Procedural vs. Handmade Content

tim-cain procedural-generation level-design game-design rpg fallout arcanum skyrim forever-games

Abstract

Problem: Should game content be handcrafted by designers or procedurally generated — and must it be one or the other?

Approach: Tim Cain draws on 25+ years of experience — from Arcanum's procedural dialogue opcodes to an unreleased 3D space exploration game called Universe — to argue for a hybrid approach.

Findings: Handmade content is finite but emotionally resonant; procedural content is infinite but often feels hollow. The solution is combining both: a handcrafted core with rich procedural systems that generate meaningful encounters, factions, conflicts, and even new mechanics. Cain calls these "forever games."

Key insight: Procedural generation shouldn't replace handmade content — it should extend it infinitely by creating emergent stories, relationships, and conflicts that arise naturally from systemic rules.

Source: https://www.youtube.com/watch?v=jkXd503iCRI

1. Terminology: "Handmade," Not "Bespoke"

Cain deliberately avoids the word "bespoke" because it implies content that is both handmade and unique — used exactly once. In reality, much handmade content is reused: NPC barks, house models, dialogue templates. "Handmade" is the more accurate and honest term.

2. Cain's History with Procedural Generation

This isn't a reaction to the AI hype cycle. Cain has been thinking about procedural content for 25–30 years:

2.1. Arcanum's Dialogue Opcodes

In Arcanum, Cain built a system where narrative designers could insert opcodes into conversations that would expand into entire dialogue subtrees — handling common interactions like bribing, bartering, asking for rumors, and greetings. The greeting system alone accounted for over a dozen contextual states.

2.2. Arcanum's Unrealized Random Encounters

Cain wanted Arcanum's random encounters to go far beyond "here's a pack of monsters in a tile sector." His vision included merchants, explorable ruins, and functional settlements — all procedurally placed and then permanently pinned to the world map once discovered. A new playthrough would generate entirely different discoveries.

2.3. The Road Problem

Procedurally generated roads between Arcanum's settlements were planned but never built. Hand-placing roads required artists to touch tens of thousands of tiles, and each modified tile saved a delta that bloated the world map file by megabytes — space they simply didn't have.

Universe: The Lost Space Game

Before Fallout, Cain prototyped a 3D space exploration game called Universe. It generated a 3D star map you could rotate and fly through, with procedurally generated solar systems (using a star system generator he built separately). The tech wasn't fast enough, and the project was shelved — but the ideas lived on.

3. The Core Argument

Handmade content is unique, original, and emotionally resonant in ways procedural content hasn't yet matched. But it is finite. Even a huge team working for a decade produces a bounded amount of content. DLCs add more, but it's still finite.

Procedural content is infinite — but players complain it feels like filler: random, repetitive, algorithmically transparent. Cain's answer: the algorithms need to be richer. They need to explore design space, producing outcomes even their creators wouldn't have anticipated. Research into AI-driven discovery has existed for decades; it's time to apply it to games.

4. The Vision: Applied to Fallout

Cain's most detailed example imagines a Fallout game with deep procedural systems layered on top of handcrafted content:

  • Generated locations: Caves, military bases, ruined towns, and functional settlements appear as random encounters and get permanently pinned to the map once discovered.
  • Generated NPCs and quests: Procedural towns have NPCs with personalities who give quests — find an item, rescue someone. These quests may lead to previously explored locations (repopulated) or entirely new generated ones.
  • Emergent events: You might arrive at a town during a raider attack and choose sides — help the town, help the raiders, kill everyone, or just watch.
  • Faction friction: Two generated towns that expand into each other's territory accumulate "friction points." When enough friction builds — contested farmland, kidnapped children, disputed sectors — war breaks out organically. The conflict has legible causes the player may have directly participated in.

5. The Vision: Applied to Skyrim

Cain sees Bethesda's radiant quest system as a starting point that could be taken much further:

  • Radiant quests leading to new dungeons: Instead of sending you back to cleared locations, the Thieves Guild could point you to a procedurally generated dungeon on a mountainside you've never visited.
  • Repopulation: Cleared dungeons get taken over by new factions — bandits move into a draugr ruin, or a wiped-out town gets resettled by nearby villagers.
  • Emergent threats: A powerful NPC warlord arises with diverse followers (archers, tanks, a wizard ally). A high priest of a Daedric cult builds temples on the outskirts of existing towns. A draugr ruin awakens a lich of unprecedented power.
  • Expanding the map: Existing caves can extend deeper. Offshore islands accessible via fast travel. Floating cloud castles that drift across the map with limited windows of access. Portals to new realms.

6. The Vision: Applied to Space Games

Drawing on his Universe prototype, Cain imagines space games (Starfield, Star Citizen) taking these ideas even further:

  • Getting paid to survey the galaxy for M-class worlds
  • Learning stellar cartography to narrow down promising systems
  • Discovering alien races that unlock new crafting tech trees
  • Encountering faction battles in deep space
  • Searching for extinct civilizations to salvage technology

7. "Forever Games"

Cain coins the term "forever games" for this hybrid model: games that ship with substantial handmade content and receive handmade DLC, but also include rich procedural generation across encounters, locations, NPCs, quests, factions, and even mechanics.

The promise: if you love Fallout or Skyrim, you can just play it forever. There will always be new discoveries, new conflicts, new ways to push your favorite systems further.

Cain hints at even more ambitious ideas — procedurally generating game mechanics themselves, like new perks — but saves that for another discussion.

8. Closing Thought

Cain believes games are "on the cusp of something big and amazing" with procedural generation. The technology and research already exist; what's needed is the ambition to build richer algorithms that create meaning, not just volume.

9. References