An image visible in a preview is not yet proof that the distributable EPUB contains that image. Ruby annotation has the same two-stage problem: plausible source notation does not guarantee the expected element after export. For the query “text editor preview ruby images EPUB,” this guide separates human visual review from package inspection. In the fixed EPUB, the measured package result was four chapters, four ruby elements, twelve image references, twelve packaged images, and zero missing resources.
Choose One Representative Specimen Per Chapter
Build a specimen matrix before looking at the output. Each row should name the chapter source, ruby base text and reading, source image, special notation, and expected XHTML destination. The controlled book used here had four chapters; each chapter supplied one ruby specimen and three image references.
Preselecting examples prevents reviewers from approving only the easiest success after export. If the source deliberately contains an old path or difficult character, record the expected failure before running the test. Keep correction and inspection in separate passes so the specimen does not change merely to fit the result.
Use Preview for Meaning and Visual Placement
In the graphical preview, a human reviewer should decide whether the reading belongs to the intended base text, the image has the intended orientation and alternative description, and a page break occurs at the editorial boundary. The useful question is not whether something appeared, but whether it expresses the manuscript’s intent.
The deterministic record for this article did not reproduce the live preview interface. It therefore does not establish font overlap, rendered size, scroll position, or selected state. Those observations require an application version, a fixed window or viewport, and captured before-and-after evidence. Keep that visual result separate from the package result below.
Open the EPUB and Map XHTML to the Manifest
After export, open the EPUB as a ZIP package and inspect each chapter XHTML file. In the fixed result, every chapter contained one ruby element, for a total of four. Each chapter referenced three images, producing twelve references. The OPF manifest contained the corresponding twelve image resources, and the missing-resource count was zero.
Equal totals alone are not enough. Connect every XHTML reference to a manifest item and then to a real file inside the package. Twelve unrelated images could satisfy the count while leaving the chosen specimen absent. Compare chapter, referenced name, packaged name, and media type in one table.
Inspect the ruby content rather than counting tags alone. Each ruby element should contain the expected base text and reading text in the same chapter named by the specimen matrix. A misplaced but well-formed annotation still increases the count, so the four semantic pairs must be compared individually with their source rows.
Reconcile Chapter Order in nav and NCX
Complete resources do not guarantee correct reading order. The fixed EPUB’s navigation document and NCX each contained four chapter entries in the same label order. The inspection also connected the OPF spine to the chapter XHTML files, looking for an omitted or duplicated chapter.
Do not collapse navigation, resources, and notation into one pass box. A book can package every image while ordering chapters incorrectly, or present the right order while missing a ruby element. Independent columns make the smallest necessary correction visible.
Repair an Image That Previews but Is Not Packaged
A preview may load a local absolute path even when the relative reference needed by the EPUB is broken. When a missing resource appears, place the XHTML reference string, OPF manifest href, and actual packaged filename next to one another. Check capitalization, extension, escaping, and directory depth, then export a new file.
After recovery, rerun all twelve mappings rather than only the chapter that failed. A shared path or export-setting change can affect other chapters. Preserve the earlier EPUB under a different name and tie every inspection result to a SHA so reviewers cannot confuse versions.
Also reopen the new EPUB from its final handoff location. A package that passes in a temporary export directory can be confused with a previous build during delivery. The filename, SHA, export time, and inspection report should identify the same artifact; otherwise the technically correct file may never reach the next reviewer.
Keep Local valid=true Within Its Boundary
The controlled structure check returned valid=true. That result means the tested package met the conditions of that local inspection. It does not establish retailer acceptance, copyright clearance, or identical rendering in every reading application.
Rune Studio’s public product material describes Markdown image, link, and EPUB workflows. That documentation supports the product capability boundary; the four-chapter artifact supports the specific values reported here. Do not blend those sources or infer an untested device display from a clean package report.
Hand Off Visual and Package Results Separately
The final matrix should retain a visual-review column for every chapter and separate columns for ruby elements, image references, manifest resources, navigation order, and NCX order. The deterministic package findings are four ruby elements, twelve references, twelve image resources, no missing targets, and matching four-entry navigation lists.
Leave the live preview rendering marked unmeasured until a reviewer actually checks the base text, reading, images, and page breaks on screen. Open one representative chapter on another reader as a distinct compatibility sample and record any device-specific difference. Output review is complete only when editorial appearance and package integrity have both received their own evidence.
If the alternate reader disagrees, retain both observations with the reader name and version. Do not edit source merely to make one application look like another until the project has chosen its supported target. Compatibility evidence informs that choice; it is not a universal promise about every reading system.