Novel Character Graph

Problem It Solves

Long novels can reach millions of characters, and mixed encodings across txt, epub, pdf, docx, and mobi files can cause truncation or unreadable output during single-pass analysis. Characters, aliases, mentorship, hostility, factions, cultivation systems, realms, and foreshadowing are spread across chapters, making manual review prone to missing villains, heroines, mentors, or recurring NPCs.

How It Works

• Preprocessing: detect file encoding, convert common novel formats into parseable text, then split by line count or chapter to reduce context pressure.
• Incremental extraction: collect newly introduced characters, aliases, nicknames, relationship changes, and key plot nodes while updating techniques/artifacts/pills, factions, sects, nations, and worldbuilding.
• Standard modeling: classify characters into S/A/B/C tiers and build character cards, using consistent markers for ❤️ romance, 👨‍🏫 mentorship, 🤝 comradeship, 👑 subordination, and ⚔️ hostility.
• Graph delivery: output two Markdown files—one for the full character lineage and one for core relationships—and generate an index.html page with relationship graph, character profiles, faction map, and comic design tabs.
• Optional theme song: when only an opening track, OST, or score is needed, select a style based on genre and mood, generate Suno Style and Lyrics tags, then combine audio and video with ffmpeg.

Boundaries

This skill is better suited to structured character analysis of long novels than to simple plot summaries. If the source file has severe encoding errors, unclear chapter boundaries, or many aliases, repair the text and confirm character merging rules first to avoid duplicate entities or ambiguous relationships.