Objectives, triggers, puzzles
Level designers use scripting tools to program specific game rules and behaviors for a level. Some common examples of level scripting include:
- Triggers execute logic when entering an area, to lose health or gain points, etc.
- Gates block progress until the player solves a problem in the game world somehow
- Game objects / entities that move or react, like buttons, doors, elevators, trains, traps, etc.
(visual scripting example)
In past generations of game engines, there was a big technical difference between the game engine code vs. the scripting logic in a game level.
For example, Quake 1's game engine idTech2 was written in C / QuakeC, but level scripters relied on premade entities with narrow predefined behavior. A func_button entity could only function as a button, and a func_door entity could only work as a sliding object. If a designer built the door as a flat platform that "opened" upwards, then the door was basically an elevator; these resourceful hacks were necessary because level designers could not easily code new types of entities. Entities were level-specific visual scripting logic that existed as data interpreted by the game engine.
editor screenshot of a very complex entity setup (168 different objects!) for spawning weapons in Quake 1 by Benoit "Bal" Stordeur https://twitter.com/BafuBal/status/1261974130624167937
For modern game engines, scripting has become increasingly textual and code-like. Unreal Engine 4's Blueprint visual scripts match C++ functionality, Unity's Playmaker / Bolt visual scripts match the C# API, Fallout 4 / Skyrim scripters use a Papyrus language, and other engines routinely embed Lua. It's all basically code now.
So: a level script is simply a coded behavior limited to the scope of a game level. Scripts usually interact with specific objects or places in the map, binding to specific entities. They have limited functionality outside of this map or across different maps. Think of it as "softer" code unique to a given level, compared to the "harder" core game code shared by all levels across an entire game.
A scripting hack is a one-off brittle solution that uses a game system in a way that wasn't intended, and thus cannot support a wider range of behavior nor be easily reused elsewhere in the game.
For example, in Fallout 3's Broken Steel DLC, the player fixes and rides a metro train during a mission. However, the game does not have any actual vehicle or train system. Instead, Fallout 3 uses a clever hack to simulate a train. Via script, the game forces the player to secretly equip an inventory item called "DLC03MetroCarArmor" -- an armband that executes a camera script that makes the player feel as if they're on a moving train. In Fallout 3's GECK game editor tool, the armband looks like a giant train-shaped hat. (see image below)
This scripting hack may seem absurd, but it works, and it saved the developers a lot of time from needlessly engineering a complex feature. So sometimes hacks are appropriate, but you can only decide on a hack like this when you have done enough planning for your project scope. Here, the Bethesda designers had planned their DLC and they knew there was no point in coding a train system they would never use again, so instead they hacked a fake train and successfully shipped their project.
The Fallout 3 "train hat" DLC03MetroCarArmor bracelet, as previewed in the GECK editor tool. From
quests, missions... usually made of triggers and gates
Branching flags model
Fallout / Skyrim numbered quest stages (FSM?)
visualize as a flow chart, can get very complicated
triggers let us define volumes and areas, and apply behavior to those zones
triggers know when something enters, stays, and exits... a trigger is an EVENT
basic trigger is traps, deaths, lava, etc. or opens a door
for singleplayer, triggers should be invisible and seamless and magic, and sometimes just one-off triggers? ... dear esther, triggers used for ghosts and voice over...
complex triggers: dynamically resizing triggers like in Gone Home parents closet?
for multiplayer, triggers should NOT be invisible!!... bomb site areas in CS, capture points in Overwatch
A gate is something that impedes or blocks the player's movement through the level, usually along the critical path.
Despite their name, gates are not just doors. Almost anything in a level can function as a gate, because any obstacle that impedes progress or flow, either by a little or a lot, is a type of gate.
A hard gate is when progress is impossible without completing the encounter or puzzle. When we just say "gate", we usually mean a hard gate.
