Riferimento della riga di comando

Opzioni ed esempi per MergeRAW

Dove trovare lo strumento

Il programma di installazione di MergeRAW include lo strumento a riga di comando nella cartella mergeraw.lrplugin. Lo strumento non viene aggiunto al percorso di ricerca dei comandi della shell. Per usarlo non è necessario che Lightroom sia in esecuzione.

Installa il programma gratuito Adobe DNG Converter (download per macOS o Windows). MergeRAW lo usa per convertire in DNG i file RAW della fotocamera e per -demosaicIn. I file DNG e i TIFF supportati possono essere letti direttamente.

macOS — Intel e Apple Silicon

Il programma di installazione colloca mergeraw-mac64 nella cartella Modules di Lightroom a livello di sistema. Apri Terminale ed esegui:

alias mergeraw='/Library/Application Support/Adobe/Lightroom/Modules/mergeraw.lrplugin/mergeraw-mac64'
mergeraw -h

Windows — Intel e AMD (x64)

La posizione predefinita è la cartella mergeraw.lrplugin all’interno dell’installazione di Lightroom Classic. Apri PowerShell ed esegui:

Set-Alias mergeraw 'C:\Program Files\Adobe\Adobe Lightroom Classic\mergeraw.lrplugin\mergeraw-winX64.exe'
mergeraw -h

Su Windows ARM64, usa mergeraw-winA64.exe nella stessa cartella. Se durante l’installazione hai selezionato un’altra cartella di Lightroom Classic, modifica il percorso di conseguenza.

Questi alias valgono per la sessione corrente di Terminale o PowerShell e ti permettono di usare mergeraw esattamente come negli esempi qui sotto. Se hai installato il plug-in manualmente, usa l’eseguibile all’interno della tua cartella mergeraw.lrplugin.

Per un aiuto nella scelta della modalità di allineamento o del metodo di fusione, consulta Per iniziare.

Il riferimento delle opzioni qui sotto è in inglese, esattamente come lo stampa lo strumento.

MergeRAW Version 3.1.2, Written by Ryan Mack <mergeraw@mackman.net>

USAGE

  mergeraw [options] -out output.dng input1 input2 ...
  mergeraw [options] -out output.dng -files input_paths.txt

  Put options before input paths. Quote paths containing spaces. In option
  descriptions, <value> is required and [value] is optional.
  Inputs: camera RAW/DNG, or unsigned 8/16-bit RGB/grayscale TIFF. TIFFs load
  directly and produce a Linear DNG. Embedded RGB/gray matrix ICC profiles are
  honored; untagged RGB uses sRGB, gray uses gamma 2.2. No CMYK or alpha. Keep
  TIFF and camera RAW/DNG inputs in separate merges.

YOUR FIRST MERGE

  These recipes use the Lightroom plugin's alignment/blend choices and Auto
  demosaic. input_paths.txt lists one input path per line. For multiline
  commands, \ continues a line on macOS/Linux; on Windows, join the lines
  and remove \.

  Long exposure on a tripod (No Alignment + Normal):
    mergeraw -demosaicAuto -out long.dng -files input_paths.txt

  Handheld burst to reduce noise (Auto Align + Normal):
    mergeraw -demosaicAuto -alignRotate -out burst.dng -files input_paths.txt

  Remove people from a busy scene (Auto Align + Median):
    mergeraw -demosaicAuto -alignRotate -median -out empty.dng \
      -files input_paths.txt

  Low-light handheld burst (Night Align + Median):
    mergeraw -demosaicAuto -alignRotate -clip -sharpGate -quietWindow \
      -alignGate -clipFeather 4 -cropToCoverage -out night.dng \
      -files input_paths.txt
  Use matching exposures. Night Align's Median uses a clipped mean (-clip).

  More recipes, including focus stacks and star trails, follow the options.

