0.51.0
- New: Plustek OpticFilm 8200i SE over USB — a native backend drives the scanner directly, no SANE, and the Scan tab is now enabled on Windows. Pre-scan with a crop box, two-pass IR aligned to the RGB scan, and Autofocus, Auto-exposure and bit depth shown only where the scanner reports them. The driver is its own project, pyopticfilm. @jboneng
- New: Render Exposure slider (HDR Slides) — a merged bracket's exposure is now continuous, 0 EV to −4 EV in the Process panel, because a bracket a stop apart rarely contains the exposure that looks right. The right-click menu still snaps to a captured frame; setting one clears the other. The merge is cached without the exposure applied, so dragging costs one multiply rather than a re-decode. @seanharding
- New: Unload from the canvas right-click menu — already on the toolbar overflow and the thumbnail menu, now on the image itself; same confirm, same saved edit kept. @thetalkingdrum
- Change: A control that cannot apply grays out with the reason — Narrowband and sensor unmix on a transparency (RGBScan.icc characterizes narrowband capture of negative dyes, and the unmix corrects for narrowband light, so neither means anything on a slide), and Normalize on a merged bracket (the stretch meters the frame and divides the merge's Render Exposure straight back out). They stay visible instead of disappearing while still live — the settings are sticky, and a hidden live setting could only be cleared by switching mode back. Unmerging restores Normalize. @seanharding
- Change: The SANE dependency group is now
sane—scannerno longer named what it installs, now that the Plustek and PIEUSB backends ship. The install hints name the group, which a plainuv syncnever installed. - Change: The in-app guide covers every panel — the tour picks up Alternative Processes and the Process panel's Calibration section, and the steps that had drifted from the controls they describe were corrected. The User Guide behind each panel's ⓘ was rewritten shorter along with it.
- Fix: Auto Crop takes the shape of the frame it found — Ratio now defaults to Free, so a detected 35mm, 645, 6x6, 6x7 or 6x17 keeps its own shape instead of a 6x6 cut to a 3:2 strip; picking a ratio still forces every frame to it. Each edge is read line by line and out to 10% of a side, so a holder mask, a clear rebate, a sprocket rail or a neighboring frame gets trimmed instead of left as a bright band, and a blown highlight beside the subject is no longer mistaken for the light bed. A frame whose picture runs to its own edge is still bounded at 3%. @seanharding
- Fix: Batch Autocrop judges each frame on its own evidence — trim, placement and deskew came from the roll consensus, which left some rolls uncropped and straightened the most crooked frames to the roll median. Tilt is not a roll property; residual tilt across five test rolls drops to a hundredth of a degree. @seanharding
- Fix: A bracket's ratios are solved on one white balance — the merge pins every frame to the primary's, but the solve that produces the ratios did not, so a camera left on auto white balance gave each frame its own multipliers and skewed the result. Visible only with Normalize on. @seanharding
- Fix: Slides stay positive in the film strip — batch thumbnails re-detected the film process from an 8-bit preview, and warm content (sunset, skin, autumn) tripped the orange-mask test, so some transparencies appeared inverted until clicked. The frame's saved mode now decides; detection runs only where nothing has. @seanharding
- Fix: Batch export finishes past a composite — a stitched or merged frame carries a
#stitch/#hdrsuffix, which the half-frame lookup read as a number and raised on. One composite in the visible set stopped the whole run, ordinary frames included. @seanharding - Fix: A damaged thumbnail no longer stops the app — a truncated cache entry (a kill or a full disk during the write is enough) passed the header check, then broke on the UI thread on that frame and on every launch after. It is now decoded before it is served, so a bad entry is a miss that regenerates. @seanharding
- Fix: ICC color management is back on Flat and Print exports — 16-bit TIFF/JXL exports shipped unmanaged on macOS arm64, silently, taking any custom Input/Output profile with them. @thetalkingdrum
- Fix: A crop handle can be grabbed wherever the cursor says it can — the hover cursor hit-tests by distance, but the press had to land inside the image, so a handle on the frame edge looked live and would not move. With a print border on, no handle of a full-frame crop could be grabbed at all. @seanharding
- Fix: Camera scanning reads shutter labels that carry their unit — a Nikon's own ladder was dropped, and NegPy no longer writes a shutter speed the body has not published. @light-sntchr
- Fix: Resource paths on Windows — generated install scripts carried backslash paths and resource lookups mixed separators. @MohammedAlkindi