Writing interactive stories

A story you wrote once but a hundred players will live differently. That's the design problem. This page is the author's guide.

Start from the wrong place

The most common mistake is to plot the story first. Don't. Plot is what the AI generates; structure is your job. Decide:

1. Where does this story start? (the inciting situation)
2. What's the central tension? (the thing the story is about)
3. What are the possible roles the player could take?

If you can answer those three, you have enough to start authoring the structure.

Chapters — tones, not plot events

Chapters in Reverie aren't "Chapter 1: Meet the witness." They're tonal acts. Three to seven of them.

Examples for a noir detective story:

• Chapter 1 — The unease. Routine ends. Something is wrong.
• Chapter 2 — The descent. The detective gets in too deep.
• Chapter 3 — The trust break. Someone close betrays.
• Chapter 4 — The reckoning. The truth surfaces; choices have weight.

Each chapter sets a mood for the AI. It doesn't dictate scenes; it colours them.

Milestones — moments that matter

Milestones are the bright moments the AI knows to drive towards. Four to seven per story, written as one-liners:

• "Detective learns the senator's daughter was the one who tipped off the killer"
• "First moment of physical tenderness between detective and witness"
• "Detective makes the choice to bury the report"

Milestones don't all have to happen. The AI tries to reach them when the player's actions head that way, lets them slide when not. Replay value comes from missing some.

Order matters. Author them roughly in the order you'd expect them to happen. The AI uses this as a soft sequence.

Data panels — state the AI reads

Data panels are small dashboards the story tracks. The story editor has a Data Panels tab where you author them — name, fields, types, starting values. Common examples:

• Affection with each character (0-100)
• Clues collected (growing list)
• Suspicion level (low/medium/high)
• Inventory (item list)
• Trust per character

The AI reads the current state before each generation and updates it after. That means a scene where the player ignores the witness's plea will see suspicion go up — and the AI's next generation reflects it.

The editor includes an AI Generate option to draft data panels from a prompt if you don't want to start from scratch.

Keep panels small. A few well-scoped values across all panels is usually enough.

Roles — same story, different player

Roles let one story support multiple protagonists. Each role has its own opening hook, its own pool of available milestones, its own data-panel views.

For our detective story:

• Detective — sees suspect interrogations, has badge authority
• Suspect — running from the detective, sees flashbacks
• Witness — knows the truth, decides who to tell

Three roles = effectively three stories with shared world and shared cast. Massive replay value for one author-week of work.

Character roster

Pick from existing characters or author roster-only characters that exist within this story. Each character keeps their voice and personality but gets the story's context.

The blurb and cover

Treat the blurb like the back of a paperback. Tone, premise, central question — three sentences. The cover image is generated or uploaded; it's what players see when browsing.

Testing your story

The Story editor has a Playtest button that opens a fresh play session in a side pane. Play through twice as different roles; watch where the AI succeeds, where milestones miss, where data panels feel forced.

Common iteration after first playtest:

• Some milestones never fire → shrink the list, or move them earlier
• One chapter feels flat → revise the tone description
• Data panel never moves → cut it

Publishing

When ready: Story editor → Publish. Stories go through the same image/quality review as characters. Once live, they appear on the Stories tab and can be featured in the discover feed.

Earning from stories

Story plays earn the author character-creator revenue the same way chats do — every credit consumed in the play sends a share back. Featured stories earn more by virtue of higher traffic.

Revenue streams →