- Kill all monsters / boss to open a door
- Press all the buttons
- Collect the key
- Only the NPC can open the door, when they want to (e.g. out of combat)
Or chain all these together.... kill a monster to get a key, press a button to release a monster that drops a key, etc.
the player could bypass the gate without completing the encounter or puzzle, but they are heavily discouraged from doing so (e.g. too many monsters)
A bit more sophisticated and less hand-holdy, supports speed runners better... but might cause a cascade of confusion if player keeps bypassing gates without realizing, isn't sure what's happening, or the gate might be so soft that it doesn't even do anything
- (very soft) blocked sightline, hairpin turn (see typology)
- (soft) simple traversal / jumping puzzle, stacking boxes
- (softish) Valve-style slowly opening door gate
- (ranges from soft to game breaking) confusing layout, unclear how to progress (might end up feeling bad-confusing instead of good-confusing though)
pathfinding nodes, monsterjump, cover nodes, clip volumes, leashing
Cover hinting for NPCs and AI
editor view of cover node hinting markup for Gears Tactics by Splash Damage (from https://www.splashdamage.com/news/blocktober-gears-tactics/)
"The tall red boxes are standing cover (otherwise known as “full” cover, providing the most protection) and the smaller pink boxes are low cover (half-cover in-game). The yellow targets on each cover node show which angles a soldier can shoot from when placed in that cover node. The yellow curved arrows above some of the low cover nodes show that a soldier can vault over the cover node into a node on the adjacent side. Careful placement of cover nodes and strict adherence to the metrics is essential to create a clean, bug-free cover setup that works as players expect it to. If a piece of cover is too high then you will not be able to vault over it, and if a cover node isn’t placed correctly at a corner then the soldier occupying that node will be unable to shoot around the corner. Both of these examples create frustration for the player and must be avoided!"
many combat games also automatically generate markup and fine-tune by hand (e.g. Uncharted 4)
When we script NPCs to perform specific actions outside the scope of their normal AI, especially for in-game cutscenes and storytelling, then that type of scripting is called choreography. On large team projects, character choreography could be the responsibility of a scripter, animator, or dedicated cinematics artist.
Common choreography scripting actions include:
- NPC actor body control like looking, turning, walking, talking. This stage blocking choreography should hook into the same body systems as the main game AI, i.e. maintaining two different walking systems would introduce extra technical debt into your project.
- Animation control. Overriding the default idle stance or the current full-body animation completely. For example, if two actors hug once in your game, you should not engineer a dedicated hugging system -- instead, you should just temporarily override their animations with hugging motions.
- Props and attachments. Temporarily bind an object to a character joint, like attaching a hat to a head, or a cup to a hand.
- Timing delays. Waiting for X seconds, useful for dramatic pacing or pauses.
- Camera control. Force a first person controller to look at something, or spawn additional third person cameras with other camera angles or cinematography.
- Game state management. For systemic or combat-oriented games, it is best practice to disable the choreographed actors' combat AI and physics, as well as pause / block / delete nearby enemies and make the choreographed NPCs temporarily invincible. If you don't manage and clean the game state for the choreography, then your actors could unexpectedly collapse or die during the cutscene.
screenshot of Faceposer choreography tool used for Half-Life 2 (2004) with 3D preview and track-based timeline interface
Taylor Swope, QA lead for Obsidian RPG "The Outer Worlds" (2019), spent a long time trying to debug an NPC object interaction scripting issue https://twitter.com/_taylorswope/status/1205252714680045568
We released [...] a fix for the dreaded "the game thinks my companions are dead" bug, which I believe I spent more time investigating than I have for any other individual bug in my career.The gist of the bug was that, for some players, a companion quest would be marked as Failed in the quest log, with the reasoning that the companion was dead -- despite the fact that the companion was very much alive and well. One reason it was so hard to pin down is that it was impossible to tell when the bug actually happened -- all of the cases we had were essentially "hey something bad happened in the last ten hours and now my quest is broken".Investigating it involved figuring out the location of every script and line of code that could possibly make the game think that a companion was dead. The only logical culprit was a bit of scripting that runs when a companion's health reaches zero: if they're in the party, it waits for combat to end and revives them; otherwise it marks them as dead "for real". [But] the only place in the game when a companion is present but not in the active party is when the player is on their ship. The problem is, when companions are on the ship, they are undamageable. Eventually we figured