Skip to content

Releases

v6.4.1

Choose a tag to compare

@github-actions github-actions released this 13 Nov 13:53

✨ New Features

  • Added support for hebrew folder names - Thank you @HowToTricks

What's Changed

New Contributors

Full Changelog: v6.4.0...v6.4.1

v6.4.0

Choose a tag to compare

@github-actions github-actions released this 23 Sep 22:02
77a31f6

✨ New Features

  • Preserve Google Photos captions/descriptions in EXIF (native JPEG writer, consolidated ExifTool writes) — Takeout JSON sidecars carry each item's caption (the note typed onto a photo/video in Google Photos) in the description field, but GPTH previously only read these sidecars for photoTakenTime/GPS and never preserved the caption — once the sidecar was gone, so was the caption. extractAllFromJson/extractAllFromJsonCached (Step 4) now also capture description from the same combined JSON read used for date/GPS (no extra file I/O), subject to the same issue #139 own-sidecar guard so a caption is never borrowed from an unrelated photo's sidecar. MediaEntity gained a description field (mirroring gpsCoordinates: not serialized to progress.json, since it's cheap to re-extract on resume). Step 7 embeds the caption into EXIF ImageDescription (tag 0x010E), with routing that never spawns an extra ExifTool invocation for a caption: JPEG files are written natively with the built-in Dart writer (image package — no ExifTool call at all, even when ExifTool is installed; the write shares the file's single native metadata pass), and only fall back to ExifTool (queued into the file's existing tag batch) if the native write fails; all other formats (PNG, HEIC, videos) queue the caption into the file's single already-queued ExifTool write. An entity with only a caption and no date/GPS is no longer skipped by Step 7. Captions are embedded as UTF-8 bytes and survive alongside embedded thumbnails and JFIF headers (issue #132-safe). The XMP-dc:Description tag remains fully implemented but is currently not written (captions go to EXIF only); see README "Caption/Description Metadata Writing".

What's Changed

  • Preserve Google Photos captions/descriptions in EXIF/XMP by @nathanzm in #151

New Contributors

Full Changelog: v6.3.0...v6.4.0

v6.3.0

Choose a tag to compare

@github-actions github-actions released this 30 Aug 19:11

🐛 Bug Fixes

  • still mode no longer double-moves a sidecar claimed by a .MP entity (race: PathNotFoundException + "Primary file is still unaccounted for") — In PixelMpTransformService._transformPixelPrimariesToStill, when a .MP entity was redirected to a sidecar (e.g. photo.MP → photo.MP.jpg), _removeEntityDuplicates() was supposed to drop the separate entity that owned that sidecar so only one entity would move it. But the owner-exclusion guard used value equality (e == owner), and MediaEntity.== keys on sourcePath/targetPath/isShortcut. Because the owner's primaryFile.sourcePath had already been mutated to the sidecar path before the dedup call, the owner and the genuine sidecar entity now compared equal under ==, so the guard if (e == owner) return false; wrongly skipped the duplicate. Two entities therefore remained pointing at the same file; the parallel mover raced — one move succeeded, the other threw PathNotFoundException (source already gone, errno 2 / "The system cannot find the file specified") and the post-finalize integrity check flagged it as "Primary file is still unaccounted for after strategy finalization". This was reproducible cross-platform (seen on macOS /var/folders/... and Windows %TEMP%..., notably in the sidecar_priority test). The owner-exclusion check now uses identity (identical(e, owner)) in both the primary-duplicate scan and the secondary-strip loop, which correctly excludes the exact owner object regardless of its mutated sourcePath. The fix is platform-agnostic.
  • --transform-pixel-mp jpg now produces a valid Google Motion Photo V2 container — Previously LivePhotoCreatorService._createLivePhotoContainer() concatenated the JPEG and MP4 bytes with no XMP offset marker, so while the file played as a plain JPEG, other apps (Google Photos, Android gallery) could not locate the embedded video and treated it as a static image. The container builder now strips any stale motion-photo XMP from the still image (via MotionPhotoExtractorService.stripMotionPhotoXmp), injects a fresh XMP APP1 segment immediately after the JPEG SOI marker declaring GCamera:MotionPhoto="1" and GCamera:MicroVideoOffset (the byte count from the end of the file to the appended MP4 — the exact semantics the existing _parseGoogleMotionPhotoV2 reader consumes), and then appends the MP4 verbatim. This mirrors what Google Photos natively exports and round-trips through GPTH's own MotionPhotos.isMotionPhoto() detector. The interactive prompt's "experimental / known issues" warning for the jpg mode has been softened accordingly.
  • still mode sidecar selection now prefers .jpg over .jpeg — _findPreferredStillImagePath() previously picked the largest file among all sidecar candidates, which could select a larger .jpeg duplicate over the real .jpg still (e.g. a 150 KB photo.MP.jpeg duplicate wrongly beat a 100 KB photo.MP.jpg actual still). Candidates are now grouped into priority tiers by extension (.jpg/.JPG before .jpeg/.JPEG); the first tier containing an existing file wins, and size is only used as a tiebreaker within a tier.

