Book Digitization Assumptions Lab DESTRUCTIVE FEED

Ingestion Pipeline Configuration 2,500 Volumes

Destructive fixture: rights review needed

The fixture assumes total physical loss. Ownership, rights searches, conservation condition and lawfulness of any use are not investigated.

Physical Ingestion & Material Disposition Flow 450 Pages / Min
785,000
Modeled Page Count
628 M
Assumed Token Yield
$23,550
Modeled Scan Cost (USD)
2,500
Assumed Destroyed Volumes
Severe (94/100)
Authored Loss Band
1,120
Assumed Orphan Proxy Count
Fixed Example Inventory (10; batch size 2,500) Unverified fixtures
METHOD COMPARISON
Guillotine Speed: Assumed 11.25×
Cradle Archival: Assumed Intact
Unit Cost Gap: $0.35/p
Model Flag: Destructive fixture

Audit the evidence behind a digitization estimate

Which outputs follow assumptions, and which require inspecting the books?

This browser worksheet multiplies volume counts, average pages, scan-price assumptions and chosen token densities. It performs no scanning, OCR, tokenization, acquisition, owner search or training. The ten inventory entries and their rights/disposition labels are fictional fixtures.

Prices, speeds, destruction rates, the 80% orphan proxy and 72/94 loss bands are authored examples. No legal risk probability, rights clearance, preservation quality or chain of custody is established. Choosing a cradle does not establish copyright permission.

Try a worked example

The default 2,500 volumes × 314 assumed pages gives 785,000 pages. At $0.03/page, modeled cost is $23,550. At 800 tokens/page, yield is 628 million. The out-of-print input gives 1,400 volumes; multiplying by the invented 80% proxy gives 1,120. That count is not a researched list of unlocatable rights holders.

Switching only the workflow to cradle keeps pages and token yield unchanged, changes modeled cost to $298,300, and sets modeled destruction and orphan-at-loss count to zero. Neither toggle inspects an object or grants permission.

Check the inputs before treating them as a corpus

The three era percentages are independent controls, not a normalized partition. Setting all three to 100% gives 300% in displayed categories. Counts may overlap or exceed intake. The pre-1928 input changes its display but is not used in the numerical metrics. Presets also change hidden average pages: 314, 380 or 260.

Token-density selections use fixed 800, 1,150 or 620 tokens/page, then round millions. The ten-row ledger remains fixed across presets and selections, including its stored 800-token label; its page counts and era mix do not drive the batch calculation. The flow’s link widths stay 14 pixels and do not encode quantities. Listed 450/40-page speeds do not calculate project duration.

Separate publication history from rights findings

The Copyright Office’s duration guidance distinguishes creation, publication, authorship and older protection/renewal rules. The retained pre-1928 control is an era fixture, not a current universal cutoff or verified public-domain determination for a particular edition. An out-of-print checkbox and an assumed orphan fraction do not establish permission. U.S. Copyright Office: Duration of Copyright

The Office explains that fair use considers purpose, the work’s nature, amount used and market effects through a fact-specific inquiry. No predetermined page percentage or scanner choice resolves that inquiry. This worksheet’s flag is an authored display rule, not a legal decision about digitization or AI training. U.S. Copyright Office: Fair Use Index

Assess quality with actual evidence

FADGI’s technical guidance pairs imaging practices with evaluation targets, software, testing and monitoring. A cost estimate or a zero-loss assumption is not a conformance test. Record the actual edition and object identifier, capture settings, image checks, OCR errors, permissions and changes to the physical item separately from this arithmetic worksheet. FADGI: Technical digitization guidelines

Read the exports as local comparisons

JSON records current configuration and computed assumptions with the fixed example ledger. CSV contains only that ledger and its assumed disposition; it does not contain the batch inputs or all metrics. Changing OCR density does not transform or measure the sample text. Browser storage retains inputs; Reset restores the default fixture. Neither export restores a session or constitutes an acquisition, preservation or rights dossier.

Sources and further reading
Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.