How to prepare manuscript images as figures for publication

A publication figure should show the evidence needed for a specific claim at a size that lets readers assess it. Preserve the fuller working photograph separately.

Start with the publisher's current image specifications and the repository's reuse terms. Record the crop, processing, caption, credit line, and rights basis for every submitted file.

Choose the evidence the figure must show

The first decision is not technical. For each figure, ask what the reader must be able to see for your argument to land: a corrected word, a change of hand, a collector's stamp on a flyleaf, the ruling pattern of a page. Then choose the crop that shows that thing. A full-page reproduction is the right answer when the argument is about the page, layout, decoration hierarchy, the mise-en-page of a Qur'an or a book of hours. When the argument is about six words of interlinear gloss, a full page shrinks those words to illegibility and wastes the space the press gave you. Crop to the claim.

This is also the moment to check that the photo can carry the weight. A shot that was fine for your own reading, slightly soft, shadowed in the gutter, taken through glass, may not survive enlargement in print. Better to discover that now, while re-photographing or ordering new images from the repository is still possible.

Resolution: the arithmetic

A common press requirement for color and grayscale photographs is 300 pixels per inch at reproduction size. Line art may require 600, 1,200, or another value. Use ppi for raster-image sampling; printer dpi describes output dots. Your press's specification controls.

Multiply the planned width in inches by the required ppi to find the pixel width. A five-inch figure at 300 ppi needs 1,500 pixels across. Apply the calculation to the final crop rather than the uncropped source file.

Upsampling adds pixels but cannot recover detail the source did not capture. A production workflow may still request resampling, so follow the press's instructions and do not treat enlarged pixel dimensions as new evidence. When the source lacks detail, reduce the print size, obtain a better photograph, or request an appropriate reproduction from the repository.

Formats: masters and working copies

Use the file format named in the publisher's instructions. TIFF can store raster data losslessly. JPEG uses lossy compression, and repeated saves can add artifacts around penstrokes, letters, and surface texture.

The working practice that follows:

• Keep an untouched master of every image, the file as it came from your camera or from the repository. Never edit it; edit copies.
• JPEGs can serve as working copies for reading, sorting, and draft layouts. Even one save can create visible artifacts at low quality or around fine detail, so retain the source and avoid repeated JPEG editing.
• When you prepare the final figure, go back to the master, make the crop and adjustments once, and save the result as TIFF. If the repository supplied a JPEG and that is all that exists, a TIFF saved from it doesn't restore what compression discarded, but it does stop further loss.
• Record the source file, crop, adjustments, output dimensions, and export settings for each submitted derivative.

Details, context, and honest adjustment

When a figure shows a detail, consider whether the reader needs to know where on the leaf it sits. For a paleographical point, usually not. For an argument about placement, a marginal note keyed to a passage, an addition squeezed into a specific blank, a detail floating free of its page loses the evidence. The standard solution is a pair: the detail at legible size, plus a small full-page locator image or a caption that gives the position precisely. Some presses count that as two figures; check before you promise.

Adjustment becomes an integrity question when it changes evidence. Cropping, straightening, tonal correction, color correction, sharpening, denoising, and compositing should follow the publisher's image policy and the needs of the argument. Do not suppress relevant damage, invent missing detail, or alter a material feature whose appearance supports the claim. Describe substantial processing, spectral composites, and digital reconstruction in the caption or methods section. Disclosure is required, but disclosure does not make a misleading image acceptable.

Captions and credit lines

Caption requirements differ among presses, journals, repositories, and license documents. A useful record often includes:

• Repository and shelfmark: the city, institution, and the manuscript's call number, the coordinates that let anyone find the object.
• Folio: the leaf and side shown (f. 24v), or the page number if the manuscript is paginated. The conventions are covered in what recto and verso mean and foliation vs. pagination; for a detail, add the position (lower margin, lines 4–7).
• Date and origin, when known, and a short identification of the content if the figure needs one.
• The credit line: the acknowledgment required by the rights grantor. Copy required wording from the license or repository record.

Follow the publisher's model and keep a repository-mandated credit line verbatim. A press may place the credit outside the descriptive caption.

Permissions: start early

Rights requests can take longer than image preparation. Ask the publisher when documentation is due, then request any needed permission once a figure becomes probable. Record the grantor, reference, permitted uses, expiry, and required credit line. The guide to tracking image reproduction rights covers that record. Check media, territory, edition, language, term, and open-access conditions rather than assuming that permission for one publication covers another.

The figure list

Maintain one row per proposed figure once the project contains several images. The table should connect each figure number to its file, manuscript location, rights basis, and credit line:

Add columns for print size, processing notes, delivery, policy URL, review date, territory, term, or other conditions your project needs. The rights cell should distinguish a license, permission, public-domain assessment, CC0 dedication, statutory exception, pending request, and unresolved status. Do not compress those different bases into “cleared.”