FILES AND RUNTIME

  -out <file>                 Output DNG filename.
  -files <list.txt>           Read UTF-8 input paths, one per line; ignore
                              blank lines and # comments. Useful for large
                              stacks and the Windows command-line limit.
  -checkDarks                 Before merging, check embedded previews for
                              dark outliers. Exit 20 means review is needed;
                              retry without this flag to use the full set.
  -detectDarks                Read-only raw confirmation of supplied candidate
                              DNGs. Reports indexed dark/light/unknown results.
  -scratchDir <dir>           Existing folder for temporary frames and
                              converted DNGs. Default: output folder. Use a
                              fast local disk for large stacks.
  -dngconverter <path>        Path to the Adobe DNG Converter executable
                              (inside the .app bundle on macOS).
  -threads <n>                Worker count. Default: performance-core count on
                              Apple Silicon; otherwise half the hardware
                              threads.
  -h, -help                   Show this help.
  -v, -version                Show version information.

BLEND AND HIGHLIGHTS

  The default blend averages the inputs, matching their exposures. No
  alignment runs unless requested. Input indices start at 0.

  -deghost                    Average still areas; preserve moving subjects as
                              they appear in the target frame.
  -median                     Take each pixel's median. Removes objects absent
                              from most frames; supports alignment.
  -medianInMemory             Keep median frames in RAM instead of scratch
                              files. Cannot be combined with alignment.
  -maximum                    Match exposures, then keep each pixel's
                              brightest value (Lighten).
  -minimum                    Match exposures, then keep each pixel's darkest
                              value (Darken).
  -range <mode>               Highlight source selection: auto (default) for
                              HDR means; on/shared forces shared-color weights;
                              perchannel selects colors independently (can
                              fringe at motion); off/full blends all eligible
                              samples. If every
                              source clips, average the least-exposed group.
                              default is an alias for auto.
  -rangeThreshold <value>     Source saturation threshold, 0 < value <= 1
                              (default 0.99). Applied before exposure gain.
  -noBracketRescale           Disable automatic highlight headroom for
                              bracketed mean merges. Recovered values above
                              the target's white point clip instead of being
                              rescaled into the output.
  -target [index]             Set the exposure, alignment and deghost
                              reference. With no index, use 0. Normally picks
                              the metered bracket frame or the first
                              darkest-EV frame; sky/night modes have their own
                              automatic choices.
  -metaFrom <index>           Copy capture identity and crop settings from
                              this input, independently of the target.
                              Defaults to the first input with sky tracking.
                              With -sharpGate and no -target, prefer this
                              frame as the target unless it is blurry.
  -sourceCrop <on> <T> <L> <B> <R> <angle> <warp> <unit> <process>
                              Override the metadata pick's catalog crop.
                              Bounds are raw-oriented normalized coordinates;
                              on, warp and unit are 0/1; process is the Adobe
                              ProcessVersion. Pass unchanged to editor finish.
  -updateEXIF                 Update output ISO and exposure time for the
                              combined exposure (alias: -exif).
  -exposureTilt <stops>       Leave out a frame whose exposure, matched
                              against the target, still tilts across the
                              image by more than this (a flash caught by the
                              shutter, one light misfiring). Default: focus
                              stacks leave out a frame whose tilt is an
                              outlier against the rest of the stack, other
                              modes keep every frame; 0 keeps every frame in
                              every mode. Frames shot at another exposure
                              setting and Sky modes are never affected.

ALIGNMENT AND PIXEL SHIFT

  -align                      Align by translation.
  -alignRotate                Align with sub-pixel shifts and rotation
                              (Lightroom Auto Align).
  -alignBicubic               Use sharper Catmull-Rom resampling instead of
                              bilinear. Slower; may overshoot at strong edges.
                              Experimental.
  -noFocusBias                With -alignRotate, disable the default
                              preference for aligning the in-focus parts of
                              each frame.
  -updateAlignmentTarget      Align each new frame to the accumulated result
                              instead of a fixed reference.
  -pixelshift                 Assemble Nikon 4/8/16/32-shot or Fujifilm
                              X-Trans 20-shot sequences. The first 5 X-Trans
                              frames give full color at native resolution.
  -noPsGridCheck              Skip the pixel-shift grid check, which normally
                              warns if camera movement has displaced frames
                              from the commanded sensor positions.
  -psKeepOrder                Merge the shots in the order given instead of
                              sorting them by file name.
  -psShots <n>                With -pixelshift -focusStack: shots per focus
                              plane (4, 8, 16, 32; 5 or 20 for X-Trans).
                              Default: detect from lens position or focus
                              distance, else from the frame count.

