Open Source AI Platform

NovelVids
Turn Novels into Short Drama Videos

NovelVids is an AI-powered production platform that automatically converts novel text into style-consistent short drama video content — from entity extraction to final video synthesis — in one seamless workflow.

NovelVids Four-Step AI Production Pipeline

Every chapter processed by NovelVids follows a structured four-step AI workflow. Each stage feeds directly into the next, ensuring character appearance and scene consistency across every generated video frame.

01

Entity Extraction

AI performs deep analysis of novel chapters, automatically identifying and extracting key entities — characters, scenes, props, and their attributes — using advanced Natural Language Processing and Named Entity Recognition.

NLP NER
02

Asset Management

Extracted entities are organized into a unified asset library. AI generates reference images for each character and scene, establishing visual identity and consistency for all downstream production stages.

Image Generation Stable Diffusion
03

Storyboard Generation

Novel content is transformed into detailed storyboard scripts. Each shot receives precisely crafted prompts informed by the extracted entities and reference images, providing shot composition and narrative direction for video generation.

Text-to-Prompt Storyboarding
04

Video Synthesis

Based on storyboard scripts and reference images, NovelVids calls video generation models to produce style-consistent short drama clips. A dedicated merge service then assembles individual clips into complete episodes.

Seedance Sora Veo Vidu

Production-Ready NovelVids Capabilities

NovelVids is not a demo or proof-of-concept. It is an industrial-grade application built with a rigorous layered architecture, comprehensive test coverage across four layers (API, Controller, Model, Service), and production-ready engineering.

Novel Management

Full novel lifecycle management — metadata editing (title, author, synopsis, cover image), and AI-powered intelligent chapter splitting that automatically detects chapter boundaries for one-click segmentation of long-form novels.

Multi-Model Configuration

Flexible AI model configuration system allowing independent model assignment for text generation, image generation, and video generation tasks. Hot-swap between models with a single click — no application restart required.

Video Library

Global video management dashboard across all projects with real-time status tracking — queued, processing, completed, or failed. Supports in-browser video preview and batch management operations for efficient content review.

Entity Consistency

A dedicated asset resolver (asset_resolver.py) maintains character and scene consistency across all production stages. Once a character's visual identity is established, it persists throughout every generated frame.

Unified API Interface

All AI model integrations leverage an OpenAI SDK-style unified interface, enabling seamless compatibility with most major large language model APIs and effortless switching between providers without code changes.

Async Task Scheduling

Built-in asynchronous task executor (ai_task_executor.py) manages AI workloads efficiently. Long-running operations like video generation run in the background with real-time progress tracking and status reporting.

Supported Video Generation Models

NovelVids integrates with leading video generation models through its modular service layer. The services/video/ directory houses individual adapters for each model, making it straightforward to add new providers.

Se

Seedance

High-quality video generation with strong motion control and fluid character animation. Seedance delivers consistent scene rendering well suited for drama-style narrative content.

So

Sora

Advanced video generation supporting complex scene composition and extended sequences. Integrated via compatible API interfaces for flexible deployment options across different infrastructure setups.

Ve

Veo

Google's video generation model offering photorealistic output and precise prompt following. Ideal for scenes that require high visual fidelity and cinematic production quality.

Vi

Vidu

Developed by Shengshu Technology, Vidu excels at character-driven video generation with strong identity preservation across frames — critical for maintaining visual consistency in short drama production.

NovelVids Modern Tech Stack

NovelVids adopts a decoupled frontend-backend architecture. The backend enforces a strict four-layer separation (API, Controller, Model, Service) with RESTful API design, while the frontend delivers a responsive UI built on React 19 with TypeScript type safety.

Backend

FastAPI High-performance async web framework
Tortoise ORM Async ORM (SQLite / PostgreSQL)
Pydantic Data validation & serialization
OpenAI SDK Unified AI model interface
Uvicorn ASGI server

Frontend

React 19 UI framework
TypeScript Type safety
Vite Build tooling
Tailwind CSS Utility-first styling
shadcn/ui Component library

Backend Architecture

API Layer (api/) RESTful routing, parameter parsing, response formatting
Controller Layer (controllers/) Core business logic orchestration
Model Layer (models/) Data persistence via Tortoise ORM (SQLite / PostgreSQL)
Service Layer (services/) External AI model calls — LLM, image generation, video synthesis

NovelVids Use Cases

The AI short drama market is experiencing explosive growth. On TikTok alone, #ChineseDrama content has surpassed tens of billions of views. AI technology has dramatically reduced production costs — from 500,000 CNY per title to as low as 1,000-2,500 CNY per minute. NovelVids serves a broad range of professionals within this expanding landscape.

Short Drama Production

Individual creators and small studios use NovelVids to produce short drama episodes at a fraction of traditional costs. The AI-automated workflow eliminates the need for large production teams, cameras, and physical sets — turning a novel chapter into a video segment in minutes instead of days. Industry data shows AI short drama ROI typically reaches 1.1-1.8x.

Web Novel Promotion

Publishers and web novel platforms convert popular titles into visual drama trailers or episodic clips. These short-form videos serve as powerful marketing assets on platforms like TikTok, Instagram, and YouTube, driving readership back to the source material and amplifying the influence of novel IPs across video-first audiences.

Social Media Content

Content creators leverage NovelVids to rapidly generate short video series for YouTube, Instagram Reels, and TikTok. The consistent visual style across generated clips creates a branded, professional look that helps build audience engagement, sustain publishing schedules, and diversify content formats.

Education & Training

Educators and training professionals transform complex textual material into animated short dramas. By converting narratives, case studies, or historical accounts into visual stories, NovelVids makes learning content more accessible and engaging for diverse audiences across institutional and self-directed learning contexts.

Get Started with NovelVids

NovelVids requires Python 3.12+ for the backend and Node.js 20+ for the frontend. The platform runs on standard cloud servers — no local GPU is required, as video generation is handled entirely by cloud-based AI APIs, keeping infrastructure costs minimal.

1

Clone the Repository

Get the latest NovelVids source code from GitHub.

git clone https://github.com/Anning01/novelvids.git
2

Set Up the Backend

Install Python dependencies (uv recommended) and configure your AI model API keys in the .env file.

uv sync
3

Set Up the Frontend

Navigate to the web directory and install Node.js dependencies.

cd web && npm install
4

Launch NovelVids

Start the backend server and the frontend development server, then access the application.

uvicorn novelvids.app:app --reload --port 8000

Start Building with NovelVids

NovelVids is open source and free to use. Clone the repository, configure your preferred AI models, and begin transforming novels into short drama videos today.