OFFLOAD · VERIFY · REPORTS
ENGRAM PRO
Offload, verify and report for the DIT cart.
DEVIN DEV · NOT ON LIVE YET
Built and in testing in the dev build. Not in the live app yet; it can still change.
1.3.1CartDATE08 NOV 2026CHANGES47IN DEV · NOT ON LIVE YET
The viewer runs on the GPU with Metal (same picture, a third of the CPU), the transcode decodes on the GPU, drive tiles wait for a real answer.
IMPROVED 02
A viewer clip that is not 4:2:2 10-bit Rec.709, or is played at ½ or ¼, keeps the old way.
ProRes decodes on the GPU (VideoToolbox) now: within 1 code value of the CPU decoder on every byte, about 30× less CPU. Report frame grabs decode on the GPU for the same proven formats (not ProRes) and use 2 cores, never all of them
FIXED 45
A drive that had not answered yet showed READY (a locked drive could read READY, be ticked and count in RUN ON N)
Now: it shows CHECKING, can't be ticked, and the first check lands before the window opens on the card. Drive checks have their own queue
WILL-IT-FIT still scanned the volumes on the main thread
Now: it reads the same background drive check
COPIES showed old or empty counts when opened right after an offload
Now: it redraws when the real counts land
Opening a DIFFERENT big bin (over 400 clips) flashed scan order, then jumped
Now: the list waits (SORT shows an hourglass) and the viewer opens the sorted first clip
During an offload the viewer could stay on the wrong first clip
Now: it moves to the sorted first clip when the newest sort lands
UNDO on Duplicate Structure that was refused (HAS CLIPS — KEPT) could still remove its drive folders
Now: only a real UNDO removes them. The drive checks left the main thread
RETRY of a partial offload whose file pick was gone copied the whole card
Now: it refuses and copies nothing. Old pick files (over 14 days) are cleaned at launch
The scrub cache could count a folder twice and grow past its limit
Now: each folder counts only what it added
HEIF stills used their 320 px built-in preview and looked soft
Now: they decode in full. A RAW whose built-in preview is too small for the cell decodes in full too
A scrub strip that froze at the end of a clip was kept
Now: a frozen tail (5 s or a quarter of the clip, the same picture) is rebuilt with every frame
Scopes copied a 4K frame twice on the main thread
Now: once, straight from the frame's own buffer. Measured each run (one copy vs the old redraw)
A scrub or toggle repaint waited behind the playing frame and a LUT change
Now: the repaint has its own GPU lane and the LUT is baked outside it
GPU buffers piled up per size and were kept after playback
Now: freed on STOP and when macOS asks for memory, and capped in total
A transcode could say "decode GPU" when VideoToolbox could not start and ffmpeg fell back
Now: the note is corrected to CPU when that happens
Every progress tick redrew every screen
Now: progress redraws at most 5 times a second, and only the progress views
Boxed controls with a fixed 22/24 pt height
Now: all use the one box height and scale with the UI
Playing a 4K clip with a look and zebra on took 1.4–1.6 CPU cores
Now: 0.5 of one core, same frame rate (FX6 137 % → 49 %, ProRes 157 % → 47 %)
Every viewer step (rotate/crop, re-squeeze, look, false colour, zebra, peaking) read the picture back to the CPU and sent it to the GPU again
Now: they run one after another on the GPU and the picture comes back once. Same picture, every byte
Scopes redrew every frame through the CPU on the main thread before they drew
Now: a straight copy instead. Same scope, every byte
Transcode from an MXF card decoded on the CPU with no core limit
Now: the GPU decodes it (FX6 XAVC-I), the rest uses all performance cores. 9.2 → 5.0 CPU-s per 4 s of 4K. The delivered file is the same picture, every byte. The transcode note says which decoder ran
The library file could lose a probe after a crash mid-write (the next probe got glued to the half line)
