NovelGen AI is a personalized AI story generator for readers who have a story they want to experience but lack the time, skill, or confidence to write it themselves.

Instead of helping professional writers manage a manuscript, NovelGen AI focuses on turning your favorite settings, characters, genres, and story ideas into complete readable novels with minimal effort.

Enter a seed, generate or refine a plot, and let the app produce a full chapter-by-chapter story that matches the kind of novel you personally want to read.

• Desktop AI Story Generating Environment

  • Standalone Rust/Tauri app with no dependencies (except for Edge WebView2 runtime which is included in Windows).
  • Supports local generation through LM Studio and Ollama, and cloud generation through Google Gemini API, Ollama Cloud, OpenCode Go, and OpenCode Zen.
  • Generates stories in Korean, Japanese, or English.

• Hands-Off Novel Creation

  • Enter a Seed, choose the chapter count, generate a plot, and novel generation.
  • Use 🚀 Batch Start to automatically create multiple story variations from your idea.
    • Pick the version that best matches your taste.
  • Seamless Previewing: Review your work instantly with a feature-rich built-in viewer supporting Markdown, KaTeX, word wrap, per-pane font sizing, Comfort mode, and persistent Light / Dark themes for a truly pleasant experience.

• Customize the Story You Want to Read

  • Full automation is available, but you can still edit the plot, adjust the tone, refine chapters, or guide the story when you want more control.

• Precision Plot Planning & Refinement

  • Manually sculpt your story by creating and editing chapter-by-chapter plot to your exact specifications.
  • Add highly specific custom refine instructions to control pacing, tone, relationships, conflict, expansion targets, or details to preserve.
  • Automatically generate professional plot improvement ideas using the ✨ Auto Instructions feature when you need inspiration.
  • Target plot refinement by story part with optional Start Part and End Part fields, including multi-part Auto Instructions. (Start Part: 0 for settings)
  • Refine long plots in structured chunks, starting with setup sections and progressing through each story part in order.
  • For long outlines of 13 chapters or more, generate plots in stages: setup sections first, then each story part in sequence.
  • Monitor estimated plot token usage with a CJK-aware counter.

• Advanced Generation & Continuity Management

  • Maintain absolute consistency with layered context, chapter summaries, and sliding-window memory under the hood.
  • For novels of 13 chapters or more, generation uses a focused plot context: the current part plus adjacent parts are kept detailed, while distant chapters are reduced to title and content.
  • Resume interrupted generation from the exact last written chapter with continuity fully intact.
  • Save all generated novels, plot files, and metadata locally for manual editing and safe-keeping.

• Targeted Novel Revision Tools

  • Refine completed drafts chapter by chapter against your specific plot requirements.
  • Limit manuscript refinement to an exact chapter range using optional Start Chapter and End Chapter fields.
  • Provide specific manual instructions or use ✨ Auto Instructions on a single chapter to automatically review it against adjacent chapters and generate 10 targeted improvement points.
  • Precisely improve scene purpose, emotional progression, dialogue clarity, prose flow, and fix repetitive expression patterns.

• Batch & Productivity Tools

  • Queue multiple generation jobs and process them sequentially.
  • Optionally auto-refine plots before generation and novels after generation in batch mode.
  • Import .txt / .md files by drag and drop into prompt, seed, plot, and novel editors.

• Frontend: React 19 + Vite + TypeScript (Modern, Component-based Architecture)
• Backend: Rust 🦀 (Safety & Performance)
• App Framework: Tauri V2
• State Management: Local persistence via localStorage, native File System, and Windows Credential Manager (for API keys).

You can download the latest version from the Releases Page.

1. Node.js (v18 or higher)
2. Rust & Cargo (Required for building from source)

1. Clone or download this project folder.
2. Navigate to the project directory:

   cd Novelgen

3. Install dependencies:

   npm install

4. Launch the application:

   npm run tauri dev

To package the app into a single native installer or executable:

npm run tauri build

The final binary will be located in src-tauri/target/release.

To run the application (without building from source), ensure the following are available:

• Microsoft Edge WebView2: Required for the UI to render.
  • Windows 11: Pre-installed.
  • Windows 10: Usually pre-installed via Edge updates.
  • If the app fails to launch, download and install it from Microsoft's official site.

• Local: LM Studio (port 1234), Ollama (default port 11434)
• Cloud: Google Gemini, Ollama Cloud, OpenCode Go, OpenCode Zen, Cerebras (API key required, stored securely in Windows Credential Manager)

You can generate your novel using two distinct workflows:

This mode allows you to refine the story's direction before final generation.

