Before the digital era we’re living in, storing and preserving media was no trivial task. Archives all over the world have been designed to preserve human works of all kinds so that future generations could experience them centuries after they were made.
From early-day movies to newspapers, visual arts, and audio recordings, everything that was created at a certain point in time is the embodiment of that era and, as such, a crucial element to understand and appreciate a time in history that’ll never be back.
The rise of AI technology is now offering us the possibility not just to restore these media but to bring them back to their original beauty, thanks to AI algorithms specifically designed to breathe new life into them, whether they’re Classic Era movies, century-old folk songs, or public speeches that changed the course of history.
Today, we’ll focus on the technology behind AI audio restoration tools, how they work, and why they’re so important for the preservation of human heritage.
About me
I’m a musician and record label owner with a passion for audio-related technology.
Over the years, I’ve collaborated with many tech startups focused on audio restoration, which made me appreciate the value of the preservation of sound recordings.
I’m a proud owner of a record collection that was passed on to me by my parents, so sound restoration has always been a topic very close to my heart, which is why I dedicated countless hours researching the topic and finding the best tools to achieve the highest-quality restoration results.
The importance of sound preservation
To pass on sound recordings to future generations means preserving the narrative of human history.
Old analog recordings of all kinds allow future generations to connect with voices from the past that have been silenced by time, to hear them if they were still with us today, and to experience history in a direct and immersive way.
For decades, the preservation of audio recordings had to deal with physical degradation, technological obsolescence, natural decay of storage media, and negligence, which is how we already lost many recordings which we’ll never hear again.
Think about the original recordings of Native American and Australian Indigenous languages and music made on wax cylinders in the late 19th and early 20th centuries. Many of these recordings were the only documentation of languages and traditions that have since vanished.
Or the immense list of lost records from the Library of Congress. By simply skipping through the list, you’ll see many records by legendary artists (Bing Crosby, Charley Patton and Bertha Lee, Seth S. Weeks) are currently unaccounted for.
As they state on their website, out of the more than 3,000 cylinder recordings made by the North American Phonograph Company (NAPCo) from 1889 to 1894, it is believed that only about 2% have been preserved or documented.
Losing such recordings is not just a loss for a specific culture but for humanity’s collective heritage.
Thanks to AI technology, now we can ensure that the voices and sounds of history are not lost to the silence of time.
Challenges with traditional restoration methods
Analog restoration means directly working on the physical media, whether it’s vinyl records, wax cylinders, or magnetic tapes. This process usually involves cleaning the media to remove dirt, repairing physical damage, and using specialized playback equipment to minimize wear and enhance sound quality.
There are many limitations to the analog restoration process. As you’re working on a fragile physical object, it’s often difficult to completely mitigate scratches, warping, or chemical degradation.
Removing unwanted noise can sometimes compromise the original sound, making it dull and lacking clarity. Furthermore, the restoration process can compromise the media itself, sometimes irreversibly.
The relatively recent development of digital methods offers more precise control over the sound, giving engineers the possibility to improve the quality of the recordings without compromising the physical object.
This requires skills and time, and striking a balance between preserving the authenticity of the recording while enhancing its quality is a complex task and a controversial one, too (more on that later).
AI technology and its innovative approach have revolutionized the field of sound restoration, offering a way to overcome the challenges of both manual analog and digital restoration.
The rise of AI technologies to restore audio recordings
In the past decade, AI technologies have become increasingly popular in the field of media restoration, mostly thanks to advancements in machine learning and neural networks.
By learning from datasets that comprise both clean and degraded audio samples, AI models can isolate and eliminate unwanted noise while preserving the original sound’s integrity and character.
One of the most critical accomplishments of this new technology is the Great 78 Project, a project run by archive.org, which has used AI to rejuvenate over 400,000 early 20th-century recordings, a task almost impossible to achieve with digital, let alone analog, restoring techniques.
With AI restoration, the process starts with preparing a comprehensive dataset of audio recordings, which trains the AI model to recognize patterns of degradation, after which the model processes the degraded recordings using the learned capabilities to remove noise and enhance sound quality.
Nowadays, there are countless AI software designed to improve audio quality, accessible to anyone, and each with their own AI algorithm and learning model, which gives them a unique approach to restoration.
While these software are designed to be as automated as possible, the presence of a skilled audio engineer is still what truly enhances the power of these tools and breathes new life into records.
The cultural impact
Using AI to preserve culture brings up complex issues about authenticity, who owns the content, and keeping cultural heritage intact.
A big worry is whether restoring recordings with AI changes them too much, especially when those recordings are fundamental for history, culture, or art.
While it’s important to pass on these recordings to future generations, it’s paramount to keep their original spirit, and the line between authenticity and transformation is often blurred.
There’s also the question of who owns the restored recordings. As these recordings might become more valuable or significant once they’re restored, figuring out who has the rights to them gets complicated.
Technically speaking, the original owners, the people who developed the AI, and the engineers who did the restoration are all partially involved in the creation of the restored record. So, if the restoration is successful and the album gets published, who should get the royalties from the work?
There’s also the complex issue of restoring recordings from cultural minorities or cultures that have since disappeared. In those cases, these recordings could be rare links to languages, music, or stories for many people, so preserving them without affecting the original meaning of those recordings becomes even more important.
The future of AI audio restoration
I have no doubts that one of the most significant prospects of AI audio restoration is its constant improvement in quality and accuracy.
The technology is getting better at telling the difference between background noise and the real sound in the recordings, which means we can fix a wider range of problems without messing up the original audio.
The cleaned-up recordings will sound clearer and more like they originally did, letting us hear history just as it was.
AI tools become easier to use and cheaper, so anyone from local historians to hobbyists will be able to restore old recordings, which opens up the chance to save and hear a lot more of our audio past.
Finally, as AI gets better at restoring audio, we might be able to mix these sounds with virtual reality or augmented reality. This means we could experience concerts, speeches, or everyday life from history in a way that feels real and alive.
It’s a whole new way to connect with the past, and it’ll happen very soon.
Final thoughts
The future of using AI to breathe new life into the sounds from our past could change the way we understand and enjoy history.
We’re going to hear audio from a bygone era and connect with our culture and others we’d never had the chance to interact with. We could use these incredible tools ourselves, to preserve the history of our city, neighborhood, and family, and ensure that those moments will be accessible to those who come after us.
But as we get excited about all these revolutionary changes, we need to make sure that as we explore and innovate, we do it responsibly, so that everyone benefits from it and our shared history gets the respect it deserves.