Now: a torn line is dropped on open and the next probe starts on its own line
A probe whose write failed (disk full, drive gone) could vanish from the library
Now: the library only leaves a probe out when it is really in the probe file
Opening a project after an older ENGRAM saved it lost every probe
Now: they are filled back from the probe file, and the old file is backed up before ENGRAM rewrites it
A big bin flashed in raw order before the sort landed, and the viewer could open the wrong first clip
Now: old clips keep their place, new ones follow, and the viewer moves to the sorted first clip
Every batch of landed clips re-read every clip in the bin to sort it
Now: only the new clips are read
Clips recorded the same day sorted by file time
Now: they sort by their timecode. Backup time on old projects no longer moves when a file's permissions change
At the largest UI size the TASKS filter row ran past the window edge at the smallest window
Now: ISSUES and CLEARED show their icon and count there (the word is in the hover)
A drive turned red because of an unticked folder on it, and FITS N MORE counted a red × folder
Now: a drive is judged on its ticked folders only, and only usable ones share its room
Redrawing the offload window checked every drive on the main thread
Now: drive health is read in the background every 3 s while the window is open; START still checks fresh
Switching projects lost a drive's custom route, and switching during a copy saved the pool under the wrong project
Now: each drive keeps its route per project, and a switch during a copy waits for the copy to end
A card you had browsed and then pulled out could still be taken as the source
Now: only a card that is still plugged in counts
UNDO right after Duplicate Structure could leave empty folders on the drives
Now: they are removed when they get made. CMD+Z (Edit › Undo) now undoes Duplicate Structure too
A folder on the card's own drive said "This is the card itself"
Now: it says "Same drive as the card. A backup on the card's own drive is not a backup."
RETRY of a partial offload could copy the whole card if the file pick had been cleaned up
Now: the pick is kept in ENGRAM's own folder, and RETRY refuses with "The first run's file pick is gone. Stage the card again." if it is missing
A queued card with no bin could show the selected bin in its BIN PATH
Now: it shows OFFLOADS
A removed drive folder inside another listed folder was treated as the same drive
Now: it stays its own drive, so a missing drive still warns
Long-GOP H.264 scrub strips could go blank after the last keyframe
Now: every I-frame is kept, and a strip that comes out blank is rebuilt in full before it is kept
The meter said GPU 0% when it could not read the GPU
Now: it shows a dash. The resource panel adds ENGRAM's own GPU share next to the whole-Mac number (helpers can't be read one by one on macOS)
Reports used every core even while a transcode ran
Now: 4 workers beside other work, else up to 8
RAW photo thumbnails in the grid were fully decoded
Now: they use the camera's own preview; the viewer still shows the full RAW
The scrub cache janitor re-measured the whole cache after every clip
Now: it keeps a running total
Every library change redrew the whole app mid-offload
Now: only selection changes reach TRANSCODE; dragging over the same row redraws nothing
Changing the Sony look in Settings only reached the next clip
Now: the open clip changes at once
The update window kept old UPDATE / RESTART buttons when a job started or ended while it was open
Now: it follows the job live
The offload Data Sheet PDF could cut the row on a page seam (clip 288 of 300)
Now: rows never split across the page break
LIVERELEASED
1.3.0CartDATE07 NOV 2026CHANGES107LIVE
Reports read the offload's own checks, one offload window with ticked live drives, the next card queues, Sony looks right, less CPU, SORT, buttons that stay put, and updates inside the app.