HANDHELD NIGHT BURSTS

  Use matching exposures and -align or -alignRotate. These controls select
  frames and reject motion before averaging.

  -clip                       Reject per-pixel outliers with a sigma-clipped
                              mean. Keeps moving objects out of the
                              background; uses temporary frame storage.
  -clipSigma <k>              Clipping threshold in noise sigmas for -clip and
                              sky stacks. Default: 2.5.
  -clipDeghost [s]            With -clip, preserve motion from the target
                              instead of the stack majority. Compare all color
                              planes with the target's local range plus
                              sensor-noise tolerance. Default: 4 sigmas.
                              Experimental.
  -clipFeather <px>           With -clip/-clipDeghost, soften and slightly
                              expand rejected regions for smoother
                              transitions. Replays stored frames; re-reads
                              inputs if scratch space is tight. Experimental.
  -sharpGate [f]              Drop frames below f times the best sharpness
                              score. Default: 0.75. Uses temporary storage;
                              automatically replaces a blurry target unless
                              -target was given.
  -noAutoTarget               Disable -sharpGate's automatic target selection;
                              keep the first frame unless -target is given.
  -quietWindow                With -sharpGate, keep the stillest stretch of
                              the burst and use its middle frame as target. No
                              effect if no suitable stretch exists.
  -alignGate                  With -alignRotate, drop frames whose feature
                              alignment failed. Uses temporary frame storage.
  -cropToCoverage             With -clip/-sharpGate/-alignGate or sky
                              tracking, record a reversible crop to the
                              covered area. No pixels are discarded; Reset
                              Crop restores them in Adobe apps.

FOCUS STACKING

  Shoot matching exposures. Automatic brackets keep capture order; manual
  stacks use comparable lens position or distance, then capture time.
  Without usable metadata, input order is preserved. The default alignment
  target is the first resolved frame. Output is a Linear DNG.

  -focusStack [depth|hybrid]  Merge each region from the sharpest frames
                              (alias: -focus). Default: depth. hybrid recovers
                              extra fine detail with more time and memory.
                              Aligns neighboring frames to handle focus
                              breathing.
  -focusScaleOnly             Correct focus breathing by centered scaling only
                              (Lightroom Tripod). No shifts or rotation.
  -focusNoAlign               Skip alignment for already registered inputs.
  -focusRadius <px>           Sharpness window radius. Default: 24. Smaller
                              favors detail but adds noise/halos; larger
                              suppresses halos.
  -focusFeather <n>           Blend across this many neighboring focus slices.
                              Default: 1. Use 1.5 for the former wider blend.
  -focusWeakPeaks             Recover weak depth peaks using coarser evidence.
                              Default: on; -noFocusWeakPeaks disables it.
  -focusDetailConfidence      Reduce depth confidence at broad edges. Default:
                              on; -noFocusDetailConfidence disables it.
  -focusTrustedBlend          Select confident source pixels with a 2-pixel
                              feather. Experimental.
  -focusLocalExposure / -noFocusLocalExposure
                            Smooth local exposure flicker (default on).
  -focusDepthSession <dir>   Prepare an interactive focus depth session.
  -focusResume <dir>         Finish from its cached aligned frames and state.
  -focusDepthMapSave <file>   Save a 16-bit gray depth TIFF to this path.
                              Black = first resolved frame; white = last.
  -focusDetailMaskLoad <file> With hybrid, retain extra detail through an
                              8/16-bit sensor-grid mask: white retains,
                              black uses depth only.
  -focusDepthMapLoad <file>   Blend using an edited 8/16-bit gray depth TIFF.
                              Keep its full image size and sensor orientation,
                              and use the same inputs and resolved focus order.
                              -focusFeather still sets the blend width;
                              hybrid still adds its automatic fine detail.

SKY TRACKING

  For untracked night-sky sequences with raw mosaic inputs. Star-aligned
  output is a Linear DNG; one-frame outliers are clipped by default.

