a continuity editor for novelists
Upload the manuscript exactly as you wrote it — Word, markdown, PDF. Plottolo splits it into chapters, builds the story bible (characters, places, objects, relationships, dated facts, narrative threads) and from then on watches the book's internal consistency as the draft grows. When chapter 20 contradicts chapter 12, you hear about it — with the quotation attached.
This is the job a continuity editor does on well-funded productions. Here it sits on your desk, at the twentieth rewrite as much as at the first.
protagonist's presence per chapter — draft 3
the product
Plottolo generates no prose. It produces analysis, critical resistance and instruments: the page you are writing is yours, and stays yours. That is a stated position, not a technical limitation.
the list of noes — written in stone
A product is defined as much by what it refuses to become. These are not «not yet»: they are out of scope, and any exception would be a declared decision, never a drift.
how it works
There is no tool to abandon. You keep writing in the program you already use, and whenever you want a snapshot of the work you upload the whole file. Plottolo works out for itself which chapters changed and re-reads only those.
drafts are never overwritten
Every upload creates a new, immutable draft: an ordered list of references to chapters, not a copy of the book. Identical chapters are the same shared block across every draft that contains them. «Go back to draft 2» rebuilds the list; it deletes nothing.
what one more draft costs you
what you see
The bible is not a summary: it is a searchable archive, built chapter by chapter, in which every attribute knows where it came from. Ask «where is the locket mentioned?» and you get the passages, not a list of word matches.
character record
| name | Elena Vaccaricharacter · appears in 41 of 110 chapters · POV in 12 |
|---|---|
| stated age | 34 in ch. 1«thirty-four, and only just» — ch. 1, draft 3 |
| family | One brother, Marco«her brother Marco walked in without knocking» — ch. 19 · conflicts with ch. 4 |
| distinguishing mark | Scar on the left forearm«the white line running along her forearm» — ch. 6, ch. 33, ch. 71 |
| knows that… | The lighthouse was put out deliberately — from ch. 28before that chapter she cannot refer to it |
| relationships | marco → brother · agata → mentor · the keeper → antagonist |
tension curve — chapters 1–24
density of dialogue, description and action
| ele | mar | aga | kee | nin | tom | |
|---|---|---|---|---|---|---|
| ele | ||||||
| mar | ||||||
| aga | ||||||
| kee | ||||||
| nin | ||||||
| tom |
narrative threads — where they open, where they close
interview a character
Question a character and they answer from their own record: only what the book says about them, with the knowledge they have at the chapter you pick. It does two jobs — it uncovers their history where the text stops short, and it uncovers where the characterisation has a hole. If they can't answer you, the hole is in the record, and they say so.
under the hood
Handing your novel to a single model means taking its word for everything. Plottolo uses more than one: each task goes to the engine suited to it, and on claims that matter a second engine checks the first. When they agree, the flag arrives with high confidence; when they don't, it arrives hedged — or doesn't arrive at all.
The reasoning is plain: one false alarm on something serious costs more than ten missed true ones. If the product wastes your time on an error that isn't there, you stop trusting it — and from then on you ignore the correct flags too.
the right engine for the job
No task pays for the most expensive model unless it needs it, and no critical task is left to the cheapest one. The link between task and concrete model is configuration: a price list changes, a line changes, the product doesn't.
| task | class | why |
|---|---|---|
| Extracting the bible, chapter by chapter | flagship | Everything rests on it: if the bible is wrong, everything downstream is wrong. |
| Consistency checking | standard | It works on the structured bible, not the prose: small input, bounded judgement. |
| Interviewing a character | flagship | Quality here is felt directly by whoever reads the answer. The context is one record, so it costs little. |
| Attaching a note to a character | economy | It is classification over a closed set: the bible itself. |
| Chapters with ambiguous boundaries | economy | Automatic segmentation does the bulk; the model is called only on the doubtful cases. |
| Transcribing a voice note | economy | High volume, high tolerance for error: you re-read it anyway. |
how they measure each other
// one reading per chapter, one schema only
{
"chapter": 28,
"pov": "Elena Vaccari",
"tension": 8,
"split": {
"dialogue": 41,
"description": 35,
"action": 24
},
"entities": [
{
"name": "Elena Vaccari",
"type": "character",
"role_in_scene": "protagonist",
"attributes": [
{
"key": "knows_that",
"value": "the lighthouse was put out
deliberately",
"quote": "she understood someone had
put it out, and meant to",
"chapter": 28
}
]
}
],
"events": [
{ "summary": "Elena discovers the sabotage",
"internal_when": "evening of the third day" }
],
"relationships": [
{ "from": "Elena", "to": "the keeper",
"type": "suspicion" }
],
"threads": [
{ "code": "the-darkened-lighthouse",
"state": "advanced" }
]
}
the three promises everything else rests on
in one line
Upload the manuscript. Get the bible, the contradictions, the threads left open and the pacing of the book. Rewrite, upload again, and pay only for the chapters you touched.