NEW 04
New changelog page at engram.metawrks.com: every release, searchable, by area (Offload, Reports, Viewer & colour, Library, Transcode)
New option: Settings › Functionality › "Viewer as a Rec.709 monitor (gamma 2.4)", so the viewer matches your SDI monitor. Off by default
New Library › New Day (SHIFT+CMD+D): copies the latest DAY's folders as the next DAY, empty, with UNDO
New Library › Wrap Day Report: the whole DAY's report from the library, using the offload's own checks — nothing is scanned again
IMPROVED 02
An offload could never run without its <drive>/…/<CAM>/<MAG> folders now — if they are missing, it says so and copies nothing
Note blocks that explained the code are off the screen. Controls, values and real warnings stay
FIXED 101
There was no way to update from inside the app, and Settings said "No update channel configured yet"
Now: ENGRAM › Check for Updates… and Settings › System › UPDATES › CHECK FOR UPDATES. It also checks by itself at launch and every 4 hours (switch it off in the same place)
A new version was never announced
Now: a calm window shows the new version, its build, the date and the changes in a text block, with UPDATE, REMIND ME LATER (asks again tomorrow) and SKIP THIS VERSION (stays quiet until a newer one)
An update could have landed mid-job
Now: nothing installs, and the window doesn't pop up, while an offload, verify, report or transcode runs or cards are queued; it says what it's waiting for
An update had no way back
Now: UPDATE checks the package and gets it ready; RESTART TO USE swaps it in when you choose, keeps the version you had, and puts the old one back by itself if the new one doesn't check out. ROLL BACK in Settings goes back one step in one press
About said "1.2" on every build
Now: About and Settings show the real version and build: 1.3.0 (build 130)
Sony S-Log3 clips showed ENGRAM's own S-Log3 curve, about a stop hot in the highlights
Now: they show Sony's own LC-709TypeA LUT, taken from your Resolve LUT folder. Pick another Sony look (LC-709, Cine+709, SLog2-709, s709) in Settings › Functionality › SONY LOOK. ENGRAM's own curve is still in the LUT list, now named "ENGRAM generic, not Sony"
S-Cinetone clips (C CAM day 1) got an S-Log3 LUT on top and went half black
Now: S-Cinetone, Rec.709, Cine and HG clips show as recorded, with no LUT. HLG reads as HLG. S-Log2 asks you to set it by hand
FX6 MXF clips fed the Sony LUT the wrong code values (crushed blacks, hot whites) and didn't match FX3/FX30 MP4s through the same LUT
Now: the LUT gets the values the camera recorded; FX6 and FX3 now agree within half a code value. On the D002 demo clip the picture is within ΔE 0.44 (avg) of Sony's own LUT, from 2.21 before
S-Gamut3 clips (not .Cine) went through a .Cine LUT as-is
Now: they're converted to S-Gamut3.Cine first (Sony's S-Gamut3 LUT is used instead when you have it)
False colour with the LUT on read the LUT's picture as log: +2 stops showed as +6
Now: false colour always reads the camera's own image: +2 shows +2.0, −4 to +4 within 0.05 stop, LUT on or off
Posters, scrub frames and the viewer decoded about 2 code values dark
Now: decoded exactly (old cached frames rebuild by themselves)
A paused frame was a lower-quality JPEG of the clip
Now: it's the real full-quality frame, the same one playback shows
33-point LUTs lost detail between their points in the viewer
Now: they're applied on a finer grid
Transcodes with a Sony S-Log3 LUT on FX6 MXF had the same wrong code values
Now: they get the camera's values too
Pressing START (or plugging in the next card) while a card copied did nothing
Now: the card QUEUES. When the copying card is certified, the next one starts by itself. If it isn't certified, the queue stops and tells you which card; START NEXT or CLEAR on the queue bar. A card that needs you (a red × drive, a duplicate) stays open in the offload window
After-copy options were buried in Settings
Now: the offload window has an AFTER row: EJECT CARD · REPORT · TRANSCODE, ticked right there
Nothing told you a card was done with
Now: a lime SAFE TO PULL bar and a notification ("SAFE TO PULL — A005 · 3/3 DRIVES VERIFIED") once every drive verified and the card re-read clean
A card going in only showed a small bar (and only if you had turned it on)
Now: it opens the offload window, filled in with the card. It never starts by itself unless you turn on "Then start by itself after 5 seconds" (off by default); STARTING IN 5 shows in the window, STOP holds it
Routes had no camera letter