🛠️ Maintenance

  • Regression test for issue #120 — An e2e test (test/e2e/e2e_missing_coverage_test.dart) reproduces the issue #120 input layout (a Pixel .MP file with its .MP.jpg sidecar) under --transform-pixel-mp mp4 and asserts the output contains exactly one .mp4, the .MP.jpg sidecar, and no stray (1).jpg collision file. The double-move that produced the stray file (the sidecar being moved once as a secondary of the .MP entity and again as its own primary) was already prevented by _removeEntityDuplicates() in PixelMpTransformService; this test guards against its return.
  • removed some dead code
  • fixed comments where the comments did not reflect the real behaviour

✨ New Features

  • Local timezone conversion for Google Photos re-upload (--local-timezone) — Google Photos Takeout stores photo capture times as UTC in the photoTakenTime.timestamp field. GPTH wrote this UTC clock to the EXIF DateTimeOriginal tag together with OffsetTime=+00:00. However, Google Photos ignores the OffsetTime tag when re-uploading and interprets the naive DateTimeOriginal clock as local time, so photos appear hours off (e.g. 8 hours early for a GMT+8 user — issue #145). The new --local-timezone flag (e.g. --local-timezone +08:00) and interactive prompt convert UTC photo dates to the user's local timezone: the local clock (UTC instant + offset) is written to the EXIF date tags together with the correct OffsetTime (e.g. +08:00), and date-based output folders use the local date, so the re-uploaded timeline matches the original. File timestamps (Step 8) are intentionally left as the true UTC instant — the OS already displays them as local time on the user's own machine. The flag accepts ±HH:MM (e.g. +08:00, -05:30), ±H (e.g. -5), HHMM (e.g. +0530), and Z/UTC aliases, and is also available as an interactive prompt. When not set, the previous UTC behaviour is preserved (backward compatible).
  • Custom date folder format for --divide-to-dates (issue #142) — Previously --divide-to-dates only accepted the four presets 0 (one folder), 1 (year), 2 (year/month), 3 (year/month/day). You can now pass an arbitrary folder-structure template using the tokens yyyy/yy/mm/dd, e.g. --divide-to-dates "yyyy/yyyy-mm" produces 2026/2026-08/ (two folder levels). / separates folder levels; any other characters (-, _, ., …) are kept as literal text within a folder name. The legacy 0-3 presets continue to work unchanged (backward compatible). The interactive mode gained a new [4] - custom format menu option with a validation loop. Albums remain flattened regardless of the chosen format. A null date still falls back to the date-unknown folder.

