Making history documentaries with AI: archives, maps and the rules that keep them credible

A guide to producing history documentaries with AI help: where the archive footage comes from, how historical maps get their borders right, what to never generate, and how editors' templates carry the genre's style.

History is the biggest documentary genre on YouTube and the one where AI is most useful and most dangerous. Useful because the archives are deep and public. Dangerous because a generated "photo" of a real event is a lie, and history audiences are unforgiving about it.

The rule of the genre

Never generate the past. Every image of a real event, person or place should be real: archive film, photographs, documents, maps, paintings. Generated footage belongs only in clearly marked reconstructions of things that were never recorded, and many history channels avoid even that.

Where the material is

• Newsreels and government film on the Internet Archive and national archives.
• Photographs from the Library of Congress, Wikimedia Commons and museum open-access collections.
• Paintings and engravings for pre-photographic eras (museum CC0 collections are enormous).
• Documents, letters and maps as images.
• Historical maps from the David Rumsey collection and national libraries.

See public domain footage sources.

Maps do the heavy lifting

Pre-photographic history is told with maps: borders for the year, movement of armies and trade, cities appearing. Correct borders matter; a modern base map under a 17th-century story is the most common error in the genre. Glisse builds maps from historical border data for the year named in the narration, with camera moves between the places mentioned and labels arriving on the words. See historical maps for documentaries.

The pacing of history

History documentaries hold longer than any other genre: stills for 5–10 seconds with a slow push, maps for 12–20. The audience is reading the image. Music is sparse and the narration is measured. A fast, cutty history film feels like a listicle.

Where AI helps

• Archive search matched to each narration line, with licences checked.
• Maps generated for the right year.
• Narration performed in a measured register with word timing for the cut.
• Sound: room tone and foley for silent film, the crackle that makes archive feel present.
• Re-cuts when the script changes.

Templates for the genre

History editors have recognisable styles: the paper-map channel, the archive-and-narration channel, the painting-and-document channel. In Glisse, those editors publish their films as templates, and a creator with a topic gets a history film cut in that editor's pacing and look, with real archive footage sourced for it. The genre's rule holds: real material first, generated only for what was never recorded.

FAQ

Can AI make history documentaries?

AI can source archive material, build accurate maps, narrate and cut. It should never generate images of real historical events.

What footage can I use for a history documentary?

Public domain newsreels, photographs, paintings, documents and maps from archives, libraries and museums, with the licence checked per item.

How do history channels make their maps?

With historical border data and animation tools, or generated in Glisse from the year and places in the narration.