Now: new CAM LETTER token: A005 lands in "A CAM" by itself
SAFE TO PULL showed after RUN ON 1 when one of your usual 2 drives was offline ("1/1 DRIVES VERIFIED" while the card was only on one drive)
Now: SAFE TO PULL only shows when the card is on every drive you usually use; the count matches the result line (1/2 stays amber, no SAFE TO PULL)
A queued card went into whatever bin was selected when it started, not the bin you queued it into
Now: the bin is locked in when you press QUEUE and shows on the card's chip (B001 → A CAM). If that bin was deleted, moved to another show or the project changed, the queue stops with NEEDS YOU instead of copying
The library file was rewritten whole, about once a second, while a card copied (61 MB each time on today's job)
Now: only a small catalog (≈2 MB) is written, at most every few seconds during a copy; each clip's metadata is added as one small line instead. Your old library file is kept next to it as a backup the first time
Opening or closing a folder in the library rewrote the whole library
Now: it saves only which folders are open
Clip metadata landing during an offload redrew the library once per clip
Now: it lands in batches, about once a second
Scrub previews, posters and the viewer decoded on the CPU, up to 4–6 cores per clip
Now: Sony XAVC-I, XAVC-L and XAVC HS (H.264 / H.265) decode on the GPU, with the exact same picture (checked byte for byte). ProRes stays on the CPU, because the GPU decoder is 2–3 code values off
One scrub-preview builder could take 4–6 cores and 1.2 GB
Now: it uses 2 at most, runs at the lowest priority, and reads only keyframes on long-GOP clips
After each card, scrub previews for the whole drive were built while the next card copied
Now: they wait until the copy is done, and stop if a new card starts
Each drive's copy was checked one drive after the other
Now: all drives are checked at the same time, on separate cores
Clip metadata was read one clip at a time
Now: several at once across the cores; fewer while a card copies, so the copy keeps its speed
The CPU meter showed the app only, not ffmpeg or the copy engine
Now: it shows the app plus everything it started, and the GPU next to it (click it for the app's own share)
Scopes kept redrawing on a paused frame
Now: they rest until the picture changes
Picking a source opened one window, then a scan window, then the offload window
Now: it is ONE window: card list → scanning → setup, all in the same window. Nothing asks for a destination after the card
You had to pick the card again even when you were already looking at it
Now: a card you're browsing in the grid (or the one card plugged in) opens the offload window straight on it. SOURCE in the window swaps the card
Drives showed OFFLINE only, and a dead drive could still be ticked
Now: every drive is a tile with a tick box and a status: READY (and how many more cards of this size fit), or a red × OFFLINE / READ-ONLY / SOURCE / FULL. A red × drive can't be ticked
START could make folders before finding a drive it couldn't write
Now: every ticked drive is checked again at START, before any folder is made; a red × drive holds START and RUN ON N copies to the rest
× on a drive was final
Now: × is instant and UNDO puts it back (place, tick, route)
The drive list was the same for every project
Now: each project remembers its own drives; a new project starts with the drives you have
The finished result could show inside the offload window
Now: the window never shows a result or DONE; the copy's progress and verdict are in TASKS and on the library strip
There was no way to copy a day's folders
Now: right-click › Duplicate Structure or CMD+D: DAY 003 → DAY 004 with every OCF / CAM folder, empty (no rolls, no clips, nothing marked verified), placed below the last day and ready to rename. UNDO in the note at the bottom. It only makes the folders on your drives when you tick "Also Make On Drives" (off by default)
GRID/VIEWER/LIST/DATA slid sideways and shrank when you opened VIEWER (a BACK chip appeared in front of them)
Now: the four buttons sit in the exact same spot and size on every page; GRID, Esc and CMD+1 still go back
The GRID/VIEWER/LIST/DATA bar sometimes dropped to icons only on one page
Now: it always shows its words, every button the same width
Switching LIBRARY/TRANSCODE/TASKS made the whole top band drop and fade in
Now: the page crossfades in place; nothing in the band moves
Buttons came in different heights (18, 20, 22, 26, 28, 30)
