A character who only chats runs dry in about three days. A story line is what gives them somewhere to go — chapters that unlock as your relationship grows, secrets that surface in a deliberate order, and scenes that actually happen instead of being described. This guide walks through designing one end to end, using a character we built ourselves as the worked example.
If you haven't set up a character yet, start with the character setup guide — personality, speaking style, reply mode. This article picks up where that one ends.
The part most people skip
Most people finish the persona screen and stop. The character is witty, has a good voice, remembers your name. And by day three the conversation is a loop: how was your day, what are you doing, tell me about yourself.
Nothing is broken. There is just nowhere to go.
A persona answers who he is. A story line answers what you two are going to go through. Those are different jobs, and the second one is where the feeling of a relationship actually comes from.
Three pieces, and they do different things
Before designing anything, it helps to separate three ideas that are easy to blur together.
| What it is | How the reader gets it | |
|---|---|---|
| Chapter | One scene of the story | Read — unlocked, then opened |
| Gate | The condition that opens the next chapter | Earned through the relationship |
| Beat | A point the scene has to reach | Talked into existence, in chat |
The one-line version: a chapter is what you read, a beat is what you talk your way to, and a gate decides whether the next chapter opens at all.
Most authoring mistakes come from confusing beats with chapters. A chapter is prose you write once. A beat is a goal you hand to the engine, and the character improvises his way there.
Step 1 — Design the person before the plot
Our worked example is 吴邪 (Wu Xie), an antique-shop owner in Hangzhou who has spent ten years waiting for someone to come back. Before writing a single chapter, we wrote a character bible. Six questions, in this order:
First impression. What does someone notice in the first minute? Warm, easy to talk to, a little bookish. Doesn't look like a man who has been anywhere dangerous.
Visible habits. Small, concrete, repeatable. Drinks tea, never coffee. Talks in circles — two lines of small talk before the point. Frighteningly good memory for detail.
The contradiction. This is the one that makes a character feel real. He looks like the most agreeable person in the room, and he is the one who listens to no one. "Sure." "Fine." "I hear you." — from him, those usually mean he has already decided to ignore you.
What he's actually missing. Not what he wants — what he lacks and can't ask for. Someone who doesn't need rescuing. Every relationship he has is built on "something went wrong, so he showed up." He has never had one that didn't start with danger.
The core secret. One thing, and it should be heavier than it first sounds. Ten years ago someone told him to wait. Everyone knows he's waiting — that isn't the secret. The secret is that he is no longer sure whether he's waiting for that person or for the waiting itself. If they came back, the ten years would be over, and he doesn't know what's left of him underneath.
What actually moves him. Not flattery, not need. For 吴邪 it's permission to stop: "You don't have to look today." That sentence is more dangerous to him than any confession, because it threatens the foundation he's stood on for a decade.
Write these six before you write chapter one. The plot gets much easier afterwards — you're no longer inventing events, you're finding the events that would press on exactly these places.
Step 2 — Turn the secrets into a ladder
A character is interesting because you don't get all of him at once. So take the core secret and break it into five layers, cheapest first:
| Layer | 吴邪 | How it surfaces |
|---|---|---|
| 1 | Drinks tea, talks in circles | Just by chatting |
| 2 | An old injury that aches when it rains — he never mentions it | After a while together |
| 3 | A few days every year he closes the shop and doesn't answer the phone | Once you're a few chapters in |
| 4 | A page torn out of his uncle's ledger — he tore it himself | Deep in the story |
| 5 | He's afraid the ten years were something he made up alone | Only after a scene you lived through together |
Each layer gets attached to a different kind of trigger, so depth costs more as you go down: time for the shallow ones, relationship level in the middle, story progress deeper, and for the last one, a story fact that can only exist if a particular scene actually happened.
That last rule is what keeps the bottom of the ladder feeling earned. Layer 5 cannot be reached by chatting a lot. It can only be reached by having been there.
Step 3 — Lay out the chapters
Now the plot. For 吴邪 we wrote five chapters:
- A rainy night at the shop — you come in out of the rain carrying an old wooden box. He sees the mark on the corner. Free, no gate.
- The torn page — his uncle's ledger, one page missing. Acquaintance.
- The letter on the roof — an envelope with no sender, containing only a date. Friend.
- Above the snowline — a storm traps you in a tent. Flirty.
- Taking it back (side story) — he hands the box back and lets you decide. Flirty + gift points.
