There’s a version of this post that starts with a rant about Claptrap. We’ll get there. But first, a principle.
Good game VO — the kind that actually lands, the kind players remember years later — isn’t really a writing problem. It’s an attention problem. And attention, it turns out, is one of the most fragile resources in game design.
Attention Is a Budget, Not a Freebie
Players will tune out any VO line that isn’t either immersive (ambient chatter, flavor, world-building texture) or genuinely useful (telling them something they couldn’t already figure out themselves). The moment a line points out something obvious, it starts burning trust. And that decay isn’t linear — it’s closer to exponential.
The first time it happens, a player might barely notice. Maybe it even feels helpful. By the third or fourth time, they’ve quietly learned to ignore that voice altogether. And once that happens, it’s very hard to earn their attention back — even when a line is useful.
This is the central challenge of companion VO, assistant characters, tutorial voices, co-op callout systems — anything that speaks to the player unprompted. You are spending from a budget every time you fire a line, and that budget does not refill easily.
Two Jobs, One Voice
Most VO characters in games are actually trying to do two very different things at once:
1. Utility — surfacing gameplay information at the right moment
2. Personality — providing banter, color, the sense that there’s a real character in the room with you
Both of these are valid and valuable. But they compete for the same resource: player attention. And they fail in opposite directions, which means they need opposite design rules.
For Utility Lines: Consistency Over Cleverness
If a piece of gameplay information is worth calling out, it needs to fire every time its trigger condition is met. Not most of the time. Not randomly. Every time, or never.
Randomizing whether a utility callout fires is a quiet disaster. The player can’t build a mental model of when to trust it, so they stop trusting it at all. You’ve turned a reliable tool into a lottery ticket. The fix isn’t better writing — it’s better trigger discipline. Specificity and reliability are the whole job here.
The failure mode for utility VO is stating the obvious — and the result is exponential trust decay. Increase specificity, reduce the frequency of low-information callouts, and never randomize the ones that remain.
For Personality Lines: Scarcity Is the Point
This is where Claptrap comes in.
Claptrap is, line for line, pretty funny. The writing is sharp. The performance is committed. And yet Claptrap became one of the most commonly cited examples of an annoying game character — not because the lines are bad, but because there are too many of them. Frequency is the enemy of impact. A character that talks constantly, however charming, will wear out its welcome fast.
Contrast that with something like the companions in Baldur’s Gate 3, who mostly talk only when it’s earned — either the moment specifically calls for it, or the player deliberately seeks them out. That restraint is a big part of why the attachment to those characters runs so deep. Or Silent Bob, who works almost entirely because he doesn’t talk.
Scarcity makes lines land. Protect it.
Three Moments Where Banter Actually Works
When personality VO does fire, it should be aimed. There are three kinds of moments where a line has the best chance of landing rather than feeling like noise.
1. Perfect Moments — Dramatic or Comedic Timing
These are moments where the emotional tone of a line already matches what the player is feeling — and where the player has little or no active agency to lose. The classic example: the *SNAAAAAKE!!* callout on player death in Metal Gear Solid.
It works for three reasons. The player is already frustrated — the line’s tone matches for free. The player has no agency left in that instant (they’ve already died), so the VO can’t be blamed for costing them anything. And there’s enough variant pool that it stays fresh.
The design implication: look for moments where control is already taken away from the player — deaths, failure states, forced beats, respawns. These are low-risk, high-reward slots for character-driven lines, because there’s no gameplay cost to spending the player’s attention there.
2. Perfect Mind-Reading — Predictable Peaks
These are moments predictable enough that you can pre-load a line that closely matches what the player is very likely already thinking or feeling.
In linear content this is easier — you know the scripted explosion is coming, so you queue the reaction. In systemic or multiplayer contexts it’s harder, but still achievable by hooking into statistically rare, high-drama events: a multi-kill streak, a last-second clutch play, a near-death recovery. The trigger doesn’t have to be explosive or spectacular — it just needs to be infrequent enough, and dramatic enough, to justify a reaction.
The key filter: would most players, in that moment, plausibly be thinking or feeling roughly what the line says? If the answer is yes, you’re mind-reading. If the trigger is common enough that the line starts feeling repetitive, you’ve drifted back into generic callout territory — and the fatigue problem follows.
3. Establishing Banter — Bookend Moments
These are fixed, expected slots — match starts, session ends, loading beats — where players already anticipate something will play. There’s no gameplay happening, no agency at stake, and no risk of the line interfering with anything. The player has implicitly opted in.
This is the safest place in any system to spend a large volume of pure character variety. A big line pool here does double duty: it keeps the character present in the player’s mind (a consistent, low-level reminder that the character exists) and gives the writing room to be purely fun, without needing to earn it with a dramatic trigger.
One guardrail: keep these lines self-contained. The moment they start referencing specifics of what just happened mid-match, they’ve graduated to mind-reading territory — and the stricter rules around frequency and reliability apply.
The TL/DR Version
Say less. Make what you do say count more.
Utility lines earn their keep through reliability — always or never, never randomized, never obvious. Personality lines earn their keep through scarcity and timing — protected, targeted, and given room to breathe.
The characters players remember aren’t the ones who talked the most. They’re the ones who said the right thing at the right moment, and then knew when to shut up.