Now: every boxed button in a top band is one height, the same as LIBRARY/TRANSCODE/TASKS; window and footer buttons share one height of their own; both follow UI Scale
The gear and the CPU meter were taller than the chips beside them and ignored UI Scale
Now: same box as the chips
The selected tab turned bold and nudged its neighbours
Now: selection shows by colour only; nothing reflows
TASKS' CARDS/TIMELINE were two loose icons on the right
Now: one boxed switch at the left, where LIBRARY's view switch sits
INFO disappeared in DATA, and TASKS' chips jumped when CLEARED appeared
Now: both are always there, dimmed when they can't be used
Going to TRANSCODE pushed the project name sideways
Now: the project name never moves
Going LIBRARY
Now: TASKS → LIBRARY landed on GRID → it comes back to the same clip in VIEWER
OFFLOAD window tabs (LIBRARY & METADATA / BACKUP & VERIFY, rename scope, existing files) were a different size
Now: same box as the other tabs
There was no way to put clips in order
Now: SORT button in the LIBRARY band: Name, Record time, Backup time, Size, Duration, Camera, Reel / Mag, Format / codec, Frame rate; ascending or descending. One order for GRID, LIST and the VIEWER strip, remembered between launches. LIST's NAME / TYPE / SIZE headers use the same order. Clips missing a value go last. Sorting never re-reads the clips
Backup time had nowhere to come from
Now: each clip now records when it landed verified
Exporting a report re-read and re-checked every clip that was already verified on the offload
Now: the report reads what the offload already recorded; 0 clips are read from the drives for a verified bin
There was no way to force a full re-read when you actually want one
Now: RE-SCAN MEDIA in the REPORT window, off by default, for that one report only
Picking a source sometimes showed the last card's finished/report window instead of the new card
Now: every offload button closes the finished job first; the window always opens on the new card's setup (the result stays in TASKS)
After a report, View › Offload and GO TO OFFLOAD reopened the report, and CMD+R ran it again
Now: they open OFFLOAD; CMD+R after a finished report opens its setup
RETRY / RUN AGAIN could fail on files the first run already copied, or copy the whole card when you had picked some files
Now: it re-checks what is already there (Skip Identical) and copies the same files you picked
The bin's Offload buttons stayed live while a card was copying
Now: they grey out until the copy ends
Sources vanished from the list while a drive was spinning up or busy
Now: the drive stays listed and says BUSY, and comes back to normal on its own
A clip still being written (size changed after the scan) could pass
Now: that clip FAILS and is named: "SOURCE CHANGED DURING OFFLOAD"
NOT CERTIFIED didn't say which file
Now: it names the file, the reason and the fix
The REPORT tab could walk a whole drive
Now: it reports on the library selection; a whole drive only when you pick it yourself
Camera thumbnails, XML/JSON sidecars and system files counted as "files backed up"
Now: the total counts media; the rest are their own rows (still copied and in the ASC-MHL)
The grid opened full of JPGs and thumbnails
Now: motion clips first; stills behind one chip
The long drive list sat across the grid
Now: hidden behind one drive toggle, remembered
Sony cards: blank metadata on a clip whose XML landed after it, odd frame rates, a 4294967295 "serial", reel letters other than C/L missed
Now: probed again once the XML lands, exact rates, serial left blank, any camera letter
With RENAME on, the dailies transcode after an offload looked for the card's original names and found nothing
Now: it opens the renamed copies that actually landed on the drive
A START that was refused left empty <bin>/<MAG> folders on every drive
Now: the empty folders that START made are taken back; a folder that was already there, or holds a file, is never touched
RUN ON N stayed armed after a refused START
Now: a refused START asks again
If the engine itself failed to launch, RETRY could register the copy into a bin that had been removed
Now: the mag bin and its folders stay for RETRY
A card pulled after you opened it still let START make the mag bin and folders, then jumped to an empty bin
Now: START says "Source is offline" and stays put; nothing is made
If a job refused to start after START, the window closed onto an empty bin and the drive folders stayed armed for the next run
Now: the card stays open in the window, the bin and your answers go back, and nothing is left armed
