Table of Contents
Your 180,000-word epic fantasy lives in Scrivener. Three years of work. Character sheets, worldbuilding bible, scene index cards, the works. Now you want to try Sudowrite, but the thought of breaking that structure feels like setting your manuscript on fire. Good news: you don't have to. The migration is safer than you think, and your Story Bible ends up cleaner on the other side.
This guide walks through the full import process. Export format choices, what survives the move, what needs rebuilding, and how to map your Scrivener structure onto Sudowrite's Story Bible without losing the work that took you years to build.
Why writers move from Scrivener to Sudowrite (and why they hesitate)
Scrivener is an organizational beast. It excels at compiling, restructuring, and managing massive projects. What it doesn't do is help you write the next sentence when you're stuck at 67,000 words with a romance subplot that refuses to spark. That's where Sudowrite earns its keep. Features like Write, Rewrite, Describe, and Brainstorm work on the prose itself, not just the folder structure around it.
The hesitation is always the same. Years of binder organization. Custom metadata fields. Color-coded labels for POV characters. Folder structures with research subfolders three levels deep. Moving feels like emigrating with one suitcase.
Here's the reframe: you're not abandoning Scrivener. Many writers keep Scrivener for long-term archival and compile-to-ebook workflows while drafting and revising in Sudowrite. The import is about getting your manuscript and the structure that matters into a place where the AI can actually help with the prose.
What survives the migration (and what doesn't)
Before you export anything, set expectations. Sudowrite's import handles the manuscript itself extremely well. Chapter divisions, scene breaks, and the actual prose come over cleanly. What gets restructured is the supporting material.
Here's the honest breakdown:
- Comes over directly: chapter text, scene breaks, basic formatting (italics, bold), paragraph structure.
- Needs mapping: character sheets, location notes, worldbuilding documents. These move into Story Bible cards.
- Needs rebuilding: custom metadata, label colors, keyword tags, snapshots, and Scrivener-specific compile presets.
- Doesn't transfer: index card corkboard view (Sudowrite has Canvas instead), Scrivener's split-screen editor.
If you've spent five years building a character bible with 40 fields per character, you're not losing it. You're just moving it into a format where Sudowrite's models can actually read it and use it while writing. That's the trade.
Step 1: Export your Scrivener project the right way
Scrivener gives you several export formats. The wrong one will mangle your chapter breaks or strip your italics. The right one keeps everything intact.
Open your project in Scrivener and go to File > Compile. You'll see a dialog with format options on the left and a structure preview on the right. Here's what to choose:
- Format: Choose Manuscript (Times) or Default. Skip the ebook formats for now. You want clean text without compile-side flourishes.
- Compile for: Select Microsoft Word (.docx). DOCX preserves italics, bold, and chapter structure. Plain text loses formatting. RTF works but is less reliable across systems.
- Section types: Make sure your chapters are mapped to chapter section types and your scenes to scene section types. If you've been working with the default Novel template, this is already correct.
- Separators: Use page breaks between chapters and a single blank line or three asterisks between scenes within a chapter. This makes the import recognize chapter breaks automatically.
- Front matter: Uncheck the title page, copyright, and dedication. You'll add those back at the end. They confuse the chapter parser on import.
Click Compile and save the .docx somewhere you can find it. Open the file in Word or Google Docs first to verify the chapters look right and your italics survived. If a passage of telepathic dialogue you'd italicized comes out plain, fix it now. Sudowrite imports what's in the file, not what was in Scrivener.
Exporting your supporting documents
Your character sheets, worldbuilding notes, and outline live in Scrivener's binder, separate from the manuscript. Export those too, but as a separate batch.
Select your Characters folder in the binder. Right-click and choose Export > Files. Export as RTF or plain text, one file per document. Do the same for Places, Research, and your outline folder. You'll paste these into Story Bible cards in the next steps.
Step 2: Import the manuscript into Sudowrite
Open Sudowrite and create a new project. Name it whatever your book is called. From the project dashboard, click Import. You'll see options for DOCX, TXT, and pasting raw text.
