Compare Two Pitch PDFs When the New Draft Has Reflowed

When a pitch PDF has been rearranged, establish which passages correspond before deciding which words changed. A moved page is not necessarily a rewrite. A rewrapped paragraph is not necessarily new copy. And a shorter sentence can contain the most consequential change in the file.

Choose a comparison route that suits the document's organization, check it against changes you already know about, and read uncertain matches in context. The deliverable is an inspectable change record—not a certificate that the latest filename is the approved film.

Demonstration status: The original paired PDFs below were created, rendered, and checked locally. No native Acrobat Pro comparison was run. Acrobat settings are documented procedures, not observed report behavior; this article does not report a completed Acrobat run.

Start with four changes you can account for

The companion files, Last Light A and Last Light B, are an original fictional pitch exercise. They are deliberately plain, four-page PDFs. They contain no real production commitments or client material.

Before generating or comparing them, a separate change key specified the differences. Version A runs cover, story, project context, visual approach. Version B moves project context ahead of story. It also narrows the story's text area without changing its words, strengthens an access claim, and changes the visual-approach heading's appearance.

The access wording is the important trap:

A: For this fictional exercise, location access is being discussed.
B: For this fictional exercise, location access is confirmed.

Those sentences do not make the same claim. In a real pitch, the second would require a basis that the first does not assert. A comparison tool can expose the replacement; it cannot provide the missing confirmation.

The local observations and paired-page image show what was actually checked:

“Normalized” here means that runs of whitespace were collapsed before comparing extracted text. Punctuation and words were retained. The exercise prints a unique identifier on each page, and the checker paired pages by that identifier rather than inferring correspondence from layout or position as Acrobat would. It is not a replacement PDF-comparison product.

That limitation is useful to keep visible. The fixture gives us a known reference against which a later native run can be judged. It does not let us announce in advance that any particular Acrobat setting found all four controls.

Identify the pair, including what “new” means

Save the two originals and compare working copies. Record their names, where they came from, and the roles they are supposed to have. For this exercise, A is the baseline and B is the revised input. Neither is a production-approved document.

In real work, “new” may mean the most recent export, a designer's layout pass, the producer's marked-up copy, or a file someone renamed after receiving it. These roles are not interchangeable. A comparison can proceed with an agreed pair even when approval remains unresolved, provided the report says what is being compared and does not quietly resolve that separate question.

Record both physical PDF page numbers and useful identifiers such as section headings. “The change is on page three” is inadequate here: the project-context page is A3 but B2. A reviewer opening only the new PDF would otherwise inspect the story page.

File hashes are helpful for tying a report to exact bytes. The example includes them in its local observations. A hash establishes which file was checked; it says nothing about who authorized its contents. Do not turn file identification into creative authority.

Before choosing a route, open representative pages. Look at whether sections live on independent pages, prose flows through page boundaries, or visible writing exists only inside page images. Inspect the actual files rather than assuming that the word “deck” in a filename tells you how their content is stored.

Match the comparison model to the problem

In Acrobat Pro, start from All tools > Compare files, assign the old and new inputs, then open Settings. Adobe documents three relevant models: Reports, Spreadsheets, Magazine Layouts compares continuous text; Presentation Decks matches similar pages as individual documents, including moved pages; Scanned Documents, drawings or illustrations uses page-image comparison. Old and new page ranges can be scoped separately.[^1]

For Last Light, Presentation Decks is a reasonable first native trial because the story and project-context sections remain separate pages. The paragraph wraps differently within its page; it has not become continuous prose flowing from one page to the next. Do not select the continuous-text route solely because you see different line endings.

A different pitch might expand a two-page narrative into three pages, with the ending of an old page becoming the middle of a new one. In that case, treating every page as an independent unit would impose a boundary that the prose no longer respects. That is the kind of problem for which the continuous-text model is worth testing.

An image-based page raises another question: are you trying to compare readable wording, visible artwork, or both? A different scan position or background does not automatically imply changed meaning. Yet a text-oriented account alone may fail to answer whether the visual material has changed. Define the question before trusting a result that addresses only part of it.

For a mixed package, there may be no sensible single setting for every section. A full run can establish the overview; a scoped second run can address one uncertain passage or image. Record the limitation of each run instead of treating two reports as interchangeable votes.

Check the meaning change beside the reflow

Start the review with the access sentence. Find the project-context section in both originals. Compare the complete sentence, not merely the highlighted fragment. Keep its surrounding qualifications visible.