  -alignSky                   Align on the moving stars, normally using the
                              middle frame as target. Unmatched frames are
                              dropped unless the global model can recover
                              them.
  -alignSkyFg                 Track stars while keeping a separate, static
                              foreground stack. Blend the two through an
                              automatic sky mask based on motion.
  -alignSkyAI                 Like -alignSkyFg, with a downloaded DINOv3 model
                              helping to find the skyline. See -modelDir.
  -modelDir <dir>             Folder containing the downloaded model files.
                              Default: the Lightroom plugin's per-user model
                              folder. Requires -alignSkyAI.
  -noSkyGlobal                Disable the global sky-motion model that uses
                              capture times and star matches to recover
                              cloud-hidden frames.
  -noClip                     Keep one-frame sky outliers such as meteors and
                              satellites. Otherwise -clipSigma sets the
                              threshold.
  -skyMaskSave <file>         With -alignSkyFg/-alignSkyAI, save a 16-bit gray
                              sky mask to this path, in rendered orientation
                              and crop. White = sky; black = foreground.
  -skyMaskLoad <file>         Use an edited sky-mask TIFF with -alignSkyFg or
                              -alignSkyAI. Overrides automatic mask creation;
                              the AI model is not needed for a loaded mask.
  -skyMaskSession <dir>       With -alignSkyFg/-alignSkyAI, build the mask,
                              write the mask editor's session (session.json,
                              mask.tif, half-size previews of both stacks)
                              to this folder and stop without an output.
                              Finish with -skyMaskLoad <dir>/mask-edited.tif.
  -alphaMask <file>           Attach a sky-mask TIFF as output transparency:
                              foreground opaque, sky transparent. Uses the
                              -skyMaskSave geometry for Photoshop layering.
  -skyLit                     With masked sky modes, preserve frames that
                              light the foreground instead of averaging their
                              light away. Default: on.
  -noSkyLit                   Average every frame into the foreground.
  -skyLitTile <px>            Foreground lighting tile size. Default: 16.
  -skyLitMinRegion <0-1>      Smallest lit region to keep, as a fraction of
                              the image. Default: 0.01.
  -skyLitBlend <0-1>          Share of the mean retained in lit regions.
                              Default: 0.
  -skySweep                   With masked sky modes, remove a trailed star's
                              frames from the foreground along its whole
                              sweep, not only where each pixel proves it.
                              Skipped when light painting was found or the
                              foreground moved. Default: on.
  -noSkySweep                 The per-pixel star drop only.

STAR TRAILS AND LIGHTEN BLENDS

  -trailJoin                  With -maximum, recover light at the joins
                              between consecutive trail segments. Requires
                              star-shaped features moving with the sky.
  -trailJoinAuto              Enable trail joining when capture timing and
                              measured sky motion qualify the frame pair.
                              Otherwise keep the plain maximum.
  -trailChroma                Repair false color along thin star trails while
                              preserving pixel brightness. Automatic with
                              trail joining; -noTrailChroma disables it.
                              Experimental.
  -lightenFloor               Remove the noise pedestal from a maximum blend,
                              using the frame mean as a floor. Uses sensor
                              noise when supported; the fallback can mistake
                              long trails for noise.
  -noLightenFloor             Keep the plain maximum without noise-floor
                              removal. Default for ordinary -maximum merges.
  -skyTrails                  With -alignSkyFg/-alignSkyAI, use an unaligned
                              maximum for the sky and the static stack for the
                              foreground. Supports trail joining, color repair
                              and noise-floor removal.
  -noSkyTrails                Keep the star-aligned sky instead of trails.
                              Default.

DARK AND FLAT FRAMES

  -dark <files...>            Treat all remaining inputs as dark frames.
                              Average them, detect hot pixels, and subtract
                              only their excess from each light frame. Also
                              accepts -dark -files darkframe_paths.txt.
  -darkThreshold <k>          Hot-pixel threshold in sigmas of the averaged
                              dark frame's noise. Default: 5.
  -flatField <file>           Correct vignetting and broad color shading with
                              one flat-field image of an evenly lit blank
                              surface. Match the camera, image dimensions,
                              lens and aperture. Average multiple flats first
                              (see examples). Not supported with X-Trans pixel
                              shift.

