When “split” implies audit-grade granularity
Operations teams pronounce “split” when they mean traceability: one PNG per logical page tied to log entries, drawing approvals, or hearing exhibits—not merely convenience slices. That mindset elevates naming contracts (revision letters, foreign IDs, ISO timestamps) and mismatch detection above aesthetic tweaks. Mixed ISO sheet sizes inside engineering rolls produce uneven thumbnail grids that tempt humans to misread missing assets; ordering must rely on metadata, not perceived height. Ai2Done pairs splitting with explicit progress so marathon PDFs never masquerade as frozen tabs. Post-export, automate equality checks—PDF page counts versus PNG counts plus spot perceptual hashes—to catch silent resampling. Evidence-grade pipelines should retain source PDF digests and timestamps beside raster derivatives so downstream reviewers can replay provenance without trusting filenames alone.
Page-by-page split in three steps
- Publish naming templates and folder taxonomy before rendering.
- Upload the PDF, split into per-page PNGs, and watch progress for anomalies.
- Script-verify counts/hashes, then ingest into PLM, ECM, or discovery tooling.