out that "undamageable" does not mean "invulnerable" -- they can't take damage from attacks but can still get hurt from other things. One of those things: falling a great distance.The problem with that is that there are no spots in the player's ship that are high enough to result in a lethal fall. So now we had to figure out how companions were mysteriously ending up way above the level. I looked into tons of theories, including "maybe their height data is preserved when fast travelling from other maps" and "maybe a physics interaction between two companions causes one to rapidly accelerate upwards". (My personal fav was "what if a companion is standing right where a cow spawns in during a random event and they're launched into space". Was genuinely bummed when that theory didn't pan out.) By this time, the game had come out, and all hopes of this being a weird fluke only a couple devs would ever see were dashed, as players all over the place started reporting their companion quests failing.Eventually, an offhand comment in one user's review mentioned seeing a weird bug where a companion was "climbing nothing", and this comment led me to figuring the whole thing out.On the dev side of things, the system for NPCs interacting with the environment is called "furniture". Sometimes this is literal furniture, like sitting in a chair, but it covers everything from using a terminal to leaning on a wall. Somewhere deep in the complex beast that is the furniture system, we had code that disabled all NPCs from starting new furniture interactions if the player was in a conversation.The problem was that using a ladder is considered two different furniture interactions: one for getting on the ladder and starting to climb, and one to stop climbing and get off. So, if someone started climbing a ladder and the player entered a conversation before they stopped, they wouldn't be able to exit the ladder, and, well…. -- Taylor Swope (@_taylorswope) https://twitter.com/_taylorswope/status/1205252714680045568
Game engines tend to have visual track-based timeline tools for choreography like Timeline (Unity) or Sequencer (UE4). Scripters and animators setup different tracks and place events on each track. This way, the game can import the choreography as a game asset instead of code, which avoids any time-consuming re-compiles and simplifying the codebase.
Large projects with many conversational cutscenes (like RPGs) usually rely on semi-automated workflows for scripting their choreography. Their tools automatically generate basic cutscene scripts for every conversation, because it is not feasible to handcraft hundreds (or thousands) of short cutscenes. Then scripters and animators polish the most important scenes with additional choreography, and QA testers report any broken low priority choreography that needs specific attention. Voice over heavy projects will also need to adjust scene timing for every voiced language, and games with day-night cycles should automatically override global lighting with better lighting setups for cutscenes and close-ups.
If your project doesn't have or need specialized choreography tools, then it is still possible to choreograph directly with gameplay scripting code. Use something like yield (Unity C#) or Wait nodes (Unity Bolt) or Delay / Timer nodes (UE4 BP) to call functions in sequence with your desired timing. This direct method is useful when you need full access to all scripting functions as part of your choreography. Imagine choreographing 100 NPCs walking through a door one-at-a-time; this task would be slow painful tedious work in a visual tool, but trivial in a scripting language with for() loops and arrays.
screenshot of Cutscene Editor used for Dragon Age Origins (2009) with 3D camera preview and track-based timeline
For their large open world RPG The Witcher 3 (2015), developers CD Projekt Red divided their quest workflow into four phases: (1) Writing the script and choices with a screenplay format, (2) Quest scripting the possible branches and game states in a visual scripting flowchart tool, (3) Dialogue and choreography with preliminary auto-generation pass, and (4) Post Production with fixes and visual polish for cutscenes.
The dialogue designer tool's automatic cutscene generator is especially interesting here. It takes the voice over audio and automatically generates tracks and timings for actors, with semi-randomized animation variations and standard camera shots that follow basic cinematography patterns like shot-reverse-shot and 180 degree rule. The result is a basic starting point for a cutscene that is already mostly done, allowing the choreographer to focus more on adding details and mood rather than boring rote work to setup fundamentals.
diagram of The Witcher 3's cutscene generator tool, from GDC 2016: "Cinematic Dialogue in The Witcher 3"
While the weaknesses of this method weren't really mentioned in CDPR's GDC 2016 talk, we can imagine the main drawback here is that it requires you to finalize and systematize much of your script, voice over recordings, and animation set in order to feed it into the generator. Without tagged audio assets and near-finished characters, it would be difficult for the generator to understand what to do. As with any procedural generation system, it is heavily dependent on its inputs, and without good meaningful input it will just result in the "oatmeal problem."
