TypeScribe v1.0

Need to send out a hundred organic Christmas cards? Or you have a fifty page essay due and Brush Script doesn't quite cut it? Got you covered.

TypeScribe is a tool that uses Machine Learning to convert typed text into realistic handwriting by introducing customizable degrees of randomness and variations, to make it all so believable.

TypeScribe	Totally not homework	A truly heartfelt letter	Christmas Cards too!?

Features

• Realistic Handwriting Generation with a Recurrent Neural Network (RNN)
• Choose from 12 predefined handwriting styles.
• Adjust line spacing, page size, margins, ink color, pen thickness, and more.
• User-friendly interface
• Automatically splits large texts into multiple lines, paragraphs and pages.
• Scalable SVG Output that maintains quality even on resizing

Getting Started

There are two ways to get things rolling:

1. Download the Executable:

   • Head to the Releases page and download the TypeScribe-v1.0.zip package.
   • Unzip the file and double-click TypeScribe.exe to launch the application.

2. Clone the Repository:

   • Clone this repo using git clone https://github.com/rudyoactiv/typescribe-handwriting.git
   • Navigate to the cloned directory.
   • With Anaconda installed, run conda env create -f environment.yml to create an environment.
   • Run conda activate test_hand to activate your environment.
   • Run the program with python gui.py

Usage

1. Launch the application with the command python gui.py or double click TypeScribe.exe.
2. Enter your text and adjust settings via the interface.
3. Preview the layout and handwriting style.
4. Click Generate.
5. Select a destination folder for your files.
6. Wait while it generates your output. May take a minute.

Known Issues

• Antivirus programs may flag (and even delete) the executable. This is due to the way Python programs are built and handled by Windows and cannot possibly be fixed. However, there is no cause for concern.
• The application itself may freeze while your file is being generated. No need to panic, it does not crash.

Future Plans

• A more responsive design
• More customization options
• Increasing the valid character set
• Fix antivirus false positives
• Build an installer to replace the zip
• Extra text formatting options (vary alignment, ink colors in a single document).

Acknowledgements

This project is inspired by the work in the handwriting-synthesis repository by sjvasquez, which provides the foundational implementation for handwriting synthesis using Recurrent Neural Networks (RNNs).

The handwriting synthesis in TypeScribe is based on the work presented in the paper Generating Sequences with Recurrent Neural Networks by Alex Graves. [Graves, A. (2013). Generating sequences with recurrent neural networks. arXiv preprint arXiv:1308.0850.]

---

Thank you for trying TypeScribe! ✍️