TRAVEL 2040
← TRAVEL 2040
How it was made

One person, a few AI tools, and a human at every gate.

TRAVEL 2040 was made over about a week, mostly in spare time, deliberately unhurried, and much of it spent waiting for daily limits to reset: a book, thirteen posters, fourteen short films and this site. The AI did a lot of the making. The judgement stayed human.

~1 weekSpare time
200+Images, 92 used
174Book pages
14Short films
13Posters
≈ £65Extra spend
How it grew

It started as research

It began as research into where travel is heading. Like most things I work on, I wanted to share it, so it became a book of travel personas. Then it seemed more interesting to bring them to life.

Research

Where travel is heading, the technology behind it, and what’s already real.

A book to share it

Thirteen travellers and the journeys of 2040, as a coffee-table book.

Bringing them to life

Gemini joined to turn the book’s stills into moving scenes.

A campaign

A LinkedIn series and a set of YouTube short films, one story at a time.

This site

A home for it all, built on the same films, posters and stories.

One master, many formats

Each film was rendered once in landscape. Python then cropped and re-edited it for every other place it needed to go. Some scenes lost a little detail at the edges; that was a trade worth making.

16:9 masterYouTube and this site
4:5LinkedIn feed
9:16Event screens
The posters and the book were made separately: the posters as vector artwork, the book as print-ready pages.
The workflow

From a sentence to a finished book

1 · Research
Claude

The future of travel, the technology behind each story, and what’s real today.

2 · Images
ChatGPT

Photographs, plans and objects for thirteen journeys, from detailed descriptions.

3 · Book
Python, drafted by Claude

Layout, type and print files for KDP and Lulu, built by scripts.

4 · Short films
Gemini API

A Python script sent each approved still to Gemini to bring it to life.

5 · Edit
Python on a Mac

Clips, captions and music assembled by scripts, run locally.

6 · Site and posters
Claude

This site, the posters and the print files.

7 · Publish
James

Nothing went out without a final yes.

Alongside every stepA brand agent kept the style and the message consistent: the palette, the type, the tone, and the line between a vision and a forecast.
Human review gate: nothing moved to the next step without a yes
The tools

Why these three

Each tool did the job it was best at. One of them held the whole thing together.

Orchestration

Claude

The conductor. It researched the stories, wrote the Python that built the book and assembled the films, kept the brand consistent and built this site, working directly with the files on a Mac.

It kept the whole project in view, so a change in one place carried through to the book, the films, the posters and the site.

Claude Team plan
Images

ChatGPT

The photographer. It produced the most natural people and places, including warm family scenes, which mattered for a book about who we travel with.

Over 200 images were generated and 92 made it into the book. Every one was briefed in detail and checked against the story before it was used.

ChatGPT Pro plan
Motion

Gemini

The camera crew. Through its API, a script could send each approved still and get back a few seconds of motion that kept the scene as it was.

Pay-per-use meant paying only for the clips we asked for, and re-running only the ones that failed review.

Gemini API, pay per use
The cost

What it cost

Almost everything ran inside subscriptions that were already in place. The only extra spend was the video generation.

Orchestration, edit and siteTeam plan

Claude, within an existing subscription.

ImagesPro plan

ChatGPT, within an existing subscription.

Video generation≈ £65

Gemini API, including the clips that didn’t make the cut, and the ones we chose not to re-render.

Human in the loop

Gated, at the right moments

The tools were fast. The gates are what made it good. The research, every image, every page and every film passed a human review before it moved on, and most went round more than once.

Is it right?

The research was checked and challenged: what exists today, what’s plausible by 2040, and what isn’t.

Is it the story?

Each image was checked against the text: the right people, the right place, the right moment.

Is it on brand?

Colour, type and tone held to one style across the book, the films and the site.

Is it honest?

Every image is labelled AI-generated: a vision of the future, not a forecast and not a record.

Is it ready?

Nothing was printed, posted or published without a final yes.

What was hard

Where AI needed guidance, and where I had to compromise the vision

Families, yes. Films of children, no.

ChatGPT made warm, natural family scenes. Gemini won’t make videos from images with children in them, so those scenes stay as stills, or the dog and the grown-ups carry the moment.

New faces in old scenes

Both tools sometimes swapped people for entirely different people, mid-scene, between versions.

Hands and prompts

Hands that rotated in ways hands don’t, and prompts taken a little too literally. Every film was watched frame by frame before it was used.

Knowing when to stop

Not every flaw was fixed. Where another render would have cost more in credits than it added to the story, we accepted the edit and moved on. That was a human call too.