Three rules we'd give anyone laying out chapters:
Never put a paid or gift-point gate in the middle of the chain. We learned this the hard way on an earlier character. A mid-chain gate doesn't just lock its own chapter — every chapter after it is locked too, because chapters unlock in order. Four different kinds of test players all died at the same wall. Put resource gates at the end, as side stories.
Stage gates are your real pacing dial. The chapters themselves are quick; the waiting between them is what sets the rhythm. For 吴邪, chapter 4 requires the flirty stage, and that turns a 10-turn story into a 63-turn one. That's deliberate — a few days of talking rather than one long evening — but it's a decision you should make on purpose rather than discover later.
Keep a free path all the way through. At least one route to the end that costs nothing. Emotional connection behind a paywall is the fastest way to make a character feel like a vending machine.
Step 4 — Write the beats
A beat is not dialogue. You are not writing what he says — the model does that, differently every time. You're writing what has to happen.
For the first chapter of 吴邪:
- Goal: he gets a clear look at the mark on the box and asks where you got it.
- Done when: he has looked and asked, and you've answered.
- Gentle steering: if the conversation drifts, his eyes keep going back to the box on the counter.
- If that's not working: lightning outside lights up the corner of the box, and a second layer of the carving shows in the flash.
- Last resort: he turns the box around himself and says what the mark means.
Those last three are a ladder, and the engine climbs it for you: gentle nudging first, an environmental event after about eight turns, and after fifteen the character simply makes it happen.
That last rung matters more than it looks. Getting stuck is not an acceptable outcome. A reader who never happens to ask the right question should still get the story — so the fallback always fires, and it still grants everything the beat was going to grant. The scene really did happen; it just happened because he brought it up instead of you.
Write the fallback line as carefully as the others. For a lot of readers, it's the version they'll see.
Step 5 — Check it before anyone plays it
Two checks, both instant, both free, neither one calls a model. You can run them after every edit.
Check script reads your story as a graph and answers questions proofreading can't: is every gate openable by somebody, and is every chapter reachable from a fresh save? It catches gates that need a story fact nothing ever produces, chapters whose prerequisite comes after them, and chapters only reachable by paying.
Playtest plays your story to the end four times — a cooperative reader, an adversarial one who tries to talk their way up, a silent one, and a rambler who chats constantly but never on topic. It reports how many turns each took, which beats had to be force-advanced, and which ones nobody could finish.
Here's the real output for 吴邪:
Check script 0 findings · 5 chapters reachable
Playtest all four readers: 7/8 beats · 4/5 chapters
↳ "Taking it back" is locked: needs 30 gift points
Seven of eight beats for everyone, including the silent reader — that's the anti-stall ladder doing its job. The one unfinished beat is the paid side story, which is the intended design.
When a beat can't be finished, the report traces it to the chapter and gate actually responsible — and if the blockage is upstream, it says so, because the chapter you need to fix usually isn't the one that looks broken.
Questions
Do I have to write all five chapters before publishing?
No. One chapter with two good beats is a working story line, and it's a much better use of your first hour than five thin ones. Add chapters as you see where readers actually go.
Do beats cost credits?
No. Beats, gates and the two checks are all part of authoring — no model calls, no cost. You can run Check script and Playtest as often as you like.
What if a reader never says the thing my beat needs?
They still get the scene. After a set number of turns the character brings it up himself, and the beat's effects are granted exactly as if the reader had gotten there. Being stuck isn't a state we let readers reach.
Can my character just tell someone the deepest secret if they ask directly?
Not unless the story got there. Layer-5 secrets are tied to facts that only exist after a specific scene has happened, so asking early gets you the character deflecting — which is usually more interesting than the answer anyway.
How long should the whole thing take to play?
For a five-chapter line, somewhere between a few evenings and a couple of weeks is a good target. The dial is the stage gate on your late chapters, and Playtest will tell you the turn count before a single reader sees it.
Can I write a story for a character based on someone else's work?
For your own private use, that's your call. If you plan to publish a character publicly, write an original one — fan characters carry real copyright exposure, and original characters are also where the interesting design space is.
Start with one chapter
Pick a character you already like talking to. Write the six questions about him — first impression, habits, contradiction, what he's missing, the secret, what moves him. Then write one chapter and two beats, and run Playtest.
You'll know within a minute whether the thing you imagined actually holds together.