From vault to vision: What SXSW reminded me about the work that still matters

I've spent my career trying to get things out of the vault. Not physically, but conceptually — pushing back against the idea that once something goes into an archive, it's gone forever. The Raiders of the Lost Ark image always stays with me: that impenetrable warehouse, the sense that whatever's inside will never be found again. In fact, archivists and librarians have done wonders providing access despite backlogs and resource constraints — and with AI, their work is in many ways finally being rewarded.

It's what made the panel I joined at SXSW, "From Vault to Vision: Preserving Film History with AI," feel like a conversation that's been building for thirty years. Joined by my friends and colleagues Violeta Hennessy from Paramount and Carin Forman from AWS, we spent an hour talking together about the exciting ways that AI can be a force for good in bringing an archive back to life. I came away with a few things I want to share, because I think there are practical takeaways here for anyone managing collections of moving images, audio, or media in any form.

The digitization problem is not solved. It's not even close.

One of the first things we addressed is something the industry has a tendency to gloss over: only about 20% of the world's video content has been digitized. That number should stop people in their tracks. The rest is on physical media, much of it deteriorating right now, on a shelf somewhere.

Acetate-based film, which covers essentially everything made between 1950 and the late 1990s when digital production began to take hold, is at a particularly alarming inflection point. At a certain stage of deterioration, the process accelerates. It's called autocatalysis. Once it reaches that point, there's very little you can do. This is not a theoretical future risk. It's a present, active emergency for a lot of collections.

Magnetic media has its own challenge. The reason digitizing videotape and audiotape is so complicated is that there's no way to extract the content without using the original playback equipment. Imagine having to use a vintage typewriter every time you needed to scan a document — that's the equivalent of what's involved here. What that means in practice is that anyone doing this work at scale has a warehouse full of vintage playback machines, sourced from eBay, museum closets, and industry connections, maintained by engineers who understand both the technology and the media well enough to know what good output sounds and looks like. That expertise is irreplaceable, and like the equipment itself, it's disappearing.

The window to act is small and closing fast. If you manage a collection and you've been treating digitization as a future priority, this is the moment to revisit that calculus.