From a written setup to your own playthrough

Start in Discover and open a story that interests you. Its detail view shows the premise, opening choices, and available roles. Beginning a new tale creates a separate save, so returning to the same story later can produce a completely different playthrough.

The opening scene is part of the story setup and costs nothing. It gives you a place, an immediate problem, and enough information to make the first move without reading a manual.

A playable role is more than a portrait and a name. It carries a position in the world: what the character knows, who recognizes them, and what they stand to lose. You can use a story's prepared role, bring a saved character, enter only a name, or play without a profile.

You can switch characters from the tale header later. Earlier scenes remain unchanged; the new viewpoint applies to the next message onward.

Type what your character does or says. A useful turn can be one sentence: “I put the ledger back and ask who else has a key.” You can also include intent when it matters: “I agree, but I want her to think I am frightened.”

You do not need to write polished prose. Clear actions give the Storyteller something concrete to resolve. If you try several unrelated things at once, expect the scene to focus on the parts that can happen together.

• Speak in character, summarize a conversation, or combine speech with action.
• Ask to examine a person, object, room, record, or inconsistency.
• Attempt risky plans. Failure can move the story instead of ending it.
• Use ordinary language. Special commands are not required.

Suppose the opening places you in a locked records room. Footsteps are approaching, and the shipping ledger you came to find has one page cut out. You write: “I leave the ledger where it was, pocket the loose thread caught in the binding, and climb out through the courtyard window.”

The Storyteller now has several things to resolve. Can the window be opened in time? Does anybody see you cross the courtyard? What does the thread suggest, and did leaving the ledger prevent the search from being noticed? The reply should answer the immediate attempt and carry useful consequences into the next turn.

A menu might have offered “hide,” “run,” or “inspect the ledger.” Free-form play lets one action combine caution, evidence, and escape. It can also fail in a more specific way than a generic wrong choice.

The Storyteller receives the story setup, relevant notes, a compact account of older events, recent turns, and your latest action. It narrates the consequence, updates the situation, and gives control back. The response streams as it is written.

It is still AI-generated fiction. Details can be surprising and, occasionally, wrong. Story Notes give you a direct way to correct important canon instead of hoping the model guesses what should be remembered.

Each save has controls for narration length, stakes, and pacing. Narration can stay brisk or expand into fuller scenes. Stakes can favor a cozier experience or allow deadlier consequences. Pacing can linger or move quickly. These controls shape the Storyteller without rewriting the premise.

You can also choose among the Storytellers available on your plan. The selected model changes how the narration is produced, while the same safety and account rules continue to apply. Hands-free voice can read the tale and listen for the next action when you prefer speaking to typing.

A save keeps its own conversation and Story Notes. Close the app, come back later, and continue that version of the tale. Starting again from the catalog makes another save rather than erasing the first one.

Long conversations are compacted so the entire raw transcript does not have to fit into every request. Recent exchanges remain verbatim, while older events are summarized and durable facts can remain in Story Notes.