The dailies transcode after an offload could read a drive you skipped with RUN ON N, or look at the drive root
Now: it reads the copy that job just wrote and verified, in <drive>/…/<CAM>/<MAG>
CMD+N on an open card kept that card's resume answer
Now: the new card replaces it the same way the top OFFLOAD button does
CMD+R with no card did nothing
Now: it asks for the card, same as the top OFFLOAD button
Clicking a bin with no drives listed added that bin's drive, right-click didn't
Now: neither adds a drive; you add drives in the window with ADD DESTINATION
A drive that dropped without warning (the OWC RAID) could still count as plugged in for a few seconds
Now: START, RETRY and COPY TO MISSING check every drive fresh before writing; no stray folder on the Mac's disk
A destination folder on the Mac's own disk read as a different drive every card
Now: it counts as the same drive, so no false "drive missing"
The library's "Offload Into This Bin" and the OFFLOAD button up top behaved differently
Now: both ask for the card, open the same window, land on every drive in the same folders and use the same START check; the only difference is which bin it goes into
With no drive listed, the library door asked for a destination the top button didn't
Now: both open the window and you add drives there with ADD DESTINATION
The top OFFLOAD button reopened the last card without asking
Now: it asks for the card, starting on the last one, so one press re-picks it
CMD+R while a card was open skipped the drive check and could copy loose to each drive root
Now: CMD+R does exactly what START does, red line and all
A RETRY that was refused (card not back in yet) hid the result and left the next RUN copying loose to drive roots
Now: a refused RETRY keeps the result on screen and leaves nothing behind
RUN ON N left the skipped drive unticked on every card after
Now: the drive is ticked again for the next card, and warns if it is still missing
The finished-job banner could say "3/3 drives " when the job was then marked NOT CERTIFIED
Now: the banner waits for the final verdict
Checking drives on every redraw could stutter the window while a shuttle spun up
Now: drive status is remembered and refreshed when a drive comes or goes
Right-click "Offload Into This Bin" cut your drives down to one
Now: every drive you listed stays; the bin only picks the folder on each drive
New Offload (CMD+N) wiped your drive list
Now: it clears the card and slate only; your drives stay
Clicking a bin on another shuttle swapped out your first drive
Now: clicking a bin never changes your drives
Nothing warned when fewer drives were going than last card
Now: START shows a red line ("2 drives missing: …") with ADD BACK or RUN ON 1 DRIVE, one press each
An unplugged drive could vanish from the list
Now: it stays listed and says OFFLINE
"Verified" read the same for 1 drive as for 3
Now: the result, the library strip and the notification say "I004 · 3/3 drives 008 · 006 · 003", amber when short
No way to see which mags were on fewer drives
Now: each mag shows its copy count (3×, amber when short); right-click a DAY or CAM › Copies… lists the short ones with COPY TO MISSING
A card staged while the last one was still copying renamed the last card's record (C009 saved as I004)
Now: each record keeps its own bin name; a new card can't be staged mid-copy
COPIES counted a stopped or part-copied card as a full copy
Now: a drive counts only when the whole card verified on it
RETRY / RUN AGAIN dropped the card loose at each drive root and skipped the drive check
Now: it re-runs into the same <drive>/…/<CAM>/<MAG> folders and asks the same drive question START does
A failed drive still read "3/3 drives ×"
Now: the count is drives that verified, and the bad one is named ("2/3 drives × 008 · 003 — 006 FAILED")
Taking a drive off with × kept warning "1 drive missing" on every card
Now: × forgets that drive
COPY TO MISSING failed on a half-filled folder and could start a transcode
Now: it re-checks files already there (Skip Identical) and never chains a transcode
RUN ON N could skip an offline drive and then do nothing when another warning was open
Now: it is greyed until that warning is cleared, and skips nothing
A leftover /Volumes folder from a dropped drive counted as the drive
Now: only a real mounted drive counts; the leftover reads OFFLINE
1.2DATE24 SEP 2026CHANGES17LIVE
MHL read-back, a failed verify reaches the library and the next report, P2 and AVC-Intra cards read right, quieter PDF export.