Full Changelog: v6.2.1...v6.3.0

v6.2.1

Choose a tag to compare

@github-actions github-actions released this 26 Aug 20:25

⚡ Performance

  • JSON sidecar path and content caching across Steps 2 and 4 — Every media file's JSON sidecar was previously located and parsed twice: once in Step 2 (partner-sharing detection) and again in Step 4 (date + GPS extraction). Each lookup involved 5–10+ File.exists() calls trying different naming patterns (supplemental-metadata, truncated variants, numbered forms, etc.), and each parse read and decoded the full JSON file. Step 2 now caches the resolved sidecar path and the isOwnSidecar confidence flag directly on the FileEntity, and Step 4 reuses both — skipping the expensive findJsonForFileWithConfidence lookup entirely. A new LRU cache (50,000 entries) in JsonMetadataMatcherService stores parsed JSON content keyed by sidecar path, so each sidecar file is read and decoded exactly once across both steps. For a 50,000-file takeout this eliminates ~100,000 redundant filesystem stat calls and ~50,000 duplicate JSON reads. The isOwnSidecar flag is preserved through MediaEntity normalization (which creates new FileEntity instances during ranking), ensuring the issue #139 cross-photo GPS guard remains intact.
  • Parallelized Step 8 creation-time updates — File creation timestamps were updated one file at a time in a sequential loop, each call opening a Win32 handle (CreateFile + SetFileTime + CloseHandle) or POSIX utimensat call. These operations are independent and are now batched with Future.wait using the fileIO concurrency limit. For a 50,000-file takeout this turns the final step from fully single-threaded to parallel, significantly reducing wall-clock time on Windows.
  • Extension-first filter in wherePhotoVideo() — The discovery filter called lookupMimeType() for every file during recursive directory listing, including non-media files (JSON sidecars, etc.). The extension check (MediaExtensions.additional set lookup + isMotionPhotoExtension) is now performed first; only files that don't match the special-extension fast path fall through to the more expensive lookupMimeType call. For a 100,000-file takeout this avoids ~100,000 unnecessary MIME map lookups during discovery. The fast path only matches extensions that lookupMimeType does not recognize (.mp, .mv, .dng, .cr2, .cover, .mp~N), so standard extensions like .jpg/.png/.mp4 still go through lookupMimeType exactly as before — Step 1's content-based extension fixing (which reads file headers, not the filter's MIME result) is unaffected.
  • Eliminated redundant stat() + length() in MediaHashService.calculateFileHash — The hash calculation called file.stat() to generate the cache key, then file.length() separately to get the file size — two syscalls per file. The size is now reused from the already-fetched FileStat, eliminating ~50,000 redundant syscalls during Step 3 for a 50,000-file takeout.
  • Partial JPEG read for EXIF date-tag check in Step 7 — hasExistingExifDateTag loaded the entire JPEG into memory (file.readAsBytes()) just to check whether EXIF date tags exist. EXIF data is always near the start of a JPEG (the APP1 marker follows SOI + JFIF), so only the first 64KB is now read (file.openRead(0, 65536)). For a 10MB JPEG this reduces memory allocation from 10MB to 64KB per file.

🛠️ Maintenance

  • Test suite for JSON sidecar caching — Unit tests (test/unit/json_sidecar_cache_test.dart) cover the FileEntity sidecar path and confidence flag fields (defaults, set/get, JSON round-trip), extractAllFromJsonCached with cached own-sidecar (returns date+GPS), cached isOwnSidecar=false (drops date+GPS per issue #139), fallback to full lookup when no cached path exists (secondary files), graceful handling of deleted sidecar files, readJsonContentCached content caching and invalid-input handling, and Step 2 → Step 4 integration verifying the cached path is populated during discovery and reused during date extraction.

