How AI Is Transforming Audio Restoration in Archival Video: from BBC to Netflix

The past is making a powerful comeback in the age of digital streaming and 4K restoration. From classic crime documentaries to 1980s sitcoms, archival video content is finding new life and new audiences on modern platforms. Yet, amid the polished visuals, one persistent challenge remains: poor audio quality. Now, thanks to artificial intelligence, even the most damaged recordings can be brought back to life with unprecedented clarity.

Why restore archival video: when cultural preservation meets technological innovation

Audio and video don’t just represent a technical concern: it’s a core part of cultural memory. If viewers can’t hear what’s being said or struggle to see clearly what is happening, the historical value of the footage diminishes. For younger audiences raised on clean digital audio and on 4K video, poor content quality can be a dealbreaker.

This is just one motivation to restore archival content, here are some more:

• The level of technological development now allows to delete possible defects on old recordings and thus, enjoy these pieces of art in the best video and audio quality that is possible. A prime example of improved quality in an old film through restoration is The Wizard of Oz (1939). 

Warner Bros. undertook a 70th anniversary edition restoration that involved scanning the original camera negatives in 8K resolution, combined with advanced ultra-resolution software to align the RGB elements pixel by pixel. This meticulous process resulted in a gorgeously saturated and impeccably detailed version of the film, with vibrant colors and clarity far surpassing previous editions;

• Next, if one film is restored, why not restore other analog recordings and make them accessible to a wide audience? 

For example, documentaries: the popularity of historical documentaries is reflected in the strong growth of the documentary film and TV show market, which was valued at approximately $12.96 billion in 2024 and is projected to grow significantly in the next ten years. A powerful example of historical restoration in documentary filmmaking is They Shall Not Grow Old (2023), directed by Peter Jackson. Drawing from original World War I footage housed at the UK’s Imperial War Museum, the film underwent an extraordinary transformation: meticulously restored, colorized, and enhanced with surround sound and 3D effects. What was once grainy black-and-white footage is reborn as vivid, immersive imagery that brings the front lines to life with unprecedented realism. 

Instead of traditional narration, the film features voice recordings of British veterans, adding emotional depth and authenticity to the restored visuals. Thus, improving the audio or video quality of the content will have a significant impact on history as a science; on historical research, for instance. It will also mean a deeper understanding of the events of the past that is also a part of the national heritage of each nation.

A Renaissance in archival video

The resurgence of documentaries, retrospectives and remastered classics has sparked renewed interest in archival content. Streaming platforms like Netflix, Max, and Disney+ are investing heavily in restoring legacy films and series to meet growing viewer demand. 

To illustrate, Disney, to mark The Walt Disney Company’s 100th anniversary in October 2023, The Walt Disney Studios’ restoration and archive team determined to return to the original negatives and perform the ultimate restoration of not just Snow White and the Seven Dwarfs but also Cinderella (1950) and 27 classic shorts, from Trolley Troubles (1927), with Oswald the Lucky Rabbit, to the Goofy-starring Aquamania (1961), matching each film to the clarity and richness of their original theatrical showings. (All debuted on Disney+ in 2023.)”

Of course, this is not an easy process, as “restoring a film is a journey of anywhere from three to twelve months, depending on the age or length of each feature.” According to Kevin Schaeffer, Director of Restoration & Library Management at the Walt Disney Studios Film Archive, “on Cinderella they borrowed an original 1950 nitrate Technicolor IB print from the Library of Congress as their guide” as sometimes the Film Archive just doesn’t have the elements that are needed for the restoration.

For instance, global streaming revenue for documentaries grew by over 80% between 2020 and 2024, and studios are racing to digitize and upgrade their catalogs in 4K and 8K resolution.

But with higher resolution comes higher expectations, and that includes sound.

Before AI: the painstaking art of audio restoration

Old footage often comes with degraded audio plagued by hiss, hum, crackle and limited frequency range. Whether it’s a radio report from the 1930s or a vintage courtroom recording, these materials were never meant for today’s high-definition formats or earbuds. 

Before the rise of AI-assisted tools, audio restoration was a painstaking, manual process requiring hours or even days of detailed work in a digital audio workstation (DAW). Engineers relied heavily on a combination of EQ adjustments, noise reduction plugins, spectral editing, and dynamic processing such as compression to remove unwanted artifacts like hiss, hum, crackle, or distortion. 

Here is a real example of audio restoration that took days of painstaking manual work in a digital audio workstation (DAW) recounted by an audio engineer. This person worked on a critical case involving a garbled cassette recording of a threatening phone call. The original recording was made on a “pocket cassette deck” from a landline phone, resulting in very poor audio quality with heavy distortion and noise. The engineer spent two full days manually cleaning up the audio using early DAW tools, primarily equalizers and rudimentary plugins, to clarify the voice enough for the recording to be used as evidence by the police. 

As sound engineers know too well, traditional tools often come with trade-offs: removing noise can strip out important sonic details and human fatigue or bias introduces inconsistency. Moreover, the scale of media archives makes it virtually impossible to restore everything manually.

However, traditional workflows came with significant limitations. Even the most experienced sound engineers faced challenges in balancing clarity with authenticity: over-processing could result in unnatural-sounding voices or the loss of subtle sonic details. Many plugins were limited in scope, often failing to differentiate between speech and background noise with enough precision. Besides, the human factor, such as fatigue, subjectivity, and time constraints, frequently impacted consistency across large archival projects. In complex cases, much of the audio was deemed unusable, simply because the tools couldn’t recover it without sacrificing quality.