Other interesting details about CDPR's choreography workflow:
- The game was scripted and choreographed in English, but the length of the voice over audio would obviously vary with each language. So for every localization, the choreography and camera shot sequences are automatically sped-up or slowed-down; for example, French choreography runs 16% slower than English. [37:46]
- To stop NPCs and wildlife from interrupting cutscenes, choreographers blocked out an invisible collision box / clipping trigger volume / "deny area" to push non-choreographed characters away and keep the camera shot clear. [40:25]
video showing the The Witcher 3 cutscene generator tool, with two randomly generated variations on the same dialogue scene
Scripting is a creative problem-solving task that is similar to any other level design task. First you want to understand the problem, and then you want to try various approaches.
- 1.Do a little planning and research.
- 2.Prototype the most basic version and playtest.
- 3.Gradually iterate and elaborate on the script until it meets your needs.
As with any creative task, a little bit of planning helps a lot. Before you open the scripting editor, take a few minutes to define what you want your script to do. Some examples of planning documents:
- Scripting an object or behavior? Sketch how it should look and its various states / properties.
- Scripting dialogue or a cutscene? Write a script and draw a storyboard.
Next, research various ways of implementing the feature by googling. When googling for programming or technical scripting help, make sure to:
- Be technically specific. Include the game or engine name, as well as the scripting language. Surround both in quotation marks to require these terms in the search results.
- Be conceptually general. Describe the desired behavior in generic terms. For more complicated features, break the problem into smaller simpler parts and google separately.
Some examples of search terms:
how to code a ceiling fanis vague in the worst way (which code language?) and specific in the worst way (you're hoping someone else has specifically built a virtual ceiling fan)
"unity" "c#" spinning object codeis technically specific (we only want Unity C# code) and helpfully generalizing (a ceiling fan is a type of spinning thing) and we can extrapolate from the generic example for our specific use case
game dev scripting objectives systemwill maybe tell you someone's opinion about how to architect an objectives system, but without any specific help. What type of game objective do you want to script, exactly? No search engine can design your gameplay for you.
- What if we breakdown this problem into simpler and more specific steps, and google for these smaller steps separately?
"ue4" "blueprint" display text on screenhelps us figure out how to notify the player about an objective
"ue4" "blueprint" player enters triggerhelps us learn how to detect when the objective is completed, when the player has reached a certain location in the level, etc.
When to ask for help. Unless you're totally lost, avoid asking other people for help at this very early stage. Try to do it yourself with what you've found already, and see how far you get. Asking someone else works better when you have more specific questions or have some existing scripts to troubleshoot. To get a useful answer, you must first understand the problem.
After you've done some searches, you now have a better understanding of the problem and you are ready to prototype the feature in the scripting tool.
ask community members for advice. Join an engine-specific forum, website, or Discord, and ask your question. If possible, recap the research you've already done and include any error messages. Here's an example question you could post:
"I'm trying to make a working carnival carousel. I found this C# code to make an object spin (link) but I when I try
Usually involves a trigger to spawn / aggro enemies, and a gate to force the player to deal with enemies
Before you begin scripting:
- do some encounter design on paper. Have a plan before you open the editor, even a basic plan. what weapons or abilities will the player have for this encounter?
- know your combat metrics. What are the ideal enemy engagement distances? is the blockout too big or too small?
Then, place some AIs / NPCs / enemies and start playing with them
- how does the NPC react to the existing space? do they do anything you want to prevent with hinting?
- what tools are available to you? static vs. scripted vs. dynamic?
- Scripts are level-specific behaviors designed with a lightweight programming language, often with an engine-specific visual programming tool.
- Common features to script include objectives, choreography, triggers, gates, and AI.
- When googling, include the engine name and search with generic terms.
- For short term small projects (< 1 month) the simplest and fastest hack could be good enough.
- For long term big projects (6+ months) a more stable systemic method is better.
- UE4 Blueprints From Hell is a Tumblr archive of messy / hacky visual programming graphs authored in Unreal Engine 4's Blueprint system. It's fun but also a bit melodramatic; sometimes shipping an inefficient script makes more sense than engineering a time-consuming "better" design. If a cleaner method is 0.0001 ms faster but took twice as long to make, then it's arguably not a better script.