Matched by number, not by guess.
Every pair is the same sheet across two issues. Sheets missing from one side are reported as such.
Know what changed before the field finds out.
Specset compares two plan sets, or two revisions of one sheet, and returns every change matched by sheet, classified by what it touches, and shown beside the drawing it came from. Clouded or not.
Plan set comparison90% CDs → 100% CDs
31 changes · 6 sheetsSheetChanges
S2.2Framing plan · Level 2+4~8
A-101Floor plan · Level 1+2~5
A-601Door schedule~3−2
M-201Mechanical plan · Level 1+1~3
E-301Lighting plan · Level 1~2
A-103Enlarged plansRevised set only
A-101_3ModifiedDimensionCorridor 1143 RFI-1287✓ Clouded
OriginalRev 2 · Mar 12
6′-0″RevisedRev 3 · Apr 18
6′-8″3Corridor 114 clear width increased from 6′-0″ to 6′-8″ at the north wall; door 114A shifts 1′-4″ east to hold the clearance.
Matched by sheet number · title block excludedv1 · v2 open the sheets, clouds drawn on v2
Every change gets a type.
Added, removed, or modified. Dimension, note, callout, schedule, legend, symbol, or detail. Revision and callout read alongside.
Your team keeps the last word.
Open the crops, remove what is not real, re-run a sheet. The log reflects your decisions, not the first pass.
Trusted by the teams delivering public work
“The 100% set moved a footing and nobody clouded it. Specset had it in the change log before we opened the transmittal.”
90K+Drawing sheets read into project records
475+Projects in production, bid through closeout
01 · How a comparison runs
Match, detect, classify.
A revised set is not a list of changes. It is a hundred sheets where a handful of things moved and only some of them got a cloud. The comparison does the sheet-by-sheet reading so nobody has to.
01 · Match
Original setA-101A-102·S2.2M-201
Revised setA-101A-102A-103S2.2M-201
Sheets pair by number across the two sets. When several original documents carry the same sheet, the highest-precedence one is used. A sheet that exists in only one set is reported, not guessed.
02 · Detect
Both renderings are compared pixel for pixel. Differing areas cluster into regions, and the title block and alignment noise are filtered before anything is read.
03 · Classify
~Dimension 6′-0″ → 6′-8″, Corridor 114△3
+Callout Detail 5/A-501 referenced at the vestibule△3
−Schedule Door 118 removed from the door schedule
AI vision reads each region as a drawing: what changed, whether it was added, removed, or modified, and which kind of content it touches. Revision triangles and their RFI callouts are read alongside.
02 · Two ways to ask
Compare sets, or compare revisions.
Point it at two uploaded sets, or ask about a sheet the project already tracks. Either way the comparison runs in the background and the thread stays open for questions.
Set against set
Two issues of the drawings.
Compare the 100% CD set against the 90% set.
Comparing Ridgeview 90% CDs (original) to Ridgeview 100% CDs (revised).Matched 142 sheets · 6 with changes · A-103 in the revised set only.31 changes: S2.2 12 · A-101 7 · A-601 5 · M-201 4 · E-301 3.
No original chosen? The revised set is scanned against every document in the project, and the highest-precedence copy of each sheet is used.
Revision against revision
One sheet, its own history.
What changed in the latest revision of A-101?
Comparing published revisions of A-101: rev 2 → rev 3.7 changes: 5 modified, 2 added. Two are not clouded.
Show me only the dimension changes.
3 dimension changes on A-101: corridor 114 clear width, stair 2 landing depth, vestibule 101 door offset.
Sheets carry their published revisions, so no upload selection is needed. Ask for specific revisions, a list of sheets, or every sheet with more than one revision.
03 · What comes back
A change log you can send.
One row per change, grouped by sheet. Open the crops, export the spreadsheet, or open the revised sheet with a cloud drawn on everything that moved, including what the architect missed.
A-101 · Floor plan · Level 1Rev 2 → Rev 3 · 7 changes
v1v2 · cloudsFilter⇩ Export .xlsx
DetailItemTypeChangeCloudedRevCallout
Corridor 114Clear widthDimension~ 6′-0″ → 6′-8″✓△3RFI-1287
Vestibule 101Detail referenceCallout+ 5/A-501 added✓△3
Stair 2Landing depthDimension~ 4′-0″ → 4′-6″not clouded
Room 118Door 118Symbol− Door and swing removed✓△3PR-002
General notesNote 7General note~ Fire-stopping reference updated✓△3
Expand a row for the side-by-side cropFalse positives removed by ID stay removed
Each row records whether a revision cloud encircles the change. The ones without a cloud are the ones to price.
The spreadsheet carries the side-by-side crops for every row, so the log stands on its own in an RFI or a change order.
Filter by sheet, change kind, or content type in conversation. Remove a false positive by ID and it stays removed.
Common questions
What teams ask about plan comparison.
01What can it compare?
Two uploaded drawing sets, such as the 90% and 100% CDs or the bid set against the conformed set, or the published revisions of a sheet already in the project. Ask what changed in the latest revision of A-101 and it compares that sheet against its own previous revision, no document selection needed.
02How are sheets matched?
By sheet number. A-101 in the original pairs with A-101 in the revised set. If several original documents contain the same sheet, the highest-precedence document is used. Sheets that exist in only one set are listed so nothing is silently skipped, and duplicate numbers across building areas can be disambiguated by area.
03What counts as a change?
Both sheets are rendered and compared pixel for pixel, then each differing region is read by AI vision and described: added, removed, or modified, and whether it is a dimension, general note, callout, schedule, legend, symbol, geometry, or detail. Title block changes and registration noise are filtered out before anything is reported.
04Does it read revision clouds and triangles?
Yes. Each change records the revision triangle beside it, the callout text such as an RFI or PR number, and whether a revision cloud actually encircles it. Changes the architect did not cloud stand out in the log, which is usually the point of running the comparison.
05What if a detected change is wrong?
Remove it. Every change has an ID, and removing false positives from the comparison keeps them out of the log and the export. You can also re-run specific sheets; only their results update, and the rest of the comparison is preserved.
06How do we get the results out?
The comparison runs in the background and per-sheet counts appear as each sheet finishes. The log filters by sheet, change kind, and content type, exports to a spreadsheet with the side-by-side crops included, and the revised sheet opens with a cloud drawn on every change. You can also just ask the thread: which dimension changes are on S2.2?
Start with two issues of one set
Send the 90% and the 100%.
We'll return the change log with every sheet matched, every change classified, and the ones nobody clouded at the top.