Meet AI: A paradigm shift

Recent advances in AI-powered audio restoration tools are changing the game. Trained on vast datasets of vocal, music, and noise patterns, AI models can now intelligently separate dialogue from background noise, eliminate unwanted sounds, and even reconstruct missing frequencies.

Solutions like LALAL.AI have become indispensable for production studios and archival teams. Using source separation technology, LALAL.AI can isolate speech from ambient sound, clean up clicks and hisses, and prepare content for dubbing, subtitling, or remastering; all with a level of speed and precision previously unimaginable. 

Compared to traditional workflows, AI tools can reduce restoration time by up to 20x. For example, consider a 1960s cassette recording riddled with distortion and tape hiss. In the past, restoring this might take a sound engineer multiple days. Now, with AI, a high-quality version can be produced in a matter of hours.

Why high-quality sound restoration matters

Firstly, it’s the preservation of the cultural heritage: if it’s left unattended, it can easily deteriorate and just perish in time. Some great old films are already lost for us forever: for instance, the 1925 version of the Phantom of the Opera whose only copy burnt in fire in the 1940s or 75% of all silent films in general.

The quality of sound in archival video plays a vital role in preserving the authenticity, emotional resonance, and informational depth of historical material. Clear, well-preserved audio ensures that voices, ambient sounds, and spoken words are not lost to time, but instead digitally safeguarded for future generations. When we enhance and protect these auditory elements, we maintain the facts of the past as well as the atmosphere and humanity that bring history to life.

High-quality sound in any video content plays a big role in attracting younger audiences:  according to statistics,younger generations like Gen Z have a strong preference for high-quality video and audio content as 94% of Gen Z express interest in receiving interactive, high-quality video content.

Gen Z and Millenials are, by the way, major contributors to the growth of streaming which accounted for 44.8% of total TV viewership in May 2025, surpassing combined broadcast and cable for the first time. These younger audiences prefer on-demand, high-quality streaming content including documentaries and historical series.

Thus, improving the sound quality using the most modern means like AI voice cleaning, for example, is really vital and is already used by important players in the media industry.

From BBC to Netflix: real-world applications

AI-powered audio restoration is no longer just an experimental tool: it’s being actively used by some of the world’s leading media institutions. From national broadcasters to global streaming platforms, organizations are adopting machine learning technologies to breathe new life into archival content. These real-world applications show how AI is transforming restoration from a slow, manual craft into a scalable solution that meets modern production standards without sacrificing historical integrity:

• Netflix has applied AI technologies to upscale and remaster older series, including the high-profile Seinfeld remaster. In this process, not only was the video enhanced, but AI-driven tools were employed to clean the audio track, ensuring it met today’s streaming standards. Also, the AI technology has been used by Netflix to upscale older shows to HD, such as the 1980s sitcom A Different World (however, the restoration of this one has been criticized by some fans due to quality issues); 
• BBC Archives has been leveraging AI to restore decades-old radio and television content. In projects like BBC Rewind, machine learning was used to clean up audio from broadcasts dating back to the mid-20th century, making them suitable for modern rebroadcasts and public access. Another BBC project, called BBC Archive Restoration Project harnesses machine learning to streamline and support the traditionally time-consuming process of restoring legacy video content;

The Criterion Collection, ARTE, and the National Film Board of Canada have also embraced AI-driven audio solutions to restore classic films for festivals, education, and streaming. For instance, a standout example of classic film restoration is a meticulous work of the Criterion Collection on Alfred Hitchcock’s Foreign Correspondent (1940). The digital cleanup involved removing scratches, dirt, flicker, and other physical imperfections, alongside precise color correction to ensure accurate tonal balance, rich blacks and properly calibrated whites. On the audio side, engineers used spectral repair tools to eliminate hisses, pops, and distortion, blending manual techniques with AI-assisted processes. The result is a stunning restoration that preserves the film’s original character while enhancing its clarity for today’s viewers.

Where LALAL.AI fits in

LALAL.AI is particularly valuable in large-scale archive restoration, offering:

• Artifact removal (e.g., crackles, background hum, static) with the help of Voice Cleaner;
• Voice and dialogue isolation from noisy backgrounds with the help of LALAL.AI’s Stem Splitter and Lead/Back Splitter;
• Batch processing via API for handling thousands of hours of material efficiently;
• Preprocessing for dubbing, localization, and subtitling with the help of Voice Cloner and Voice Cleaner.

"The best use case for the de-noise algorithm I found that other options couldn't achieve is when a speaker in a podcast or live recording hits the microphone. LALAL.AI can separate the noise and the signal almost perfectly," says Jean-Philippe Villemure, our client from Villemure Mastering.

How to use LALAL.AI

1. LALAL.AI is available online, so just go to the LALAL.AI website and choose the product needed from Products’ section (or use a desktop/Android/iOS app).

2. Log in using one of several login options available.

3. 10 minutes are available for free for new users within the Starter pack of LALAL.AI - it works for all the tools from the set.
4. Then, click on Select Files/Choose Audio or Video/Choose song depending on the product you will be using and just follow the process! Then just wait until the audio is processed, that’s all.

The convergence of AI and archival restoration is not a passing trend, it’s a necessary evolution. More production studios, streaming services, and cultural institutions embrace these tools, so we surely can expect higher quality restorations as well as broader access to historical content that might otherwise remain hidden in the static of the past.

Follow LALAL.AI on Instagram, Facebook, Twitter, TikTok, Reddit, LinkedIn, and YouTube to keep up with all our updates and special offers.