Public Beta · Web Application

StoryGEN AI

A browser-based workspace for AI-assisted narrative design, guiding authors and game designers from an initial idea to a structured, reviewed story through a six-stage workflow.

StoryGEN AI develops the research behind my master's thesis into a publicly accessible narrative design tool. Authors shape the creative direction, inspect generated results, edit their work and explicitly approve the versions used in later stages.

My Contributions

• Developed the React and TypeScript workspace, narrative graph explorer and dedicated editors
• Adapted the Python narrative engine to run inside a browser Web Worker using Pyodide and WebAssembly
• Implemented local SQLite storage with persistent IndexedDB snapshots, checkpoints and recovery
• Built version history, source dependency tracking, comparison and targeted updates
• Integrated ChatGPT / Codex and Claude connection modes, with optional Windows connectors
• Implemented project transfer, full archive backups and concept PDF export
• Prepared the containerised web release, bilingual interface and user documentation

From Idea to Story

Six stages organise the process: Preparation, Structure, Chapters, Story, Review and Summary. The workflow connects the creative brief to narrative events, characters, chapter plans, encounters and gameplay requirements, then develops and reviews the text before producing a shareable concept document.

Each stage uses approved source versions. Authors can explore narrative graphs, edit structured content, compare revisions and request focused changes while preserving a traceable history of their creative decisions.

Python Engine, Browser Workspace

The Python engine runs through Pyodide in a dedicated Web Worker. An in-process adapter handles application requests locally, reusing the narrative pipeline without requiring users to install Python or start a local server. The React interface remains separate from the engine's execution.

SQLite operates inside the worker, with persistent snapshots saved in IndexedDB. Project archives stay in the user's browser; relevant story context is sent to the selected provider for AI generation. The web service distributes the application and relays encrypted provider traffic, without hosting the story database.

Versions and Reliable Saves

Version history tracks dependencies between narrative stages. Authors can compare results, activate previous versions, protect important work and inspect the impact of deleting linked versions. Checkpoints, single-writer protection across tabs and emergency backup export help preserve work when an operation or local save is interrupted.

Public Beta

Released as a public web beta in October 2026, with an English and Italian interface, an in-app guide and project import/export. Browser AI connections require a supported provider account with model access and available quota; local Windows connectors are optional.

The current workflow uses a three-act structure and sequential chapter progression. Complete playable branching routes and alternative dramaturgical structures are future directions. Demo mode uses static example content; personalised generation requires an AI connection. Browser data does not automatically sync between devices, so projects can be transferred through export and import.

Research foundation: StoryGEN AI - Master's Thesis