AI Geography Video Generator

Documentary stock footage is the default here and the cheapest look, at about 12 credits a video: licensed clips of real coastlines, mountains, rivers and cities, with nothing to attribute. It suits country facts and physical geography, where the places exist and have been filmed. A clip shows a real place, though not necessarily the one the narrator names, and when no clip fits a line the pipeline falls back to an AI image, so check every scene before you post.

Mixed combines stock footage and AI images, typically about 14 credits a video, for videos that move between places a camera has filmed and things it can't, such as the inside of a volcano or the layers of the Earth. Its AI scenes can show a real place in a realistic image, which matters for the labels below.

Miniature World renders each scene as a tilt-shift toy diorama, made for cities, economics and logistics: how a port city works, why a town grew at a river crossing, how goods cross a border. Nobody takes a toy model for footage, so it can show a real place without passing for a photograph of it. About 25 credits.

Retro Poster turns every scene into a mid-century screen print, the look of old travel posters: one landmark, coastline or skyline per scene, which suits a country profile or the story of a name. Keep names and numbers in the narration, since lettering in an AI image can come out garbled. About 25 credits.

Flat Vector is the clean science-explainer look, for physical geography: why rivers meander, how a mountain range leaves a desert in its rain shadow, why the Earth bulges at the Equator. Its scenes are illustrations, not measured diagrams, so check that each one shows what the narrator says. About 25 credits.

Vintage Archive renders each scene as an old black-and-white photograph, for historical geography from the 1850s to the 1960s: a boundary survey, a canal being dug, a city before its skyline. It's the look most easily mistaken for a real archive picture: an AI 'photo' of an 1875 survey crew is a new picture, not a found one. Keep faces out, and see the labels below. About 25 credits.

Miniature World, Retro Poster, Flat Vector and Vintage Archive can also turn each image into a short AI video clip. That raises a typical video to about 464 credits, so keep it for a series' strongest episodes; documentary footage and Mixed don't have that option.

"Which mountain is the highest on Earth? NOAA gives three right answers: Everest is highest above sea level; Ecuador's Chimborazo, near the Equator where Earth bulges, has the summit farthest from Earth's centre, over 2 km farther out than Everest's; Mauna Kea in Hawaii is tallest from base to peak, more than 10,210 m. Documentary look, calm narrator, say which measure each answer uses."

"Is the Four Corners monument in the wrong place? 2009 news reports said 2.5 miles off; NOAA's National Geodetic Survey said it marks the exact spot where the four states meet. In 1875 surveyor Chandler Robbins set it where Congress put the line, 32° west of the Washington meridian, not 109° west of Greenwich. Vintage Archive look, survey crew from behind, end on 'In surveying, monuments rule!'"

"Why is Canada called Canada? The Government of Canada says the name likely comes from the Huron-Iroquois word 'kanata', meaning 'village' or 'settlement'. In 1535 two Indigenous youths told Jacques Cartier the way to kanata: Stadacona, where Quebec City stands today. It became an official name in 1791, with Upper and Lower Canada. Retro Poster look, ships on a river, cliffs, no people."

Each fits the 400-character limit of the box above, and each rests on an official source you can name in the post text: NOAA for the mountains, NOAA's National Geodetic Survey for the Four Corners and the Government of Canada for the name. Write the figures into the prompt yourself, keep the source's own hedges ('likely', 'more than'), and ask the narrator to say which measure a number uses.

Every figure in a geography video needs a source and a date: an area, a population, a height, the length of a river or a border. Take it from whoever measures it, such as the country's national statistics office or national mapping agency, the UN or the World Bank, and say whose figure it is and which year it describes, in the narration or the post text.

Don't lean on the CIA World Factbook. For decades it was the default reference for country facts, and much of what's online still quotes it, but on 4 February 2026 the CIA announced that The World Factbook 'has sunset'. When a figure traces back to it, replace it with the national or international figure and its date.

Numbers differ by source and by definition, and coastline length is the classic case. NOAA's official value for the US shoreline is 153,646 km (95,471 miles), and NOAA itself says shoreline numbers 'can vary depending on how the shoreline is defined'. Its figure counts the outer coast, offshore islands, sounds, bays, rivers and creeks up to the head of tidewater, or to where tidal waters narrow to 100 feet, measured on charts in 1939–40, with the Great Lakes measured in 1970. Area works the same way: Norway's mapping agency, Kartverket, gives 323,807 km² for mainland Norway and 385,206 km² for the whole kingdom, Svalbard and Jan Mayen included. And as the first prompt shows, 'highest' depends on whether you measure from sea level, from Earth's centre or from base to peak. Say which figure, which definition and whose, and say 'about' when good sources disagree.

Keep the source's hedges, too. The Government of Canada says the name 'Canada' likely comes from the Huron-Iroquois word 'kanata'; a script that drops 'likely' claims more than the source does. Many place-name stories are told more confidently than the evidence allows, so if no official or scholarly source backs an origin, call it a story or leave it out.

Never present a disputed border or territory as settled, in the narration or in a scene. Where countries disagree, say so neutrally, for example that one country administers an area and another claims it, or choose another topic. A map takes a side even when the narration doesn't, which is one more reason not to use AI images as maps.

Settled borders have their own logic. Congress set the Four Corners on paper, a surveyor set it on the ground, and NOAA's National Geodetic Survey explains that once a boundary monument has been established and accepted by the parties involved, 'the location of the physical monument is the ultimate authority in delineating a boundary'. When a claim says a border or a marker is in the wrong place, check the survey record before you repeat it.

