Thanks! Yeah, the parsing side was actually one of the messier parts.

The hardest thing wasn't just reading the PDF itself — it was making the output predictable enough for the flipbook renderer. PDFs can have very different page structures, fonts, image layers, and rendering behavior.

I eventually separated the pipeline into two steps: first normalize the input into a consistent page-image sequence, then let the flipbook renderer handle the interaction and page-turn effects.

That separation made a big difference. It also meant I could support other formats without making the renderer understand every file type.

The page-flip animation was probably the more visible challenge, but the preprocessing layer was what saved me from a lot of headaches later.