IMPROVED 05
Client sheets: a failed later VERIFY prints "VERIFICATION FAILED on N file(s)… NOT certified"
P2 cards: "1 clips · 5 camera attachments (audio files, not counted as clips)" instead of 6 clips
A proxy batch with a vanished render blocks the ASC-MHL for the whole batch
Sony/Canon posters and AUTO proxies regenerate once with correct (darker) blacks
Proxies drop sound that runs past the last video frame, normally under 1 frame
FIXED 12
A failed VERIFY never reached the library, the next REPORT or the job record
Now: it does; a REPORT is green only when the engine certified it
Proxy intake could copy files that changed after they were verified
Now: copies exactly the verified set, pinned by size, mtime and inode
HDE archive followed symlinks inside its destination
Now: refused, one clip at a time, link named
A dead card read looked like progress
Now: a stalled read is named within 30 s; STOP escalates in about 8 s
Re-runs refused on free space they didn't need
Now: preflight counts only bytes still to write
P2/VariCam per-channel audio MXFs counted as clips (1 take read as 6)
Now: attachments of the video clip
P2 proxies said "no sound to carry"
Now: they carry the AUDIO/ channel files, or refuse if any is damaged
AVC-Intra 1080 read 1088/544
Now: display size and frame layout used
Clips whose sound outlasts the picture could not be proxied
Now: reader trimmed to the picture
BT.2020/BT.470BG MXF transcodes died; 10–16-bit RGB masters came out ~19 codes green-low
Now: mapped correctly, RGB decoded as RGB
Different ffmpeg paths guessed range and matrix differently
Now: one shared decision everywhere ffmpeg decodes; Sony/Canon masters read limited unless the file proves otherwise
STOP killed the Resolve helper mid-render, leaving scratch projects and staging partials
Now: cleanup runs; partials never reach offload
1.1DATE23 SEP 2026CHANGES50LIVE
Tighter custody: a changed file can't be re-certified, client sheets only claim what the checks prove, proxies and renders no longer count as clips, camera metadata matches the camera.
