Realistic AI Handwriting PDF Studio

I’m building an AI-powered handwriting studio that turns plain-text notes into print-ready PDFs that fool the eye into thinking they were written by hand. The engine must let me switch on demand between modern cursive, classic calligraphy, and casual handwriting, while a “customizable font” panel lets me adjust stroke weight, slant, spacing, and the natural randomness of real penmanship.

What I need from you is an end-to-end solution—from model design to final PDF export—that pairs visual authenticity with solid print fidelity. High-resolution (300 dpi, CMYK-friendly) output is non-negotiable; the pages have to survive a close look after coming off an office laser printer with no tell-tale artifacts.

Deliverables
• A handwriting synthesis engine (Python, TensorFlow/PyTorch or comparable stack) that ingests text and returns vector or high-resolution PDF files
• Style library covering the three handwriting families, built for easy expansion
• Configuration interface for all font customisation controls
• Documentation plus a short demo video showing the workflow end to end

Acceptance criteria
Given a 200-word sample note, at least 8 out of 10 blind testers must say the printout looks genuinely handwritten.

If you’ve worked with diffusion models, GANs, stroke-based synthesis, or OpenCV texture overlays, highlight that experience—those techniques appear well suited for the realism I’m after. A proof of concept two weeks from kickoff would be ideal, followed by a polished release once the quality bar is met.