Estimate handwritten-source transcription from representative pages, not from page count alone. Separate image preparation, initial transcription, uncertainty resolution, and editorial use in the book. Measure how much human checking the sample requires, and preserve unreadable passages as unresolved evidence rather than paying for repeated guesses that merely sound convincing.
Inventory the actual source conditions
Group the material by handwriting, page condition, language, and document type. A clear recent notebook, a faded letter, and a densely corrected draft can require very different work. Count the pages in each group so the estimate reflects the collection's composition instead of treating all pages as interchangeable.
Record whether the images are complete and usable. Missing margins, poor focus, and overlapping sheets may prevent reliable transcription regardless of the tool. Identify material that needs a better capture before estimating repeated processing of an image that cannot support a trustworthy reading.
Choose samples that reveal difficulty
Select both ordinary and difficult pages from each meaningful group. Include abbreviations, crossed-out passages, unusual names, and changes of hand where they occur. A sample chosen only for legibility gives a misleading estimate of the work needed to prepare the complete source collection.
Define what the transcription should preserve. A reading transcript may differ from a detailed record of deletions and insertions. Decide which product the book needs before comparing effort, and document the convention for uncertain or unreadable text so reviewers apply the same standard throughout the sample.
Measure generation and verification separately
If an AI tool is used, record the actual input, requested output, and provider billing for the sample under the current applicable pricing. Do not assume that image processing is billed like ordinary prose drafting. Keep the pricing source and the date of the estimate with the project record.
Then measure the human review needed to compare the result with the image. A quick initial transcript can still require extensive checking, especially when plausible words replace unfamiliar names. Separate these costs so a low processing charge does not conceal the main effort required to make the material usable.
Plan an uncertainty route instead of unlimited retries
Create a way to flag uncertain readings and connect them to the relevant image region. A reviewer should be able to inspect the evidence without searching an entire folder. Keep proposed alternatives clearly distinguished from accepted text, especially when the choice would change a date, name, or claim used in the book.
Set a practical decision point for escalation, better source capture, specialist review, or leaving the passage unresolved. Repeatedly asking a tool for another answer does not create new evidence. Include a limited retry allowance only where the sample shows a concrete reason it may improve the process.
Separate transcription from adaptation into prose
An accepted transcript is source material, not necessarily a finished chapter. Estimate selection, contextual explanation, attribution, and manuscript editing separately. If the book quotes a letter, the workflow differs from using the letter as evidence in a narrative account, and the preparation estimate should reflect that distinction.
Keep the project's source-use records with the transcribed material. Technical access to a scan does not settle whether or how its contents can be published. Avoid folding that question into the transcription budget as if successful text extraction also established publication permission.
Scale the estimate by source group
Apply the sample findings to comparable groups and show the assumptions behind the total. Keep particularly difficult pages as a separate allowance rather than spreading their uncertainty invisibly across every page. This lets the project decide whether to expand the work, narrow the selection, or obtain better source material.
After the first substantial batch, compare actual accepted output with the estimate. Track unresolved passages and review time as well as completed pages. A realistic plan measures material that can support the book's intended use, not simply the number of pages for which a tool produced something resembling a transcript.
Frequently asked questions
Clear answers for this publishing decision.
Can I estimate the job by counting handwritten words?
Word count alone misses differences in legibility, corrections, image quality, and review needs. Use representative samples and explain how their conditions match the remaining material.
Should I accept the most common reading from several AI attempts?
Agreement among attempts is not independent evidence that a reading is correct. Compare the result with the source and preserve unresolved uncertainty when the image does not support a confident answer.
What should a sample budget include?
Include capture preparation, processing, human comparison, uncertainty handling, and the separate work needed to use the transcript in the manuscript.
When should the estimate be revised?
Revise it when source quality, transcription conventions, language needs, or the rate of unresolved passages differs materially from the sample assumptions.
What does this book cost to print and produce, and what should the list price be?
Printing cost follows from the trim, ink, and page count, the list price sets the royalty after that cost, and the production budget should be estimated before generation starts, so compare these numbers before you commit to a format or a plan.
See what a book project costs in credits