Select your compiled .docx and upload it. Sudowrite parses the file and asks how to split it. You'll get three options:
- Auto-detect chapters: Works if your Scrivener compile used clear "Chapter 1," "Chapter 2" headings or numbered breaks. Sudowrite scans for these patterns.
- Split by heading: Tell Sudowrite which heading level marks a chapter (usually H1 or H2 depending on your compile settings).
- Manual split: Import the whole thing as one document and split chapters yourself. Tedious for a long novel.
Auto-detect usually works for standard novels. If you have unconventional chapter headers (Roman numerals, named chapters, no numbers at all), use Split by Heading. Preview the result before confirming. If chapter 7 swallowed chapter 8, your separator wasn't consistent in Scrivener. Fix the export and try again. Better to spend ten minutes here than discover the problem 40,000 words later.
Once imported, you'll see your manuscript in the chapter sidebar on the left. Click any chapter to open it. The prose should be intact, formatting preserved. Scene breaks within chapters survive as three asterisks or whatever separator you used.
Step 3: Map your Scrivener notes into the Story Bible
This is where the real migration happens. Sudowrite's Story Bible is the persistent memory the AI reads while helping you write. It has six sections: Characters, Worldbuilding, Style, Outline, Synopsis, and Braindump. Your Scrivener notes need to land in the right buckets.
Characters
Click the Story Bible icon and go to Characters. Add a new Character card for each of your major players. The card has fields for description, personality, voice, and notes. Paste your Scrivener character sheet content here, but reformat for what the AI actually uses.
Here's the difference. Your Scrivener sheet might have looked like this:
NAME: Kestra Vell. AGE: 27. EYES: Gray-green. HAIR: Black, shoulder-length. OCCUPATION: Royal cartographer. BACKSTORY: Born to a fisherman in Veth, lost mother at 9, apprenticed to Mapmaker Doran at 14...
Cleaned for Sudowrite's Character card:
Kestra Vell. 27. Royal cartographer with gray-green eyes and shoulder-length black hair. Fisherman's daughter from Veth, lost her mother at nine, apprenticed to the cartographer Doran at fourteen. Wary of nobles. Speaks in clipped, observational sentences. Hides grief behind dry humor. When stressed, she sketches maps of places she's never been.
The second version reads like prose. The AI uses it as voice and personality reference when you call Write or Rewrite. Bullet points and key-value fields are harder for the models to parse into believable dialogue and behavior. Convert your stat blocks into character-as-person paragraphs.
Do this for your six to ten most important characters. Skip the spear-carrier with two lines on page 340. You can add minor characters later when they matter.
Worldbuilding
Worldbuilding cards in Sudowrite are organized by Rules, Lore, Factions, Settings, and Items. If you wrote a 12,000-word grimoire of magic system rules in Scrivener, paste it into a Rules card. Faction politics? Factions card. The capital city your story mostly takes place in? Settings card.
Don't dump all of it. Be ruthless about what's actually relevant to the book you're currently drafting. If a faction never appears on-page, the AI doesn't need to know about it. You can always add more cards as the story expands. Save Series Folder for shared lore across multiple books, which we'll cover in a moment.
Style
The Style card tells Sudowrite what your prose sounds like. This is critical and easy to skip. Write three to five sentences describing your tone, POV preferences, and rhythm. Examples:
Close third-person past tense. Sparse, observational prose with sudden bursts of lyrical description during emotional beats. Sentences vary aggressively in length. Dialogue does heavy lifting; interiority is brief. Influences: Abercrombie's cynicism, Le Guin's restraint.
If your Scrivener project had a style sheet doc, that goes here. If not, write one now. It takes ten minutes and dramatically changes how Sudowrite's Write feature continues your scenes.
Outline and Synopsis
Paste your outline (whether you used a beat sheet, three-act structure, or Save the Cat) into the Outline section. Sudowrite's Chat feature reads this and can answer questions like "what should happen next based on my outline?" without making things up.
The Synopsis is a one to two paragraph summary of the entire book. If you wrote a query letter pitch, that's your synopsis. If not, write one. The synopsis is what Sudowrite uses when it needs to understand the whole arc, not just the current chapter.