Full Changelog: v6.2.0...v6.2.1

v6.2.0

Choose a tag to compare

@github-actions github-actions released this 26 Aug 18:19

🐛 Bug Fixes

  • Shortcut mode no longer retains physical album videos — In --albums=shortcut mode, .MP4 and .MOV companion files (and other non-image media) were left as regular files in the Albums/ folder instead of being replaced by symlinks to the canonical file in ALL_PHOTOS. The root cause was the shortcut reuse check: it treated any existing filesystem entry at the target name (file, link, or directory) as a reusable shortcut, so a stale physical file from a previous or interrupted run was silently kept instead of being replaced by a link. Shortcut reuse now only accepts actual symbolic links (Link.exists()), never regular files. When a regular file is found at the intended shortcut location, identical duplicates (same content hash as the canonical ALL_PHOTOS file) are removed, and distinct files are preserved under a Shortcut Conflicts/ subfolder before the expected link is created — so Albums/ contains only shortcuts in shortcut mode. A new ShortcutIntegrityService runs after the EXIF write phase (Step 7) to verify and restore any album shortcuts that later stages may have broken or turned back into regular files: missing links are recreated from the entity's canonical output file, broken links are deleted and relinked, and unexpected regular files are moved to Shortcut Conflicts/ before the link is restored. This resolves issue #143.

⚡ Performance

  • Removed the Pool(1) mutex from MediaHashService cache access — Every LRU cache read and write in MediaHashService was serialized through a single-resource Pool(1) lock, adding a pool-scheduling hop to every hash lookup and store. Dart's cooperative async model is single-threaded per isolate, so map access between await points is already atomic — the lock provided no safety benefit and was a serialization point across all concurrent hash operations in Step 3 (merge/duplicate detection). Removing it eliminates the scheduling overhead for every calculateFileHash, calculateHashAndSize, and cache lookup call. For a 50,000-file takeout this removes ~100,000 pool scheduling cycles during Step 3.
  • Eliminated double directory traversal in Step 2 discovery — The discovery step traversed every year and album directory twice: once to count media files (for the progress bar total) and again to actually collect them. Both passes performed the same recursive directory.list() + 512-byte header read + MIME detection per file. The count pass has been removed entirely; files are now collected in a single traversal per directory, and the progress bar total is derived from the collected list. The bar now advances during the partner-sharing JSON batch phase (the slower part of discovery) instead of during file collection. For a 50,000-file takeout this eliminates ~50,000 redundant file header reads and halves the number of recursive directory listings.

🛠️ Maintenance

  • Regression test suites for issue #143 — Integration tests (test/integration/media_entity_moving_strategies_test.dart) cover HEIC, MP4, and MOV shortcut output (each album entry must remain a symlink), missing-link restoration after later processing stages, and physical-video recovery (a stale regular file at the shortcut location is removed when identical to the canonical file, or preserved under Shortcut Conflicts/ when distinct).

