For artists & art historians
You made, or you study, a work that only runs on old software — a CD-ROM piece, a Director or Flash work, a small program that phones a server that no longer exists. It half-works, or doesn’t run at all, and you want to know what it actually is and how to keep it alive.
That is exactly what this system is for. You do not need to be technical to use it.
Modern web art too — same preservation shape. This is not only for legacy pieces. An Electron application (Chromium and Node bundled together) or a browser extension has exactly the same problem: a runtime frozen at one moment and servers that may already be gone. The analysis treats them the same way.
You don’t have to set anything up
Standing up virtual machines, wiring old operating systems, capturing network traffic — that is our job, not yours. ZKM can provide the infrastructure. You bring the artwork; we run the machinery.
There is one thing we ask you to bring yourself:
You bring your own AI
Reading and understanding an artwork is done by an AI (a large language model). For any serious work that means a lot of computation, and computation costs money. So the deal is:
We are happy to provide the infrastructure. You provide your own AI.
In practice, this is usually not something you set up personally. If a museum or class runs the workbench for you, your group already has an AI budget — you just work, and the cost lands on your group. If you run the tool yourself, you point it at an AI provider of your choice (or a free local model). Either way the cost — usually a few euros for a small work — sits with you or your institution, not with us, and nothing about your artwork is shared beyond the AI you chose. See Bring your own AI for the details.
Why this way? Because running old software is cheap and we can absorb it, but AI usage scales with every artwork and every institution, and we can’t give that away for free. It also keeps you in control: your key, your provider, your bill, your data.
What you get back
- An autopsy report — in plain language: what the work is, what it was built with, what machine it needs, and — importantly for a broken piece — every place it tries to reach on the internet. If a work is dying because a server is gone, you’ll see exactly which server, so it can be restored or re-served from an archive.
- A running machine — the work booted inside an era-correct virtual computer (the right Windows or Mac of its time), where you can actually see it run again.
- A revival plan — the ordered steps to bring it back, so a conservator (now or in ten years) can reproduce it.
What it looks like
The workflow is a simple guided flow — Bring in the artwork:
- Identity — tell us what the work is (title, artist, when it arrived). Anything you know that the files can’t reveal is valuable.
- The original — upload the artwork bundle. It’s stored untouched; everything else is derived from it.
- Autopsy — we analyze it. If it looks like you accidentally uploaded an installer or a whole runtime alongside the art, we’ll pause and ask you — never reject it (a self-installing work is a legitimate artwork).
- The machine — we suggest the era-correct virtual computer(s); you can pick more than one to compare (say Windows XP and Windows 2000).
- Run it — the artwork boots in the machine, and you can watch it work.
Full walkthrough: Bring your artwork.
A note for historians
You don’t have to run anything to get value here. The autopsy report alone is a primary source: it documents a work’s technical body — its formats, its dependencies, its network reach — as evidence, generically and without guessing. The deterministic tools extract the signals; the AI judges what they mean for this particular work. That separation is deliberate, so a report about your work isn’t shaped to fit someone else’s.
One view is made for you specifically: the artist’s hand — the artist’s own words gathered from the source and the decompiled code: their comments, their authorship and copyright lines, the contact URLs, the voice inside the work. The pass extracts the words; you read them.
Where to go next
- Curious what the analysis actually does? → The autopsy
- Ready to think about the AI key? → Bring your own AI
- Want to run the whole thing yourself instead? → Self-host quickstart
Or just get in touch with the ZKM team — bring the artwork, we’ll take it from there.