DEFECT REPAIR

  -cosmicray                  Repair transient hot pixels, cosmic-ray strikes
                              and readout defects in each light frame.
                              Protects stars and trails; works well with
                              -alignSky.
  -cosmicrayScene <mode>      auto (default): classify the scene and skip
                              short exposures; astro: protect stars; day:
                              protect sharp static detail in long daylight
                              exposures.
  -cosmicrayThreshold <k>     Transient threshold in frame-noise sigmas.
                              Default: 8.
  -burstSeconds <s>           Total shutter time of the capture these frames
                              belong to, for the automatic policy's 1 s
                              threshold (a plane of a larger burst).
  -hotmap                     Build a persistent hot-pixel map from 8+
                              matching-exposure light frames with scene
                              motion. Static scenes risk mapping real detail.
                              Ignored when -dark is supplied.
  -irpdaf                     Repair IR PDAF row banding in every input. Can
                              also filter a single DNG.
  -hotspot                    Detect and subtract central glare in IR images.
                              Conservative; does nothing at low confidence.
  -hotspotTrim <0-2>          Scale the hotspot correction. Default: 1.
  -hotspotForce               Apply hotspot correction even at low confidence.
  -skipUnrepaired             With -irpdaf/-hotspot, write no output if no
                              repair was needed.

DNG OUTPUT AND COMPRESSION

  -demosaicAuto               Keep the mosaic when possible; otherwise choose
                              a demosaic for the sensor. Uses Adobe for
                              X-Trans when available. Lightroom's Auto mode.
  -demosaic                   Write a full-color Linear DNG. Alignment modes
                              already demosaic as needed; this explicitly
                              requests linear output.
  -demosaicXTrans             Use the native Fujifilm X-Trans demosaic and
                              write a Linear DNG (Lightroom Sharper). Cannot
                              combine with -demosaicAuto or sky tracking.
  -xtransMoire               Opt in to localized two-color moire repair with
                              directional anti-aliasing for native X-Trans
                              Sharper. -noXTransMoire disables it (default).
                              Does not affect native Neutral.
  -remosaic                   Demosaic, then re-encode as a Bayer mosaic for a
                              smaller output (Lightroom's Force Mosaic
                              Output; Adobe Denoise renders linear DNGs of
                              night skies purple). Not supported for
                              grayscale output; X-Trans re-encodes as RGGB.
  -demosaicMethod <1-7>       Choose the demosaic method when one is needed. 2
                              uses Adobe DNG Converter (Lightroom Safe); 7 is
                              the native default. May combine with
                              -demosaicAuto.
  -float                      Write 16-bit floating-point DNG samples. Pair
                              with -demosaic for ACR/Lightroom shadow detail.
  -split <new-dir>             Extract one multi-frame Pentax DNG or Fuji RAF
                              into a new folder, without merging/conversion.
  -rmprof                     Remove lens distortion and other DNG opcodes.
  -lossy                      Enable JPEG XL compression. -jxl_distance sets
                              loss; 0 makes it lossless. Mosaic JXL requires
                              -dngVersion latest, otherwise it uses lossless
                              JPEG.
  -dngVersion <ver>           Output compatibility: latest (default); 1.7.0 =
                              no mosaic JXL (DXO PhotoLab); 1.4.0 = no JPEG XL
                              at all.
  -jxl_effort <1-10>          JPEG XL effort. Default: 7. Higher compresses
                              better but takes longer.
  -jxl_distance <0-25>        JPEG XL loss. Default: 0.01. Higher loses more
                              detail; 0 is lossless. Used with -lossy.

EXTERNAL RAW CONTENTS EDITING

  -export <file>              Export a demosaiced TIFF for editing.
  -import <file>              Import an externally edited TIFF back into a
                              DNG.
  -edit <app>                 Launch an editor such as Photoshop.
  -editContrast <-100..100>   Lightroom contrast approximation for editing
                              and editor previews. Default: 0.
  -wb <temp> <tint>           White balance for export/import: temperature in
                              Kelvin and tint offset, e.g. 5500 +10. Default:
                              camera white balance.
  -exposure <stops>           Exposure compensation for editing, e.g. +1.0 or
                              -0.5. Adjusts the gamma curve's contrast
                              midpoint (alias: -exp).
  -orientation <1-8>          EXIF orientation: 1 = normal, 3 = 180 degrees, 6
                              = 90 clockwise, 8 = 90 counterclockwise.