In the local fixture, the extracted comparison identifies a replacement from being discussed. to confirmed.. It reports no other wording change in the context paragraph, and the explicit fictional-exercise label appears in both versions. The narrower meaning change is therefore clear: the access status became stronger, while the rest of the paragraph's wording stayed as it was.

Now move to the story pages, A2 and B3. The local normalized-text check found the same words and punctuation in the same sequence. Rendering and inspection show different wrapping: seven body lines in A, ten in B. That is evidence of reflow in this particular controlled pair, not a rule that a denser-looking page is always unchanged.

If a native comparison describes the rewrapped paragraph as deleted and inserted, do not treat that wording as an instruction to approve an entire new synopsis. Read the passages. A report's presentation of correspondence and the editor's classification of a change are different layers of work.

The reverse mistake is more dangerous: calling the whole pass “just layout” because most visible differences are line breaks. The project-context page also wraps differently after the shorter status wording. Here, changed wrapping accompanies changed meaning. Appearance alone cannot distinguish that case from the story's unchanged wording.

Use the controls to ask specific questions of the native run. Did it associate A3 with B2? Did it expose the access replacement? Did it preserve a reviewable correspondence between A2 and B3? Did it account for the heading's changed appearance when the review included formatting?

A missed control is a finding, not a reason to tidy the report before showing it. Retain the actual result and add the manual observation or targeted second run that resolves the problem.

Distinguish an excluded comparison from a hidden result

Adobe's Compare text only option excludes graphic differences. Separately, the report's Filter controls determine which difference categories are visible; Adobe says annotations, formatting, and background differences are off by default.[^1]

Those are different questions. What did you ask the comparison to examine? What are you currently allowing the report to show?

The visual-approach page makes that distinction concrete. Its heading says the same thing in both files. The local text comparison therefore reports equality. Its font and color changed, and the rendered page differs. A record restricted to wording cannot serve as a record of the complete visual pass.

For an editorial-only review, that restriction may be intentional. Say so. For a review that also includes image replacement, typography, or presentation changes, make those concerns explicit in the settings record and inspect the relevant source pages.

Do not report “no formatting changes” when you have only established that no formatting changes are visible in the current view. Likewise, do not assume that turning on a display category repairs an earlier decision to exclude the relevant material from comparison. Check which stage imposed the restriction before rerunning or drawing a conclusion.

The accompanying local test does not reproduce Acrobat's filter behavior. Its heading control is an input for that test, not a screenshot of a successful UI operation. A completed native demonstration should retain the settings and actual report states rather than an idealized reconstruction of what they ought to look like.

Resolve ambiguous correspondence without inventing certainty

Our exercise uses unique page IDs. Real decks can have several pages titled “Story,” repeated character names, or nearly identical project notes. A similar-looking page may be the wrong counterpart.

Suppose a real package contains two location paragraphs, one about a proposed cinema and another about a workshop. Both say access is being discussed. In the revision, only one says confirmed, but the headings have also changed. Searching for “confirmed” locates the stronger statement; it does not establish which old paragraph corresponds to it.

Read the project name, location description, surrounding sentences, and any reliable source-page identifiers. If those still do not settle the match, label the correspondence unresolved. Do not silently merge the paragraphs, infer that both locations are confirmed, or choose the counterpart that produces the neatest report.

A scoped comparison can help where it isolates a known pair. Be careful to map the scopes correctly: A2 against B3 is the story comparison in this exercise; A2 against B2 is not. Document both ranges. A report that compares the wrong pages can be internally consistent while answering the wrong question.

Text extraction needs similar care. PyMuPDF's documentation warns that extracted PDF text may not follow normal reading order and may contain unexpected line breaks.[^2] Our whitespace normalization is adequate for the deliberately simple example after checking its pages. It is not a safe instruction to erase every structural distinction in a complex deck.

Columns, captions, and repeated headers deserve inspection beside the page. A line break can be layout noise, but a boundary can also distinguish a caption from a claim or one character's account from another's. Do not normalize away the context needed to interpret the result.

For scans, inspect the actual visible passage when extracted text is absent or questionable. An empty extraction is not evidence that the page contains no words. Any recognition or image-based comparison adds another operation to document; do not present an unperformed OCR step as a clean text comparison.

Turn differences into a reviewable record

The useful handoff combines the source pair, the tool's actual report where available, and a short classification of findings. Adobe documents saving the Compare Results document and reviewing changes side by side.[^1] Saving that report does not eliminate the need to name its settings or unresolved issues.

