic Solutions in the late 1980s. A special shoutout to Andy Moorer who was the brains behind the effort.

We occupied Studio D at the famed Fantasy Studios in Berkeley, CA. If you don’t know the history of the facility, you should do some online surfing. You’ll discover credits that range from Sam and Dave to Credence and Bobby McFerrin, just to name a few. You will also discover that the facility is responsible for some incredible soundtrack work for major films, including Amadeus, The English Patient and The Right Stuff. Fantasy’s dub stage is currently producing audio for Games and VR soundtracks in 9.1 and mulling the adoption of Atmos. In short, this studio is superlative looking back and looking forward!

Records and a poster at Fantasy Studios

Records on display at Fantasy Studios

There were approximately forty attendees, and our conversation ranged from how to identify noise types in order to choose the best tools to address them to how to decide how much noise to leave in to achieve best results.

Jessica Thompson spent some time discussing how she thinks about noise in her work as a restoration engineer. She thinks of it as part of the program, and when looking at it through that lens, it is not always something to be gotten rid of, and almost never something to be completely eliminated. She talked about how noise can refer to a time in history. If recordings are documents, then noise is part of that snapshot in the way that parchment might be part of a historic document from centuries ago. It would look weird to see the US Constitution on printer paper! But she went beyond, indicating that the noise could add to the beauty of a recording.

We all agreed that in order to be able to effectively reduce noise, certain things are important:

  • Train your ears to recognize the artifacts of noise reduction so you can figure out how much is too much

  • Learn to identify types of noises so you can quickly select the best tools to remove the noise. Clipping distortion and intermodulation distortion might seem the same, but they require very different types of tools to attack them.

  • Train your ears so you can focus on good sound instead of focusing on the noise. The more you understand what makes something sound good, the better you’re able to work toward that goal.

  • If noise is a distraction, try to remove it. If it isn’t a distraction, let it be.

We also touched on new machine learning informed tools that are bringing new approaches to the work.

Until fairly recently, we simply had to try to attack noise with steady state and broadband noise reducers or various types of interpolators. In the last few years, we have been able to use a class of tools called source separators (see RX 6 Advanced’s Deconstruct module) that intelligently separate a tonal signal from a noisy signal, or a sustained sound from a sound that has strong transients, and in separating them, play with levels and remix them.

Now we also are seeing machine learning provide yet another approach, whereby the machine can identify the signal to be maintained and reduce the level of everything else (see RX 6 Advanced’s Dialogue Isolate Module).

On the second day of the conference, we walked around the corner to Coast Mastering where Michael, Jessica, and I gave a mastering clinic to another forty eager audio engineers and producers. We discussed workflows, how to set mastered levels for streaming, CD and lossy files, and the importance of working in a high resolution. We all use at least 24 and 32 bits all the time, and when we can, sample rates higher than 48kHz.

A theme that persisted was how helpful the collaboration between an informed mastering engineer is to a mix engineer, giving the mix engineer perspective on her or his work. Whenever possible, all of us commonly offer feedback and advice to mix engineers before the mixes are finished so the mixes can arrive for mastering in better shape. That’s a win-win situation.

We all agreed that there seem to be two trends present in audio production that are at odds with each other:

  • Consumer audio levels are coming down as streaming services begin to dominate the consumer marketplace.

  • Mix levels are still very high (mostly too high) because of the availability of high-quality limiters that make it easy to raise the level on meters.

This was one of the impetuses for this workshop—to help engineers understand how to get the best results.

All of us also agreed that it is a common practice now to generate one version of a master without limiting (or with very little) that would go to streaming and be vinyl friendly, and to generate another hotter level for projects going to CD or for download sales.

We also debated the viability of high res in the consumer marketplace and how it might affect our work.

It was agreed that high resolution was important through the mastering process, but the extent to which consumers would spend money on high resolution is a subject for debate (or we can just wait and see!).

I was impressed by the level of interest and sophistication of the audiences, and all the more, by the incredible hospitality I was shown.

Thanks to Coast Mastering and Fantasy Studios!