Most of what gets written about commissioning a game describes the process from the outside: brief, art, build, delivery. This is the other version. One build, the most detailed we have made, and the problems that actually had to be solved to make it hold together.
It is called Edward & B.Bear, and it was made for a grandfather's birthday. A boy and his teddy cross a painted London by tube, and at the end of it they fly to Cape Town, because that is where the man it was made for flies in from.
A note on why you cannot play this one here. The finished file is twenty megabytes, which is exactly right for a family opening it on a link and wrong for a page somebody might only be reading. So what follows is the record of it: the artwork, the motion, and the problems that had to be solved. The arcade has two smaller builds you can play in a click.
The shape of the thing
Twelve tube stations. Three interlocking lines. Eight kinds of London street, each painted several times over. Two airports and a turning globe. Sixteen keepsakes hidden along the walks, a satchel to keep them in and a notebook that fills with clues.
All of it arrives as one HTML file: 502 painted images and eighteen pieces of music, inlined, about twenty megabytes. Once it has loaded it asks the network for nothing at all. That matters more than it sounds. A game handed to a family on a link should still work in five years, on a laptop nobody has updated, in a house with bad wifi. Anything that reaches out to a server is a thing that can be switched off.
The map was a tree, and a tree is boring
The first version had a network that looked fine written down and was miserable to play.
Camden, Lambeth, South Bank, Rivergate and Shoreditch each had exactly one destination. Arriving at any of them left you nothing to do but go back the way you came. Two stations, two walks, over and over.
The fix was to stop thinking like a diagram and start thinking like a tube map. There are now three lines that each close into a loop, and the rule is that every station sits on at least two of them, offering at least three ways onward, with the whole map two or three hops from anywhere.
That is a design problem, not an art problem, and no amount of beautiful painting would have rescued the version before it. It is the clearest example of why the brief matters more than the asset list.
Eleven stations were sharing eight walks
The next problem was the one underneath it. Euston and Westminster showed the identical six paintings. So did Embankment and South Bank. So did Waterloo and Rivergate. Arriving somewhere new and recognising the street you walked ten minutes ago is the fastest way to make a world feel small.
So a station stopped being a location. It now draws a walk of six rooms from its own kind of street plus a second kind nearby, shuffled on a seed made from the station itself. Every station gets a different six in a different order, and it is stable: King's Cross is always the same King's Cross.
Turning a corner from a Georgian terrace onto a market street is what walking across London actually feels like, and that is the effect the shuffle is buying.
Rooms, not one long strip
The instinct with a side-on game is to stitch the paintings into a continuous belt and slide it past. It never works. The joins show, and every painting ends up serving a job it was not composed for.
Each painting is instead its own room. You walk across it, reach the edge, and step through to the next. There are three ways a room can be drawn, and the choice is made per painting:
- Direct. The painting is the room. It fills the screen and barely moves, and you walk across the floor it already contains. Right for station interiors and short intimate scenes.
- Parallax. The painting sits back and drifts slowly, with a consistent foreground of pavement, kerb and railings running in front of it at full speed. This is how a scene gets longer without inventing a stitch.
- Depth. The path recedes, so the pair turn their backs and walk into the picture.
Nobody stands in the painting
This is the trick that took the longest to arrive at, and it is the reason the wide shots work.
A painted street in one-point perspective has a floor that recedes. Put a character on a line halfway up that floor and they are walking in the air. Every attempt to place them in the picture failed for the same reason.
So they do not stand in the painting at all. A path is drawn across the front of it, nearer the camera than anything painted, and that is the ground they walk on. The painting falls back behind it. Two numbers do the whole job: where the painting's floor stops and the path begins, and where the boots go down that path, with paving visible in front of them and behind.
Ground lines, measured by eye
A related and duller problem, solved the dull way.
Every painting needs a ground line: the height, as a fraction of the picture, where boots belong. Not where the pavement technically starts, but a believable place to stand on it.
An automatic detector was tried and abandoned. It kept finding kerbs, slab joints and the dark lip at the very bottom of a frame, putting the line at 0.96 and dragging the whole painting up to five times its proper size. So every one of them was read off a ruler laid over the picture, by eye, one at a time. The canal scene that is almost entirely water has its line up at 0.70, on the far bank, so the water falls behind the towpath drawn in front of it.
There is no clever version of this. Somebody looks at each painting and decides.
The walk cycle belongs to the ground
A small thing that made everything feel wrong until it was fixed.
The animation originally picked a frame from the character's running clock multiplied by a frame rate, and that frame rate was recomputed every frame from current speed. Once the clock had been running thirty seconds, nudging the rate from 9.00 to 9.06 moved the result by nearly two whole frames, instantly. Since speed changes on every frame of every acceleration, the legs raced, stuttered and changed tempo for no reason.
Now a walk cycle covers a fixed distance, about a body height, and the phase advances by how far the character actually travelled. Tempo is exactly proportional to ground speed, the feet never skate, and no frame can be skipped. It is four lines of arithmetic and it is the difference between a character who walks and a puppet being dragged.
The platform could not be a painting
Every platform picture in the folder was a one-point perspective looking down the length of the platform. A tube train seen from the side cannot live in one of those. Drawn over the picture that already had a train arriving at an angle, you got two trains about to collide.
So the platform is the one underground shot built rather than painted: assembled side-on from the side-on art that already existed. Tiled wall with its poster and bench, the MIND THE GAP arch, the columns, the roundel, the platform edge with its yellow tactile strip. Flat wall behind, train along it, yellow line, paving in front. The four line groups tile their platforms in different colours so one stretch of wall does not have to serve every station identically.
What the mouse is for
The game needed a way to tell you where to look without a quest marker, which would have been wrong for the whole thing.
A clockwork mouse carries the letters. He whirrs up, stops at Edward's boots, and hands over a folded note: down the mews behind the red doors, a glint between the cobbles. That clue is written into the notebook, and the notebook is one half of a spread you open with a single key. The satchel is the other half: sixteen slots, filling as you find things.
The cast of soft toys around him was cut hard. A bunny, a corgi, a pillar box and a sock puppet all sat outside the world the paintings build, so they went. The plush pigeon went too, because the real pigeons already did that job better. What survived is a clockwork mouse and a knitted double-decker, which are things you could believe were dropped on a London pavement.
The last chapter leaves London
Everything above is scaffolding for the bit that matters.
Grandpa lands at Heathrow, having flown from South Africa. He scoops both of them up at once, and from that point he leads. The three of them walk out to the aircraft, board, and the flight crosses a turning globe from London to Cape Town, with the aircraft rotated onto the tangent of the route rather than flipped, so it always points where it is going. Table Mountain is waiting through the glass at the other end.
None of that is a game mechanic. It is the reason the game exists, and every problem above was worth solving only because it gets you to this.
What this says about commissioning one
Three things worth taking from it.
The brief decides more than the budget. The two largest improvements in this build were a network redesign and a shuffle, both of which came from playing it and noticing it was dull. Neither needed a single new painting.
Scale is measured in places, not hours. A single level needs one setting to be convincing. This needed twelve stations, eight kinds of street, two airports and a globe, and that is where the cost of a build like this actually sits.
The personal detail should be load-bearing. The flight to Cape Town is not decoration on the end of a London game. It is the destination, and the whole journey is arranged to arrive there.
If you want the shorter version of how a commission runs, the process guide covers brief to delivery, and the cost guide covers what moves the price. If the person you are buying for is a child with a favourite toy, turning a teddy into a game is the closer fit. And the arcade has two smaller builds you can play right now.
