ComfyUI Script-to-Video Suite

Transform PDF scripts into AI-ready video generation prompts through an intelligent pipeline.

A ComfyUI custom node suite that converts long-form PDF scripts into structured storyboards and detailed video generation prompts using AI-powered parsing and scene breakdown.

Badges

Features

PDF Script Processing: Extract and chunk text from PDF screenplay/script files with configurable overlap
AI-Powered Storyboarding: Generate detailed storyboard panels using Gemini AI via relay server
Prompt Engineering: Convert storyboard scenes into optimized video generation prompts
Modular Pipeline: Three independent, chainable nodes for maximum flexibility
ComfyUI Integration: Seamless workflow integration with custom output types

Installation & Setup

Prerequisites

ComfyUI (latest version recommended)
Python 3.10+
PyMuPDF for PDF processing
Git

Installation Steps

Clone the repository into your ComfyUI custom nodes directory:

cd ComfyUI/custom_nodes/
git clone https://github.com/Long-form-AI-video-generation/script-parser-custom-node.git

Install dependencies:

cd script-parser-custom-node
pip install PyMuPDF requests

Restart ComfyUI to load the custom nodes
Verify installation - Look for "Script To Video Suite" in your node menu

Relay Server Configuration

The suite uses a Gemini AI relay server for processing. The default endpoint is pre-configured.

To use your own relay server, update the RELAY_SERVER_URL in the gemini_relay_client.py file.

Troubleshooting

Nodes not appearing in ComfyUI:
Ensure the repository is cloned into ComfyUI/custom_nodes/ and restart ComfyUI.
Relay server connection errors:
Check your internet connection and verify RELAY_SERVER_URL in gemini_relay_client.py.
PDF parsing issues:
Confirm that the PDF is not encrypted or image-based (use OCR first if needed).

Architecture Overview

System Design

┌─────────────────┐
│   PDF Script    │
│    (Input)      │
└────────┬────────┘
         │
         ▼
┌─────────────────────────┐
│   1. PDF Chunker        │
│   - Extract text        │
│   - Create overlapping  │
│     chunks              │
└────────┬────────────────┘
         │
         ▼
┌─────────────────────────┐
│ 2. Storyboard Generator │
│   - Process via Gemini  │
│   - Create panels       │
│   - De-duplicate        │
└────────┬────────────────┘
         │
         ▼
┌─────────────────────────┐
│ 3. Prompt Generator     │
│   - Split into scenes   │
│   - Generate prompts    │
│   - Format for AI video │
└────────┬────────────────┘
         │
         ▼
┌─────────────────┐
│ Video Prompts   │
│   (Output)      │
└─────────────────┘

Core Components

The suite registers three primary nodes with ComfyUI's node system, each designed for a specific stage of the script-to-video conversion pipeline.

Script-to-Video Pipeline

Stage 1: PDF Chunking

The PDF Chunker node extracts text from PDF files and splits it into manageable chunks with configurable overlap to maintain context between segments.

Key Parameters:

pdf_path: File path to the source PDF script
chunk_size: Characters per chunk (4000)
overlap_size: Overlap between chunks (400)

Stage 2: Storyboard Generation

The Storyboard Generator processes each chunk through the Gemini relay server to create detailed storyboard panels with visual descriptions and action notes.

Processing Flow:

Iterates through text chunks
Sends each chunk with prompt template to relay server
Collects AI-generated storyboard panels
De-duplicates based on action descriptions

Stage 3: Prompt Generation

The Prompt Generator converts storyboard scenes into detailed, AI-ready video generation prompts optimized for models like Stable Diffusion Video or RunwayML.

Scene Processing:

Node Reference

1. PDF Chunker (S2V)

Property	Description
Category	Script To Video Suite
Input Types	`pdf_path` (STRING), `chunk_size` (INT), `overlap_size` (INT)
Output Type	`CHUNKS` (custom type)
Function	`process_pdf`

Purpose: Extracts and chunks PDF script text for processing

2. Storyboard Generator (S2V)

Property	Description
Category	Script To Video Suite
Input Types	`chunks` (CHUNKS), `prompt_template` (STRING)
Output Type	`STRING` (storyboard_text)
Function	`generate_storyboard`

Purpose: Converts script chunks into structured storyboard panels using AI

3. Prompt Generator (S2V)

Property	Description
Category	Script To Video Suite
Input Types	`storyboard_text` (STRING), `prompt_template` (STRING)
Output Type	`STRING` (final_prompts)
Function	`generate_prompts`

Purpose: Generates AI video generation prompts from storyboard scenes

Example Usage

Basic Workflow

[PDF Chunker] → [Storyboard Generator] → [Prompt Generator] → [Save Text]

Workflow Configuration

Add PDF Chunker Node

pdf_path: "C:/Scripts/my_screenplay.pdf"
chunk_size: 4000
overlap_size: 400

Connect to Storyboard Generator

prompt_template: "You are a storyboard artist. Create storyboard 
panels from the following script text. Use '--- PANEL BREAK ---' 
between panels..."

Connect to Prompt Generator

prompt_template: "You are a prompt engineer for an AI video generator. 
Convert the following storyboard scene into detailed prompts..."

Sample Output Format

--- SCENE BREAK ---

PANEL 001
ACTION_DESCRIPTION: Character walks through foggy street at night
VISUAL_PROMPT: cinematic shot, noir atmosphere, streetlights through fog, 
lone figure silhouette, moody lighting, 4k quality

--- SCENE BREAK ---

PANEL 002
ACTION_DESCRIPTION: Close-up of character's face showing concern
VISUAL_PROMPT: dramatic close-up, concerned expression, rim lighting, 
shallow depth of field, cinematic color grading

Development Guide

Project Structure

script-parser-custom-node/
└── comfyui-script-to-video-suite/
    ├── __init__.py                    # Node registration
    └── s2v_nodes/
        ├── gemini_relay_client.py     # API relay client
        ├── pdf_parser_node.py         # Legacy parser 
        ├── s2v_chunker_node.py        # PDF Chunker node
        ├── s2v_storyboard_node.py     # Storyboard Generator
        └── s2v_prompt_gen_node.py     # Prompt Generator

Dependencies

PyMuPDF (fitz): PDF text extraction
requests: HTTP communication with relay server
re: Regular expression for scene parsing
json: Data serialization

Environment Setup

Development Installation:

git clone https://github.com/Long-form-AI-video-generation/script-parser-custom-node.git
cd script-parser-custom-node
pip install -e .

Testing Nodes:
- Load ComfyUI in development mode
- Add nodes to test workflow
- Monitor console output for debugging

Reporting Issues

Use GitHub Issues for bug reports
Include ComfyUI version, Python version, and error logs
Provide sample PDF or workflow JSON if applicable

Maintainers & Acknowledgements

Maintainers

Long-form AI Video Generation Team

Repository: Long-form-AI-video-generation

Related Projects

Script Parser to Storyboard - Complimentary repository.

Contributions, feedback, and ideas are always welcome! let’s build the future of AI video together!

⭐ If this project helps your workflow, consider giving it a star! ⭐

Name		Name	Last commit message	Last commit date
Latest commit History 11 Commits
.vscode		.vscode
comfyui-script-to-video-suite		comfyui-script-to-video-suite
.gitignore		.gitignore
Contributing.md		Contributing.md
README.md		README.md

Yeabebe/script-parser-custom-node

Folders and files

Latest commit

History

Repository files navigation