◆
Building Ascension, told from the inside: decisions, mistakes and progress.
The Frost Wyrm is the first boss in Ascension with its full repertoire: six attacks, each with its own wind-up and recovery. There are seventeen versions between the first test and the video, and most of them fell over details nobody should ever notice.
In September I wrote that if the slime ever needed a new body, we'd change the body. Three weeks later it needed one. This is the second life of the game's first enemy, and what it cost to make a gel move like a gel.
A wolf doesn't run by speeding up its walk. Walking, trotting and galloping are different coordinations, and all three are animated separately. A hundred and twenty-eight animations later, the howl calls the pack — but only when it's alone and wounded.
The concept is from 8 June, the same small hours as Leo's. It waited four months for a production chain to exist, and once it did, it took a day and a half to be animated and fighting.
Every monster in the codex leaves a card: its own illustration, a frame that is interface rather than artwork, and text that says what happens, when, and where it stops. Two hundred and fifty-three of them, still a proposal.
Idle, walk, run, two-cut attack, jump, shield dodge, defeat. The full catalogue running inside Godot, at game scale and real speed, half of it driven by the controller rather than played back.
Two weeks ago I wrote that the right answer was to split the jobs: 3D moves, 2D draws. I was wrong. The answer was to stop patching.
For months Ascension's first enemy was a green hexagon with a red wedge. That is how its brain got built before its body, and how the body arrived afterwards: nine animations in eight directions.
The three-dimensional model was finished and still did not fit on screen. The way out was not choosing between 3D and 2D but splitting the jobs: 3D moves, 2D draws. Seventy frames repainted one by one for a single walk cycle.
It started as a sharpness problem and turned into the most expensive decision in the project: how tall the character should be on screen. Instead of picking by eye, I measured Terranigma and Eastward pixel by pixel.
A new line with its own finish, eight directions that are headings rather than sides, and a sword draw rejected once for deforming the arm despite passing technical review.
Elyndor is spread across three layers and each has its own drawn map. Here are the surface, the sky and the underdeep — and what they are for besides looking good.
Thirty-one seconds of paper, gold thread and candlelight. Ascension's world is not introduced with a map: it is introduced with a book that opens by itself.
From the 253-monster codex, ten from ten different families. What they are, how they fight, and why none of them has to be killed.
Rebuilding Leo in three dimensions so every new animation comes from the same source. It took three turnaround sheets, and the automated checks found two faults the eye does not catch: 79 clips floating and the chin pointing at the sky.
253 monsters, 22 dungeons, 110 weapons and 51 script events exist on paper before the game does. Here is the reasoning, and what it cost me.
Three months, three rebuilds and a dozen-odd revisions. Every attempt Ascension's protagonist went through, including the ones that did not survive, up to the design that stays.
The last pass over the citadel does not touch the finish: it changes the content. A stable and a henhouse to the west, grilles on the gates and seven open stalls where there were closed workshops.
The setting where the prologue begins went through a village, a full redesign and three finishing passes. This is how the work evolved, version by version.
The design taken to final art, front and profile, with the sword and shield integrated. From here Leo stops being a reference and becomes the source of everything else.
I decided that changing screens should mean turning a page. That single decision ended up defining pausing, saving, chapters and the voice narrating the whole game.
I generated Leo's dialogue bust from the game sprite, where the face is about ten pixels across. The generator invented a face, and it came out semi-realistic. The lesson forced the design that rules the face.
The design was there; what was missing was painting it at the resolution you play at. That is where the real work showed up: crisp stretches next to smeared ones, painted shadows the engine cannot move, and a list of one-off mistakes.
The problem was not the quality: it was that the enclosure had one gate, to the south. A setting can be beautiful and badly connected. This one was, and the world map gave it away.
With the measured pieces, the first complete enclosure was assembled, with physics, collisions, a camera and neighbours. It worked. And it still had to be scrapped.
The first attempt at the place where the game begins was not a setting: it was a kit of pieces. A wall, a tower, two houses, a well. And Leo in a corner, because everything is measured against him.
From concept to sprite. The system that turns a character into eight-direction sheets already worked, and Leo was the first through it: this is the Leo the prologue was first played with.
The concept everything started from, and the problem it took me a while to see: it was drawn for a still image, not for a game that has to animate it in eight directions.