Making of
The Making Of
From the first sketch to the screens on these pages.
From the game's design documents and the history of its repository.
What was decided on the way
- The decisions so far are written down: Væ Victis, a gothic action game. The first art direction, title screen and status panel follow.
- Pre-rendered sprites: a rigged 3D model becomes eight-direction palette frames. The method every figure in the game still uses.
- The haven, "the cathedral under the cathedral", is built and playable.
- Wick's voice gets a stone-hall effect chain; all 117 lines are re-voiced.
- The graybox plan: the whole procession to the final boss, 37 new maps and 15 bosses, built flat first and played through before any art.
- Anonymous, self-hosted play stats. Then a premise card is added after the intro, because players could not tell the story from the opening.
- The rename: Væ Victis, with the Æ carved into the title, and its own home at vaevictis.vip.
- The King's Labyrinth Gardens, the Height and the Pleasance, first in graybox, then dressed.
- The inventory of every generated picture the game shipped: 194 keys. Then every one is redrawn in code or cut, and a guard (tools/no-ai-art.mjs) fails the build if one returns.
The method, in four rules
- Everything is code. The pictures, the music and the sound effects are made by scripts, with recipes (see the art book and the soundtrack).
- One palette. 32 colours, fixed in the game's core. A debug overlay counts what reaches the screen.
- Graybox first. A level is a text map in one file, playable flat before any art.
- Gameplay you do not own must not change. Each agent was given one package and wrote only its own files.
The turn: no generated pictures
Then it was decided that no generated picture would ship. Every one the game had shipped was removed or redrawn: 122 pictures redrawn in code and 45 no longer shipped; Wick's HUD portrait is now rendered from his own 3D model in each of his three skins. A guard script fails the build if a generated image returns. What is still synthetic is the voices, made with open-source speech models and kept as placeholders until they are re-recorded.