For the original Last Light fixture, the following is a local manual classification supported by the saved observations, not an Acrobat report:

Cover, A1 → B1: unchanged control. The local normalized-text and render checks agree. This gives a reference for an unchanged page in this pair. It does not verify the rest of the document by association.

Story, A2 → B3: relocation plus reflow. The full normalized text matches; the measure is narrower and the paragraph has three additional lines. Wording is accounted for in this fixture. The new page still needs whatever readability review the project requires; text equality does not approve the design.

Project context, A3 → B2: relocation plus substantive wording change. The access claim strengthens from discussion to confirmation. In real work, refer that statement to the person responsible for establishing access. Keep it marked as requiring confirmation rather than accepting it as a side effect of reformatting.

Visual approach, A4 → B4: presentation change. The heading changes weight and color, without a wording change. Whether that alteration is wanted is a design decision outside the detection result.

These categories can coexist. Forcing each page into “moved” or “changed” would erase the exact problem the example was made to teach. A page can move and contain a materially altered qualification; an unchanged passage can move and become harder to read.

For the native completion of this exercise, retain the Acrobat version and operating system, both file names or hashes, document model, old/new page ranges, text-only setting, displayed categories, and each control's actual outcome. Where a result is missing or mispaired, include the source passage and the additional step used to resolve it. Do not report an expected result as a completed control.

The final unresolved question is not always correspondence. Here the fictional files make correspondence easy, and the local checks account for the planned controls. The absence of an Acrobat run remains a technical gap. In a real version review, the authority for “confirmed” would be a separate editorial gap even after perfect detection.

Keep both visible. A useful change report makes it possible for the next person to find the difference, understand what changed, and know what still requires a decision. It does not make that decision for them.

Sources

[^1]: Adobe, “Compare two versions of a PDF file (Acrobat Pro)”, updated February 26, 2026; “Compare files,” “Compare files settings,” and “Review Compare Results.” Supports documented models, settings, filtering and report handling—not a claim that Acrobat was executed in this exercise. [^2]: PyMuPDF, “Text”, “How to Extract all Document Text” and “How to Extract Text in Natural Reading Order.” The local fixture was inspected using PyMuPDF 1.26.7; the online documentation is not a claim about the version of an unperformed Acrobat run.

Frequently asked questions

What four page-level differences does the Last Light exercise use as controls, and which one changes meaning?

The cover is unchanged; the story keeps the same normalized text but reflows from seven to ten lines; project context moves from A3 to B2 and changes 'being discussed' to 'confirmed'; the visual-approach heading changes from regular black to bold dark blue. The meaning change is the stronger access claim in project context.

Why does the article suggest Presentation Decks as a first Acrobat trial for the Last Light fixture, and when would continuous-text comparison be worth testing?

Presentation Decks is reasonable because the story and project-context sections remain separate pages and the paragraph wraps differently within its page rather than becoming continuous prose across pages. A continuous-text model is worth testing when a narrative expands across page boundaries, such as an old page's ending becoming the middle of a new page.

What is the difference between Adobe's Compare text only option and the report's Filter controls?

Compare text only excludes graphic differences from the comparison itself. Filter controls determine which difference categories are visible in the report, with annotations, formatting, and background differences off by default. One is what you asked the comparison to examine; the other is what the report currently shows.

How should a reviewer handle ambiguous page correspondence in a real pitch deck?

Read the project name, location description, surrounding sentences, and reliable source-page identifiers. If those do not settle the match, label the correspondence unresolved rather than merging paragraphs, inferring that both locations are confirmed, or picking the neatest counterpart. For scoped comparisons, document both ranges: in the exercise, A2 against B3 is the story comparison, while A2 against B2 is wrong.

What limits should a native Acrobat completion of this exercise record, and what gaps remain?

Retain the Acrobat version and operating system, file names or hashes, document model, old and new page ranges, text-only setting, displayed categories, and each control's actual outcome. Where a result is missing or mispaired, include the source passage and the extra step used. The absence of an Acrobat run is a technical gap; authority for 'confirmed' is a separate editorial gap even after perfect detection.

Further reading

• Two Script Versions Support Different Film Pitches. Which One Are You Writing? — When the comparison exposes a strengthened claim or conflicting versions, use this to decide which creative basis governs and hold unsupported claims for the responsible maker.
• Update Shared Treatment Copy Without Overwriting Local Layout Edits — Use this when deciding whether a difference is formatting or meaning, and before letting an update overwrite local wording that carries an approved decision.