IMPROVED 25
Cards first offloaded by Silverstack or another tool: cover reads "N of M files hashed at offload". Real E002: "12 of 14"
Attachment caption drops "and verified" unless every attachment verified
Reconciliation line ends at the count unless every file is verified and in the MHL
Library reports: fewer CLIPS and shorter runtime where proxies were double-counted. Proxies show "Delivery: produced proxy"
RED span segments: caption says "clip segments", not "proxies"
ARRI Color Space column now filled: AWG4 (ALEXA 35/265), AWG3 (Mini LF class). Was "—"
ARRI Gamma: LogC3 on Mini/LF/SXT/AMIRA/65. Unknown camera prints its own "LOG-C"
ARRI codec: HDE clips read ARRIRAW HDE. ARRICORE reads ARRICORE
Imperial focus and drop-frame TC values change to the correct ones
BRAW rows show real audio. No false NO AUDIO. FCPXML and ALE audio fields fill too
Sony XAVC-S cards with DCIM read Sony. Cards with no brand evidence show a blank camera instead of Canon
ALEXA 35 ARRIRAW posters shift slightly in colour and grain. Now matches ARRI's render
BRAW posters from clips with an embedded LUT look different. LUT is no longer applied twice
Volume slider now drives synced sound
Transport refusal note stacks above the SYNCED AUDIO chip
LOG view on Mini/LF/SXT ARRIRAW says LogC3. E-stops read "LogC3 EI800"
LUT picker refuses unknown keywords and shaper LUTs, and names the reason
CRM PROXIES AUDIO row: "Linear PCM · 16-bit · 2 ch summed from source"
Resolve button works from a Finder launch
MXF OP-Atom with sound lands video-only, with a row note
A symlinked subfolder under the destination is refused, with the link named
A proxy folder mixing audio formats is refused whole
Sync flags more clips ambiguous when two rolls share a name
Scaled ProRes proxies look slightly sharper (Lanczos scaler)
Every transcode run line ends in "· colour=primaries/transfer/matrix"
FIXED 25
Card B's MHL re-certified card A's changed files as new
Now: each path is checked against its first digest; a mismatch blocks certification (exit 11)
Reports took any MHL digest as proof
Now: recorded size is compared; truncated or swapped files read SIZE MISMATCH, NOT CERTIFIED
A symlinked folder under the destination was followed
Now: refused before any write or skip; card can't be wiped on a false green
Source-verify re-read the card from RAM (0 MB off the device)
Now: card pages purged first; a 1 GiB file now reads 1024 MB off the card
Sheets said "verified" beside Unverified 2
Now: claims match the evidence; other tools' MHLs read "hashed at offload"
Library REPORT counted proxies as takes (13 CRM
Now: 26 CLIPS) → proxies matched by filename; 13 CLIPS; CRM posters upgrade from 160x120
A 3-segment RED take read "3 clips"
Now: one take, one clip; a missing segment is flagged
FX3 / a7S III cards with a still read CANON, reel C000–C003 by launch
Now: read Sony, one fixed reel
Every ALE "LOG-C" became LogC4
Now: LogC3/AWG3 on Mini, Mini LF, LF, SXT, 65, AMIRA; LogC4/AWG4 on 35/265 only
ARRI imperial focus 12x too far (10 ft read 36.58 m)
Now: 3.05 m
ARRI drop-frame TC off (1 h @ 29.97 DF read 00:59:56;12)
Now: 01:00:00;00
ARRICORE codec read "—", HDE read ARRIRAW
Now: ARRICORE, ARRIRAW HDE
Sync merged two different WAVs with the same name
Now: both kept, ranked; clip marked ambiguous
ProRes masters and proxies were built from 8-bit RGB (645 of 877 codes)
Now: 10-bit YCbCr path
OP-Atom MXF with sound left a 0-byte .mxf and deleted the master
Now: master kept; video-only MXF with an "audio not carried" note
ALEXA 35 ARRIRAW decoded without its unit calibration
Now: calibrated; frames match ARRI's reference to 8e-6
BRAW embedded LUT applied on top of Rec.709 output
Now: LUT off under Rec.709; refused overrides fail loudly
29 Resolve LUTs passed the picker, then failed every proxy (exit 5)
Now: viewer and engine accept the same 130 of 212 LUTs
A mixed 24-bit / 16-bit proxy folder got sealed
Now: whole batch refused, file and field named
Resolve button said UNAVAILABLE HERE when ENGRAM opened from Finder
Now: finds the bundled ffprobe/ffmpeg; works
CRM proxy window promised source audio channels
Now: "Linear PCM · 16-bit · 2 ch summed from source", plus channel warnings
Handoff ran a plain offload on staging
Now: runs the proxy-intake gate first
Volume slider didn't change synced sound
Now: it does, live
Transport refusal note sat below the SYNCED AUDIO chip
Now: above it, as intended
Playback held every frame in RAM (2K proxy: 4.28 GB in 20 s)
Now: drained per frame; growth 1024 MB → 0 MB