Registration
Auto-Detect reports a low match, or “no match”. Check that the source really is the image this print came from, and that the scan is roughly upright. Then place the quad by hand: drag each corner to the corresponding corner of the printed image area. A hand-placed quad is fully supported — the detector is a convenience, not a requirement, and the PDF report records that the corners were placed by hand.
Aspect is red / the status bar warns the aspect is off. One corner is in the wrong place. Compare in Both view and drag it.
Overlay shows ghosting even though the corners look right. The print itself may be distorted — paper expansion during wet processing is real and not always uniform. Get the corners as close as you can; the flat-region masking exists precisely so a small residual error does not corrupt the measurement.
The quad keeps moving when I load something. Only until you touch it. Dragging a corner marks the quad as yours, and later automatic detections leave it alone.
Coverage and confidence
Coverage is low and a banner says so. The image simply did not contain those tones. The correction is still valid for the tones it measured, and unchanged elsewhere. If you need a full-range correction, use an image with a full-range subject — or print a step wedge and use StepWedgeDensitone.
Lots of rows have low confidence. Try fewer Bins (32 spreads the same samples over fewer buckets), or a larger Window. If the mask is nearly empty, the image is too detailed to have flat regions: raise Edge excl. and lower Flat radius, and check the Mask view to see what survived.
White or black point says “estimated” in red. There was no highlight or no shadow band to read. Include the paper margin in the scan, and if the print has a coating border let it into the frame — those are the anchors the app prefers.
“black point from coating border”. The exposed coating got full UV with no film in the way, so it is darker than anything the process can print through a negative. The black anchor is therefore slightly too deep and measured tones normalize a little high. Check your deep shadows against the print by eye before trusting them.
“scan highlights at top of range — likely clipped”. The scanner or camera exposure clipped the paper white, which flattens the measured highlight response. Reduce the repro exposure, turn off auto-levels, and rescan.
The correction
The corrected print is worse.
The usual cause is stacking: the exported .acv replaces the curve you
loaded as Active, it does not sit on top of it. The second most common
cause is not loading the active curve at all, which makes the app correct
from linear.
Shadows in the negative all collapse to one value. Turn on Aim at print range. The literal target aims at tones your process cannot reach, and everything past that point maps to the same value.
The correction looks too strong / too flat.
Check the Aim slider — it is a look, and it travels with the export.
linear is the neutral setting.
Export is greyed out. There is no result yet: load both images, and let the analysis finish (the status pill will read Curve ready).
Proofing
Proof is greyed out. The profile is built during analysis. Load both images and wait for the analysis to complete.
Photoshop shows a stale profile / Simulate Paper Color is greyed out. Photoshop caches profiles by filename. If you exported over a profile you had already used, quit and reopen Photoshop.
The proofed source looks warm — my paper isn’t that colour. The warm cast is the D50 media white, not a measurement. The profile is neutral by design: it encodes tone and black lift, not paper tint.
Files and sessions
A session opened but an image did not. The file was moved or its volume is offline. Re-open it with Source… or Scan…; everything else in the session stays as restored.
⌘S asked me where to save even though I saved before.
That is the first save after a launch. What was restored is a working state,
not the document you named, so the app will not overwrite your last saved
.apdt on the strength of it.
Export filenames are full of hyphens. They carry your method and paper, with characters that are illegal in filenames folded to hyphens — “Platinum/Palladium” would otherwise be a path separator. Report contents keep your text verbatim.
Getting help
Bugs and feature requests:
github.com/yutakas/AltProcessDensitone-releases/issues.
An .apdt session plus the source and scan makes a report far easier to act
on.