Edit a Book with AI — Paste, Edit, Review

Most AI writing tools force you to edit one paragraph at a time or copy-paste into a chat window and hope for the best. Entangled Text's book editor handles the entire manuscript in a structured pipeline — from input to diff review — so you stay in control without doing the busywork.

That's the full cycle. Paste, instruct, review. You can run as many editing passes as you want — each one builds on the last accepted version.

The edit instruction is freeform — anything you'd tell a human editor, you can tell the AI. Here are real examples that work well:

• Fix pacing — "The first three chapters are too slow. Cut unnecessary exposition and get to the inciting incident faster."
• Rewrite dialogue — "Make every conversation between Marcus and Liora more tense. They're supposed to distrust each other — the current dialogue is too polite."
• Change tone — "Shift the narration from detached third person to a warmer, more intimate close-third. Let the reader inside Elena's head."
• Add sensory details — "Every scene in the market district should smell, sound, and feel alive. Add sensory texture without slowing the pacing."
• Shorten paragraphs — "Break up any paragraph longer than five sentences. This is a thriller — the prose should feel fast."
• Strengthen a character's voice — "Make the antagonist's dialogue more menacing. He should sound calm and reasonable on the surface, with threat underneath every sentence."
• Remove filter words — "Cut every instance of 'she felt,' 'he noticed,' 'she realized,' and 'he seemed.' Replace with direct sensory description."
• Consistency pass — "The character's name changes from Katharine to Katherine in chapter 7. Standardize to Katharine everywhere."

Feeding a 90,000-word novel into an AI in one shot would be expensive, slow, and unreliable. Feeding it one paragraph at a time would lose all context. The editor takes a middle path designed around how language models actually work.

Chapter-aware splitting

The chunking engine detects chapter headings, scene breaks, and section dividers in your manuscript. It splits along those natural boundaries so edits never cut a scene in half. If your text doesn't have explicit chapter markers, it falls back to intelligent size-based splitting with overlap so nothing falls through the cracks.

Two-pass token optimization

Before editing begins, a fast scan pass reads each chunk to determine whether your instruction is even relevant to that section. If you asked the AI to "rewrite all dialogue to sound less formal" and a chunk contains no dialogue, it gets skipped entirely. This saves tokens — and money — without sacrificing quality. Only chunks that need editing go through the full edit pass.

Context windows that overlap

Each chunk is processed with surrounding context from adjacent sections. The AI sees what happened before and what comes after, so edits stay consistent with the narrative around them. A dialogue edit in chapter 4 won't accidentally contradict a character detail established in chapter 3 if both are within the context window.

AI editing tools that silently replace your text and hand back a "finished" version are a trust problem. You don't know what changed, why it changed, or whether the AI introduced new issues while fixing old ones.

Entangled Text shows you every change. After the AI processes a chunk, you see a word-level diff — deletions highlighted in red, additions highlighted in green — exactly like a code review or a tracked-changes document. Nothing is applied until you approve it.

Per-chunk control

Each chunk gets its own accept or reject decision. If the AI nailed chapters 1 through 5 but mangled the dialogue in chapter 6, reject that one chunk and re-run it with a refined instruction. The rest of your accepted edits stay in place.

Accept all

When you're running a mechanical pass — fixing typos, standardizing names, removing filter words — you may not need to review every chunk individually. The accept-all option applies every pending diff at once. Use it when the instruction is narrow enough that you trust the output without line-by-line review.

Editing a full-length novel isn't a fifteen-minute task. You might review ten chapters, close your laptop, and come back to it two days later. The editor preserves your session state — pending diffs, accepted chunks, the current instruction — so you pick up exactly where you left off.

Your editing sessions appear in the sidebar history alongside your other projects. Open any past session to continue reviewing, run a new editing pass on the already-edited text, or copy the final version into a new book project.

Once your edited manuscript is in good shape, save it as a Story Studio project or export directly.

Save as a project

Convert your edited text into a full Entangled Text book project. This gives you chapter organization, listing metadata, and access to the Ship tab for exports and launch prep.

Export for publishing elsewhere

Download your manuscript in the format you need:

• EPUB — Ready for Amazon KDP, Apple Books, Kobo, and other ebook distributors. Metadata and chapter structure included.
• DOCX — Open in Word, Google Docs, Scrivener, or formatting tools like Vellum and Atticus for print layout.
• PDF — Share with beta readers or use for print-on-demand services.

Your text is yours. Edit it here, publish it anywhere.

The editing workspace is free. Start with built-in Kestrel — Free on eligible features — no API key required. Pro adds Kestrel — Pro (much more powerful AI with better prose + a monthly pool instead of Free’s chapter/day gate). Ultra is the same powerful model family with a larger included pool. Both paid plans unlock included AI — no setup if you want other providers (you pay those providers directly). Compare Free / Pro / Ultra.

Supported providers:

• OpenAI — GPT-5.6 Terra (default), Sol, Luna, and other models
• Anthropic — Claude Sonnet, Claude Opus
• Google — Gemini Pro, Gemini Flash
• Mistral — Mistral Large, Mistral Medium
• DeepSeek — DeepSeek V4-Flash, DeepSeek V4-Pro
• xAI — Grok models
• OpenRouter — 100+ models through one key
• Groq, Together AI, Fireworks AI — fast open-model hosts
• Perplexity — Sonar models
• Cohere — Command R+ and related models
• Azure OpenAI — your Azure deployments
• Amazon Bedrock — Claude, Nova, Llama, and more in AWS

Pro and Ultra use included Kestrel — Pro / Ultra for editing passes. Included AI — no setup.

Editing is one stage. When you want a store-ready file, chain the free tools in order:

1. Draft in Story Studio (or paste here)
2. Manuscript Cleanup — cut AI sameness
3. Consistency checks — names, POV, bible drift
4. Blurb + cover
5. Kindle & KDP formatter → Ship / EPUB