AI Generators for Handwritten SVG Output

Contact: Marco Peer

Note: The thesis is in collaboration with Forensisches Institut Zürich.

Overview

AI-based handwriting generators have made significant progress in recent years (Figure 1). However, most existing systems produce raster graphics. For reproducing handwriting with a pen plotter on paper, stroke paths rather than pixel-based images are required. Such information can be represented in SVG data (single-line vector graphics). This master’s thesis investigates whether AI-based handwriting generators exist that can produce such vector-based outputs. Furthermore, the thesis evaluates how convincing the generated handwriting appears when physically reproduced on paper.

Objectives

• Review the state of the art in AI-based handwriting generation
• Identify AI generators capable of producing single-line handwriting as SVG output
• Investigate methods to convert raster-based AI outputs into SVG stroke paths if direct solutions are not available
• Generate handwriting samples in the style of approximately 20 individuals
• Reproduce generated handwriting using a pen plotting device
• Evaluate the perceived authenticity of the generated and plotted handwriting in a study with laypersons

Methodology

• Literature Review: Survey existing approaches to AI-based handwriting generation
• Technology Analysis: Identify tools capable of producing SVG-based handwriting
• SVG Conversion: Develop and test workflows to convert AI-generated images into plausible stroke paths
• Handwriting Generation: Generate handwriting samples for approximately 20 individuals
• Physical Reproduction: Plot selected samples using a pen plotting device
• Evaluation Study: Assess whether laypersons can distinguish between plotted AI-generated samples and genuine handwritten samples
• Reporting: Document results in the master’s thesis and final presentation

Skills Required

• Programming skills
• Basic knowledge of statistics
• Interest in AI and handwriting examination

Created: 02.04.2026