How we work, and how to check it

How a source is verified

A credit line is not a licence. Before an image enters a film we record it at item level, and the checks are deliberately mechanical, because judgement is the thing that drifts.

Frame-match, not name-match

We confirm the exact image we hold is the exact image the licence describes — not merely that a file with a similar name exists somewhere with a similar credit.

The licence is re-read at build time

On our most recent documentary, every licence was fetched live from the source archive as the film was assembled and compared against the record made when the file was first downloaded. A licence can change after you save a copy of it.

Every file is hashed

Per-item SHA-256 and byte size, so the file used in the film is provably the file that was cleared, and a master can be checked by anyone who has it.

Resolution is recorded, never manufactured

Pixel dimensions are part of the record. We do not upscale historical material or invent detail that was never photographed. Where an image is small, it stays small — used as an inset rather than enlarged into something it is not.

This standard is applied in full to our most recent work. Older films in the archive were made before parts of it existed, and their records are being brought up to it rather than described as though they always met it.

The three registers

Every visual carries one of three labels, and the label decides where it may appear. This is enforced by the record itself: a generated image cannot enter an evidence gallery, because the gallery is built by excluding that register.

• Authentic evidence

  A photograph, document, engraving or object that exists in the world. It carries its creator, its collection and its licence. This is the only register that may appear in a film gallery.

• Studio schematic

  A diagram, map or chart we drew to explain something. It is our illustration of the evidence, never evidence itself, and it is labelled as ours.

• Generated

  Imagery produced by a generative model. It is disclosed as generated wherever it appears, on screen and in print. It is never evidence, and it may never supply an answer the record cannot.

What we refuse to use

Some material cannot be used, and the reasons are recorded per film alongside the material that could. A licence that forbids commercial use, a licence we cannot verify, an image whose terms are not written down anywhere we can point to, or a reconstruction that would imply something the evidence does not support — each is barred and stays barred.

One of our developed subjects was stopped entirely because every route to its core imagery was closed. The film was not made. We would rather lose a subject than publish an image we cannot stand behind.

What the record cannot say

Every film page carries a section for the questions its evidence does not answer — what is contested, what was never recorded, what remains undetermined, and what simply has not been investigated. Each is a category we track, not a phrase we reach for.

We publish it high on the page rather than at the bottom. An unresolved question is part of the evidence, not a gap in it.

Corrections

If we have got something wrong, we would rather know. Tell us what the film or the page says, what the evidence actually shows, and where that evidence can be read. We will check it against the record and correct it if you are right.

A public log of corrections is being built. Until it exists we are not claiming one.

How to reach us →

Rights and permissions

If you hold rights in material we have used and something is wrong — an attribution, a licence, a term we have misread — contact us and we will correct or remove it. If you are an institution considering whether to license material to us, our per-item records are the answer to most of what you would want to ask.

Rights enquiries →