“Input blocked: Sorry, we can’t create videos from inputs containing photorealistic children.”

Story 05: the key boxes, and a hand we kept

Hannah leaves her phone in locker 14 at the ferry office. The key boxes took three goes to look like something an island harbour might really have, and one version gave her an extra hand. In the short film, the hand passing the key isn’t quite right. Another render might have fixed it, or broken something else. It stayed.

First draft: outdoor drawers numbered 10 to 18
Version 1Drawers on the harbour wall.
Second draft: an extra hand appears from behind the woman at the lockers
Version 2Inside now, but with a hand coming out of her back.
Final: Hannah holding the key to locker 14
FinalOnly box 14 has a number, and the hands are a little off. But it was sufficient.
Frame from the short film showing the key handover
Kept anywayIn motion, the hand passing the key goes a little strange.

Two jokes that didn’t make the cut

A running Budgy Smuggler joke

I have a slight Budgy Smuggler obsession, and I’m often seen in them. The plan was for every water scene in the book to feature a pair, as a running joke. After rendering them, I realised it was too much, and most of them went. The dare, a lovely moment of jumping into the river, stayed in the book, but had to stay a still in the film: Gemini wouldn’t animate it.

James on a jetty in a pair of Budgy Smugglers
Exhibit A: the real thing, on a jetty in the real world.
An advert homage, one take too many

For the Dome, the night swimmer was a nod to a famous fragrance advert, set in Germany rather than on the Med. I iterated far too many times. He stayed in the book as a bit of cheeky eye candy, and never made it into a film.

The limits we hit

Most of the week was spent waiting: for daily limits to reset, for clips to render, for uploads to go through. The actual making was the quick part. Planning around the ceilings became part of the schedule.

Gemini20 generations a day

Our plan allowed twenty a day, so the films were made in daily batches, best scenes first.

ChatGPTA daily image cap

Image generation was rationed by day too, which made every brief count.

ClaudeMemory, not quota

We never hit a usage limit, but long sessions were condensed to fit its working memory, and small details were sometimes forgotten. The files, not the chat, stayed the source of truth.

YouTubeA daily upload limit

Even publishing had a ceiling: the films went up over more than one day.

From the cutting-room floor

A reject, and a keeper

Asked to bring the family dinner to life, the model kept the table, the lights and the valley, and invented a different family. It didn’t make the cut.

The original family dinner scene from the book
The book: the Barnes family at one long table.
Rejected AI video frame with different people at the table
Rejected: the right table, the wrong family.

Other scenes worked first time. Here, a still from the book became a moving moment, with Aoife joining from Melbourne.

Still image of the hologram call
The still, from the book.
Frame from the animated hologram call
Kept: the same scene in motion, for the short film.

A few more that didn’t make it.

Rejected washing-line image showing a real swimwear brand
A real brand on the washing lineThe model printed a real swimwear logo on the trunks. Out it went.
Three frames from a rejected clip where a hand twists unnaturally
A hand with a mind of its ownThree frames from a rejected clip: the hand turns in a way no wrist should.
Revision trail

Again, and again, and again

The hologram call

Story 06 needed one quiet moment: Aoife joining the family from Melbourne. Getting there took a village party, a lookalike cousin and a lot of patience.

First draft: a crowded village party
Version 1A village party. Too many people, and where is Aoife?
Second draft: a hologram standing in a party
Version 2The hologram arrives, but it’s still a party.
Third draft: a quiet room with a lookalike
Version 3The right room, but a lookalike cousin.
Fourth draft: the family with Melbourne behind the hologram
Version 4The family, with Melbourne behind her. Close.
The final hologram call scene
FinalAoife, from Melbourne, where it was already morning.

The economy sleeper

Story 07 needed a believable lie-flat bunk in economy, and a poster plane that looked like a real aircraft. Both took more rounds than most.

An early version of the sleeper bunk image
An early draft: one bunk by the window. Beautiful, but not yet the idea.
The final economy sleeper bunks
The book: stacked sleeper bunks down an economy cabin.
Poster draft with the wrong aircraft
Version 1The wrong plane.
Poster draft with an aircraft missing an engine
Version 2One engine short.
Poster draft with engines on top of the wing
Version 3Engines on top of the wing.
The final What Disruption poster
FinalRebooked while you sleep.
The collection

Everything we made

Looking back

“I’m happy with what was created. It really creates a feeling for each of the stories. I hope this shows how a system is needed for repeatable and efficient operations. With the right foundations, you can do a lot.”

James Lever