YouTube requires creators to disclose AI-generated or altered content that seems realistic, including content that 'alters footage of a real event or place' or 'generates a realistic scene that didn't actually occur'. Two of its examples could have been written for geography channels: 'AI generated extra footage of a real place, like a video of a surfer in Maui for a promotional travel video', and 'a realistic depiction of a tornado or other weather events moving toward a real city that didn't actually happen'. You answer Yes under 'AI use' when you upload; for videos ShortFast posts for you, the setting is in YouTube Studio. Non-realistic content needs no disclosure, and neither does using AI for the script or the captions. YouTube says: 'Disclosing AI content alone won't limit a video's audience or impact its eligibility to earn money.'

So check what each scene is. Real stock clips show real places as they are. A realistic AI image of a real place, whether a Mixed scene or a footage video's AI fallback, is the case YouTube's example describes, and so is a Vintage Archive 'photograph' of a real survey or a real city: answer Yes. Miniature World's toy dioramas, Retro Poster's screen prints and Flat Vector's illustrations are plainly not footage. YouTube's monetisation policies also list 'content that uses deceptive or misleading imagery or narratives, such as realistic visuals tricking viewers into believing a fake celebrity death or natural disaster has occurred' among what can't earn, so an AI flood over a real city, shown as if it happened, fails twice.

TikTok requires a label on AI-generated content that shows realistic-looking scenes or people, and defines those as content that looks like it could be real, 'such as AI-generated images that look like real photographs or footage'. Artistic styles, like anime, and generic text-to-speech narration don't need one. TikTok encourages labeling AI-generated content and says the setting won't affect distribution, so turning it on costs nothing.

TikTok's misinformation rules reach into physical geography. Its Community Guidelines don't allow 'misinformation that denies the existence of climate change, misrepresents its causes, or contradicts its established environmental impact', or misinformation that incites panic, 'including falsely presenting past crisis events as recent', and even with a label they don't allow AI-generated content that misleads about 'a crisis event, like a natural disaster or conflict'. Two more rules keep content out of the For You feed: 'misuse of authoritative sources to push misleading conclusions', such as quoting selected data against what a study found, and unedited media 'presented out of context' in a way that may mislead about a topic of public importance, which is what a clip of one flood becomes if the narration names another. Conversations about climate and weather are allowed 'as long as they don't deny or misrepresent scientific consensus'.

Labels and monetisation are separate questions. YouTube's 'Generic or Repetitive Content' policy 'includes content that looks like it's made with a template, or that may feel repetitive to viewers after watching several videos in a row from the same channel', and a daily country-facts series regenerated with a new country name and little else is that risk. Its examples of what can't earn include 'similar or repetitive content with low educational value, commentary, narratives, or minimal variation across videos', and a script that only reads out a statistics page matches a reused-content example: 'content that exclusively features readings of other materials you did not originally create, like text from websites or news feeds'. A recurring format is fine: the allowed list includes the 'same intro and outro for your videos, but the bulk of your content is different'. Give each country its own question and its own sources, and the bulk is different.

It doesn't draw accurate maps, and it doesn't animate them. AI images of maps and flags often come out wrong: a border in the wrong place, a coastline that doesn't exist, a flag with the wrong pattern. Don't use AI images as maps. Describe a border or a route in the narration, which the captions show word by word, and if a scene needs a real map, upload one you made or have the rights to use.

Stock footage shows real places, but not necessarily the place the narrator names: a mountain range, a coastline or a skyline can come from somewhere else. When no clip fits a line, the pipeline falls back to an AI image, so a footage video can still contain AI scenes. Check every scene, and swap any that shows the wrong place.

It doesn't check facts. ShortFast writes a fluent script, and fluent isn't the same as current: a draft can repeat an old population figure, a superseded height or a number from a retired source. Nothing in ShortFast checks a figure, whether it writes the script or narrates yours, so check each one before you render.

The narration is fixed once rendered. You can swap scenes, edit captions, change music, title and post text, and re-render, but a wrong figure in the narration means generating again. The only text ShortFast adds is the burned-in captions of what the narrator says, in one of 9 caption styles and 18 fonts: no map labels, data cards or source cards, so say the numbers and the source aloud.

An automation has no approval step. In a country-facts series the AI picks each new place and writes its script, and every episode posts automatically or arrives by email, so choose email delivery and check each episode's figures and scenes before you post it. The next two episodes are queued: you can edit their captions, scenes and post text, re-time, pin or delete them, but not change their script. The series brief, up to 500 characters, is where your rules go: one question per place, the kind of source to name, no disputed borders.

None of these channels use ShortFast; they show what geography audiences come back for. RealLifeLore ('Answers to questions that you've never asked') builds most of its videos on a 'why' question, such as 'Why The Nepal Flood Was WAY Scarier Than You Think', and its recent thumbnails are relief maps with a country or a border highlighted and a few words of text. Wendover Productions says it is 'all about explaining how our world works', from travel and economics to geography, in titles such as 'Why Botswana Got So Wealthy' and 'How the Channel Tunnel Works', with relief maps on its recent thumbnails. Both lean on maps, which ShortFast doesn't make. Geography Now says it is the first and only YouTube channel to have covered profiles on every country, alphabetically, and it's presenter-led, which a faceless video can't copy: its host is on the thumbnails, beside a country's outline or its flags. Half as Interesting ('Education-y explainer videos that are almost good enough to watch') gives one odd question a whole video, as in 'What is the World's Biggest Hole Humans Have Ever Dug?'.

What they share is one clear subject per video, a place or a question, and an answer you can check. That is the part to copy; the scenes, the voice and the captions are the part ShortFast handles.