Braindump
Braindump is the catch-all. Random notes, half-formed ideas, that one piece of research you found fascinating but couldn't fit anywhere. Dump them here. The AI reads it, but with less priority than the structured sections.
Step 4: Choose your prose model based on genre
Sudowrite runs multiple models, and the right one depends on what you write. Sudowrite's prose-modes matrix maps subgenres to recommended models. Set yours after import:
- Muse for romance, erotica, horror, and thriller. Muse is Sudowrite's fiction-trained model. It writes like a novelist, won't refuse explicit or dark content, and handles emotional intensity better than general-purpose models. If you write smut or grimdark, Muse is the default.
- Claude 3.7 Sonnet for fantasy, mystery, and YA. Strong on plot coherence, dialogue, and the cleaner prose styles common to these subgenres.
- Claude 3 Opus for literary, historical, and sci-fi. The most thoughtful model. Best when your prose carries philosophical weight or your sentences need to do real work.
- Deepseek-R1 for adventure and crime. Punchy pacing, good with action and procedural beats.
- GPT-4o Mini for non-fiction. If you're drafting a memoir or a creative non-fiction piece alongside your novel, this is the lean choice.
Set the model in your project preferences. You can switch per scene if a chapter calls for something different. A romance subplot inside your epic fantasy might benefit from a one-chapter swap to Muse, then back to Sonnet for the war council scene.
Step 5: Verify and start writing
Before you hit Write on a continuation, do a quick verification pass. Open three random chapters from different parts of the book. Check that italics survived, chapter breaks are correct, and scene divisions match what was in Scrivener. Open your Character cards and confirm the descriptions match how those characters actually behave on the page.
Then run a small test. Pick a scene mid-book where you know the characters well. Highlight a paragraph and hit Rewrite with the Show Don't Tell mode. Does the rewrite sound like your book? If yes, your Style card and Character cards are pulling their weight. If the rewrite sounds generic, your Story Bible needs more specifics. Add sensory details, voice tics, and stylistic preferences to the Style card.
For new prose, try Write in Guided mode. Give it a direction ("Kestra confronts the council, suspects the cartographer guild is being framed"). Compare the output to how you'd write the scene yourself. If it's in the right zip code, you're set. If it drifts, refine the Story Bible.
Common import problems and how to fix them
Most migrations go smoothly. When they don't, the issue is almost always one of these:
- Chapter breaks merged: Your Scrivener compile didn't use consistent separators. Re-export with explicit page breaks between chapters.
- Italics lost: You exported as plain text instead of DOCX. Re-export.
- Scene breaks missing: Your scenes weren't separated in Scrivener. Add three asterisks or a horizontal rule between scenes before re-exporting.
- Special characters mangled: Smart quotes, em-dashes, or em-spaces converted to garbage. Open the .docx in Word, use Find and Replace to clean up, re-import.
- Characters don't sound right when AI writes them: Your Character cards are too statistical. Rewrite them as prose paragraphs with voice samples and behavioral tics.
If you're writing a series
Sudowrite has Series Folder for shared Story Bible across multiple books. If you imported book three of a trilogy, set up a Series Folder, add the shared characters and worldbuilding once at the series level, and let each book inherit them. You can override at the book level when a character evolves between books.
Chapter Continuity also kicks in here. It catches contradictions across chapters and books: eye colors that changed, a character who died but reappears, a city that was destroyed in book one but somehow has a tavern scene in book two. Worth running once your full manuscript is imported.
What to do with your old Scrivener project
Keep it. Don't delete anything. Scrivener remains a strong tool for compile-to-ebook workflows, archival, and any future restructuring you might want at the binder level. Many Sudowrite users draft and revise in Sudowrite, then export the finished manuscript and run a final compile in Scrivener for ebook formatting.
The migration isn't a divorce. It's adding a writing partner to a workspace you've spent years building. Your structure stays in Scrivener if you want. Your prose moves where the AI can help you finish the book.
Sudowrite offers a free trial. Import a single chapter and a few character cards, run a few Write and Rewrite passes, and see if the prose comes back in your voice. If it does, bring the rest of the manuscript over. If not, that's a Story Bible problem, and it's fixable in an afternoon.