Fiction Workflow

GraphPaper 0.5.0 · Documentation source · Suggest an edit

Fiction workflow

Fiction has its own scene-oriented pipeline. It uses your premise and canon to develop character choices and consequences, rather than treating a story as an essay with invented names.

Establish the brief

Create a Fiction project and set genre, point of view, tense, desired voice and target length. In Canon, write the rules the draft must not casually change: what characters know, what a technology cannot do, where people are, or which event has already happened.

For example:

Memories can be transferred but never copied. Restoring one memory permanently consumes another. The archivist knows the cost; her client does not.

This is creative sample material, not a system requirement. Its value is a constraint that forces a choice. A premise and canon can seed a graph even without a document library.

Bring the right kinds of sources

Use Canon for a story bible, approved chapters and established world facts. Use Inspiration for possibilities that are not yet true in the story. Use Voice for your writing samples. Keep these roles separate so an inspiring idea does not silently become a rule.

An imported graph remains a map of possible relationships until you inspect it. Fictional truth is defined by the author’s approved material, not by a graph label.

Explore the dramatic structure

The graph can connect a character’s desire, an obstacle, a rule and a consequence. Pin the central conflict and exclude distracting threads. Angle discovery proposes story premises with stakes and counterforces. You can rewrite or replace every proposal.

Before drafting, edit the outline as a sequence of scenes. Each scene needs a purpose: a choice, discovery, reversal, loss or shift in a relationship. Word allocations are targets, not a substitute for pacing.

Understand the continuity ledger

As scenes are generated, the model updates a compact ledger of character locations and knowledge, chronology, objects, unresolved threads and possible canon conflicts. Later scenes receive that ledger along with the brief and preceding prose.

The ledger is an AI summary of the drafting pass. It is not an infallible database, and it does not supersede your canon. Whole-draft revisions and manual edits can make it stale. Read it through Story ledger, compare it with the current manuscript and correct the relevant brief or source material before another long pass.

Review for craft, not academic citations

Fiction review focuses on motivation, agency, causality, scene progression, emotional specificity, repetition, prose and the ending. It does not force invented narrative events to carry scholarly citations.

Author-voice learning and Humanize can help with cadence and diction. Deslop can flag formulaic contrasts or padding, but intentional repetition, ambiguity and unusual punctuation may be part of the story. Read proposed changes in context rather than accepting them solely because a cue count went down.

Keep the author’s choices

Save strong drafts before changing the ending or narrative perspective. Use Versions to recover checkpoints. A restored draft should also prompt a check of any continuity summary that described a later version.

Export Word, Markdown or HTML from the writing desk. The project backup carries the current project state; a full data-directory backup preserves revision history as well.