1. Input Initial Idea: Enter a brief concept in the "Plot Seed" box, and/or click 🎲 Auto Seed to let the AI brainstorm a unique starting point.
2. Generate Plot: Click Generate Plot. The AI will create a chapter-by-chapter summary. For 13 chapters or more, plot generation is split into setup sections (1-4) and then each story part, reducing omissions in long outlines.
3. Refine Plot (Optional): Click ✨ Auto Instructions to have the AI automatically review your plot and suggest 5-10 specific improvement points, or manually add your own guidance in Plot Refine Instructions. Then click ✨ Refine. The AI first rewrites the setup sections, then refines each story part in order using the revised setup, already-refined earlier parts, and the remaining original chapter outline as boundary/context.
   • Use Start Part and/or End Part to refine only part of the plot.
   • Start only: refines from that part through the end. (Start Part: 0 for settings)
   • End only: refines from Part 1 through that part.
   • Same Start and End: refines only that part.
   • ✨ Auto Instructions supports both single-part and multi-part ranges. When a range is selected, it analyzes the selected parts with adjacent boundary parts as continuity context.
4. Review & Edit: You can manually edit the generated plot directly in the UI to fix inconsistencies.
5. Save Plot: Use the 💾 Save Plot button to store your outline locally in output/plot/.
6. Start Generation: Click Novel Generation. The AI will follow your plot exactly, chapter by chapter.
7. Refine Novel (Optional): Add manuscript-specific guidance in Novel Refine Instructions, then click ✨ Refine to revise the draft against the plot.
   • Use Start Chapter and/or End Chapter to refine only part of the draft.
   • Start only: refines from that chapter through the end.
   • End only: refines from the beginning through that chapter.
   • Same Start and End: refines only that chapter. When specifying a single chapter this way, you can click ✨ Auto Instructions to have the AI automatically review that specific chapter against the plot, previous chapter, and next chapter to generate 10 specific improvement points.

Generate multiple novels or handle high-volume creation automatically.

1. Setup: Enter your concept in the Plot Seed box and set the Batch Count.
2. Optional Refinement Settings:
   • Auto refine plot: Automatically refines the outline in chunks before starting novel generation.
     • Auto instructions: Automatically generates tailored improvement points for the plot during the refinement process.
   • Auto refine novel: Automatically runs the chapter-by-chapter refinement pipeline after the draft is complete.
     • Auto instructions: Automatically generates tailored improvement points for each specific chapter during the refinement process.
3. Execution: Click 🚀 Batch Start. The system will sequentially:
   • Brainstorm unique plots for each batch job.
   • Apply requested auto-refinements to plots and manuscripts.
   • Save all files (.txt and metadata .json) to the output/ directory.
4. Queue Management: Jobs are added to a task queue and processed one by one. Monitor progress via the Queue Size display.

The application manages its configuration and storage through the following structure:

• Key Configuration Files
  • system_prompt.txt: Stores your default system prompt. Use the Save icon in the UI to update this file.
  • gemini.txt (Legacy): Previously used for API keys; now automatically migrated to Windows Credential Manager for enhanced security.
• Storage Paths
  • output/: Primary directory for generated novel manuscripts (.txt).
  • output/plot/: Storage for generated and refined plot (.txt).
  • output/json/: Essential metadata and continuity JSON files used for resuming generation or refining drafts.

Define the global persona, tone, and constraints of your AI novelist.

• Use the System Preset dropdown to switch between styles (Standard, Web Novel, Epic Fantasy, etc.).
• Selecting Custom (File Default) reloads the contents of system_prompt.txt.
• Drag and drop a .txt/md file into System Prompt Details to load text instantly.
• Click the Save icon to persist your custom prompt to system_prompt.txt.

Adjust these in the sidebar for fine-grained control:

• Temperature: Higher values (e.g., 1.2) increase creativity; lower values (e.g., 0.5) make output more focused.
• Top-P: Controls the diversity of the vocabulary.
• Repetition Penalty: Helps prevent the model from repeating phrases or sentences.

• Drag and drop .txt/md files into the Seed, Plot, and Novel editors to replace the current text quickly.
• Each Seed / Plot / Novel preview includes an independent font size slider for easier editing and proofreading.
• Enable the Comfort checkbox beside each preview font slider to use a softer reading surface designed for long-form review.
• Use the theme toggle beside the NovelGen AI title in the sidebar to switch the whole app between light and dark appearance.

NovelGen AI maintains long-term continuity using a sophisticated layered memory architecture, allowing it to generate cohesive novels of any length:

Continuity metadata is saved as JSON files in output/json/.

1. Plot Context: Short outlines use the full refined plot. For novels of 13 chapters or more, the prompt uses the current story part plus the immediately previous and next parts in full detail, while distant chapters are reduced to title and content so the model keeps the master plan without wasting context.
2. Recent Chapter Summaries: A sliding window of the last 4 chapter summaries provides high-density context for immediate narrative flow.
3. Layered Story State:
   • Facts: Established canon facts that must remain consistent.
   • Open Threads: Unresolved plot points and "to-do" items for the narrative.
4. Character Status Memory: Dynamically tracks character locations, emotional states, and evolving relationships.
5. Narrative Arc Management:
   • Current Arc: The immediate conflict or goal being pursued in the present chapters.
   • Closed Arcs: Summarized history of finished story arcs, allowing for long-term recall without wasting tokens.
6. Sliding Prose Window: The final ~1,200 characters of the previous chapter are provided to ensure seamless transitions and consistent writing style.
7. Style & Expression Cooldown: An automated trope-detection system that prevents the AI from overusing specific transition phrases or descriptive clichés in narration.

If your model runner allows changing context length, these settings work well in practice:

• Recommended Minimum: ≈ plot_tokens * 2 + target_tokens + 1000 + 3000~5000
• Short plot outlines: 16k to 24k context is usually enough.
• General long-form novel generation: 32k context is the recommended default.

This project is licensed under the MIT License - see the LICENSE file for details.