I watched Elroddd’s five-minute video, I Killed A Business Model, and immediately liked the directness of the solution.

The video starts with a three-second clip of General George Patton. Getty Images asks $400 for its lowest-priced licence, according to the example shown. The same shot sits inside a public-domain film on the Internet Archive, free to download.

That feels unfair. Getty Images adds search and delivery, but the price hides the fact that the underlying footage can already be free. Destockd solves the access problem without claiming ownership of the archive.

The archive was open but still hard to use

Public access does not always mean practical access.

A film might run for 20 or 30 minutes. Its title might say little about the shots inside it. Finding three useful seconds means loading the film, scrubbing through it, and downloading the whole file from a slow archive server.

Commercial stock sites have already solved this interface problem. They offer fast search, small previews, and individual downloads. Their convenience becomes a toll booth when they sell access to public-domain material without making the free source clear.

Destockd removes that toll booth.

This is a well-engineered answer

Destockd pulls HD films from the FedFlix collection. It detects cuts, splits each film into individual shots, and indexes those shots with CLIP, a model that connects images with text descriptions.

The pipeline is easy to understand:

archive film -> shot detection -> CLIP embeddings -> visual search -> direct clip download

That simple flow fixes the whole user journey. A person can search for fashion, computer, or factory without knowing a film title. After finding one useful result, the site can return 48 visually similar shots.

The result now contains more than 40,000 shots. It needs no account, subscription, paywall, watermark, or advertising. Every clip has a direct download and a link to its source film. Where available, it also links to the related U.S. National Archives record.

The source links matter as much as the search. They let a user check provenance and rights information instead of trusting a vague “royalty-free” badge.

One tiny shot shows the value

This two-second black-and-white shot from The True Glory is a good example. Destockd makes it available as one clip, then links to the restored source film and its U.S. National Archives record.

I did not need to know the source film before I found the shot. I can still inspect its history after I find it. That is what a good archive interface should do.

I want to test this idea for India

I would love to audit Indian archives and build a similar search layer if the footage and rights allow it.

There are promising places to start. The Films Division archive says it holds more than 8,000 documentaries, short films, and animation titles. The National Film Archive of India preserves films and related material. Prasar Bharati Archives has digitised large collections from Doordarshan and All India Radio.

India needs a rights audit before it needs a search box. U.S. federal government works are generally excluded from copyright protection there. Indian government works follow a different rule. Sections 26 and 28 of the Copyright Act, 1957 give cinematograph films and government works a 60-year term measured from the year after publication.

That means some Indian government films published in 1965 or earlier could be out of copyright in 2026. Each film still needs a source check. Embedded music, borrowed footage, unclear publication dates, and third-party ownership can change the answer.

The archives also use different access rules. Films Division sells archival footage. Prasar Bharati licenses footage for specific uses. The Press Information Bureau allows free reproduction of material on its site, with conditions and exceptions for third-party content.

So the first Indian version should be an audit, not a mass scrape:

  1. Build a source register for public collections.
  2. Record the owner, publication year, reuse policy, and source URL for every film.
  3. Review rights before processing any media.
  4. Split cleared films into shots and create visual embeddings.
  5. Keep the rights record beside every download.

This work could uncover an Indian public archive that is technically available but practically invisible. If the audit finds enough cleared footage, I would be excited to help build the search and delivery layer around it.

Destockd shows the standard: preserve the source, expose the provenance, and make every useful second easy to find.