DIAGNOSTICS

  -perf                       Print phase timings and a summary.
  -timing                     Print worker-thread timing information.
  -validate                   Run DNG SDK validation on the inputs and written
                              output; report what another reader would see.
  -alignVerbose <0-9>         Alignment diagnostics. Higher levels add
                              feature, fit and per-track detail.
  -saveMask <prefix>          Save matching debug masks as TIFFs. Use "*" for
                              all masks; quote it to prevent shell expansion.
  -testMask <prefix>          Compare matching masks with saved TIFF
                              references. Use "*" for all masks.
  -testMaxDiffPct <n>         Allow n% of pixels to differ in mask
                              comparisons. Default: 0.
  -testMaxDiff1PctPixels <n>  Allow n% of pixels to differ by more than 1% in
                              mask comparisons. Default: 0.
  -irpdafsim                  Inject synthetic PDAF banding to check -irpdaf.
  -hotspotsim                 Inject a synthetic IR hotspot to check -hotspot.

MORE EXAMPLES

  Recipes use the plugin's alignment/blend flags and Auto demosaic. Add
  -cosmicray for automatic defect repair; with 8+ light frames, add -hotmap
  too. See output options for compression and -updateEXIF for
  combined-exposure metadata.

  Keep a moving subject (Auto Align + Deghost):
    mergeraw -demosaicAuto -alignRotate -deghost -target 0 -out subject.dng \
      -files input_paths.txt

  Tripod focus bracket (No Alignment + Focus Stacking):
    mergeraw -demosaicAuto -focusStack depth -focusScaleOnly \
      -out focus.dng -files input_paths.txt
  Omit -focusScaleOnly for Auto Align; use hybrid to retain extra detail.

  Star trails or light painting (No Alignment + Lighten):
    mergeraw -demosaicAuto -maximum -trailJoinAuto -out trails.dng \
      -files input_paths.txt

  Fixed stars with a sharp landscape (Align Stars + Basic Mask):
    mergeraw -demosaicAuto -alignSkyFg -out sky.dng -files input_paths.txt
  Use -alignSkyAI for AI Mask; its downloaded model must be installed.

  Trails with a cleaner foreground (Align Stars + AI Mask Star Trails):
    mergeraw -remosaic -alignSkyAI -skyTrails -trailJoinAuto -lightenFloor \
      -out skytrails.dng -files input_paths.txt
  Requires the downloaded AI model. -modelDir can select its folder.

  Merge rendered TIFFs directly:
    mergeraw -out merged.dng frame1.tif frame2.tif

  Apply dark frames while averaging a tripod stack:
    mergeraw -demosaicAuto -out clean.dng -files input_paths.txt -dark \
      -files darkframe_paths.txt

  Correct one photo with a flat field:
    mergeraw -flatField flat.dng -out corrected.dng photo.dng

  Average several flats, then correct a tripod stack:
    mergeraw -out flat.dng -files flatfield_paths.txt
    mergeraw -demosaicAuto -flatField flat.dng -out corrected.dng \
      -files input_paths.txt
  flatfield_paths.txt lists matching flats, one path per line. Average them
  without alignment; -flatField accepts one flat image, not a list of flats.

  Save a sky mask, edit it, then rerun with the edited mask:
    mergeraw -alignSkyAI -skyMaskSave sky-mask.tif -out sky.dng \
      -files input_paths.txt
    mergeraw -alignSkyAI -skyMaskLoad sky-mask.tif -out sky-edited.dng \
      -files input_paths.txt
  Edit sky-mask.tif between runs. The load run uses your mask instead of AI.

  Save a focus depth map, edit it, then rebuild the same stack:
    mergeraw -focusStack -focusDepthMapSave depth-map.tif -out focus.dng \
      -files input_paths.txt
    mergeraw -focusStack -focusDepthMapLoad depth-map.tif \
      -out focus-edited.dng -files input_paths.txt
  Edit depth-map.tif between runs; keep the same image size and frame order.
  Map export is opt-in. Lightroom does not automatically save either map.

  Guide: https://mackman.net/mergeraw/instructions.html