What's Changed

  • Fix shortcut mode retaining physical album videos by @jaimetur in #144

Full Changelog: v6.1.10...v6.2.0

v6.1.10

Choose a tag to compare

@github-actions github-actions released this 21 Aug 09:46

🐛 Bug Fixes

  • .cover (album cover) and .mp~<digits> (edited alternate) files are now recognized as Pixel motion photos — Both are MP4-container motion-photo files that Google Photos exports with non-standard extensions: .cover for album cover images and .mp~2 (.mp~1, .mp~3, …) for edited alternate versions of .mp files. Neither was recognized by dart:mime or the hardcoded MediaExtensions.additional list, so wherePhotoVideo() silently dropped them in Step 1 & Step 2 — they were never discovered, never extension-fixed, and never transformed. A new centralized MediaExtensions.isMotionPhotoExtension() helper (the single source of truth for motion-photo extension recognition) now matches .mp, .mv, .cover, and the .mp~<digits> regex family, and all hardcoded == '.mp' || == '.mv' / endsWith('.mp') checks in the discovery gates (Step 1 & Step 2), the Step 1 skip guard, and the Step 6 transform dispatcher (mp4/jpg/still modes + secondary extension update) delegate to it. Files keep their original names (no input mutation, no collision risk when video.mp + video.mp~2 coexist), so JSON sidecar name-matching stays intact. A related fix in the Step 4 truncated-filename fixer ensures the title extension-stripping regex also handles .mp~<digits> (the ~ is not alphanumeric, so the old regex did not strip it, causing a false truncation match that doubled the extension). This resolves issue #138.
  • GPS coordinates and DateTime from an unrelated photo are no longer written to a file's EXIF — When a media file's own JSON sidecar was missing, the matcher fell back to aggressive strategies (-edited suffix removal, cross-extension MP4↔HEIC/JPG pairing, and the numbered cross-extension tier) that can match a different photo's sidecar. Because date and GPS were read from that one (wrong) matched sidecar together, the wrong photo's location — differing by tens to hundreds of kilometers — and its timestamp ended up in the output EXIF. This was the issue #139 symptom (~34.5% of GPS-tagged outputs had the wrong location, and a cluster of videos were mis-dated with another real video's timestamp). The matcher now exposes a confidence flag (isOwnSidecar) distinguishing a match that names this file (exact name, filesystem-truncated, Takeout bracket-swap, Google-added extension, Pixel .MP→.MP.jpg, same-file numbered forms) from a heuristic match that can point at a different photo. Date and GPS are now only trusted from an own-sidecar match; a heuristic match yields neither, so the file falls through to EXIF / guess / folder-year instead of inheriting a sibling's metadata. A related photo's date is not an acceptable heuristic here, so both fields are gated (not just GPS). This resolves issue #139.
  • Issue #133 album recovery is preserved — The cross-extension / numbered matching tiers that #133's orphan-recovery work added (and which #139's fix gates for date/GPS) are not used by the album-recovery path itself. Album association recovery reads sidecar title/photoTakenTime via getMediaNameCandidatesForJsonName in Step 2 and attaches membership as metadata; it does not call findJsonForFile, so the new confidence gate does not affect it. The matcher still finds cross-photo sidecars (for Step 1 extension fixing and album recovery) — it only refuses to use their date/GPS for EXIF writing. E2E tests confirm every album mode still materializes recovered orphan memberships on disk.
  • DateTime EXIF data is no longer overwritten from a different video in batched writes — When Step 7 flushed several files needing EXIF writes in a single ExifTool batch, the batch args were built as commonWriteArgs() + [tags₁, file₁, tags₂, file₂, …] on the assumption that each file's -Tag=Value block applied only to the filename immediately following it. ExifTool does not work that way: in argv, argfile (-@), and stay-open modes, all -Tag=Value assignments accumulate and the final set is applied to every file in the invocation. So the last file's tags overwrote every earlier file's tags — the exact issue #140 symptom, where IMG_5948.MOV ended up with its sibling IMG_9304.MOV's date in its XMP tags (DateTimeOriginal, DateTimeDigitized, ModifyDate) while QuickTime:CreateDate survived (written via a different path) and GPS stayed correct (ruling out the #139 cross-photo JSON-matching bug, which would have corrupted GPS too; the per-entity dateTaken in progress.json was already correct). ExifToolService.writeExifDataBatch now isolates each file by routing the batch through per-file writeExifDataSingle calls (each its own stay-open -execute round-trip, so there is no Perl-startup cost per file), and writeExifDataBatchViaArgFile delegates to it (the temp-argfile fallback, which had the same global-assignment problem, was removed). Step 7 also snapshots each file's tag map (Map<String, dynamic>.from(tagsToWrite)) when queueing, so later in-place mutation in the InteropIFD retry path (_stripOffsetTags / _retagEntryToXmpIfJpeg) cannot alter a sibling's queued tags. This resolves issue #140.

🛠️ Maintenance

  • Regression test suites for issue #138 at three levels — Unit tests (test/unit/motion_photo_extension_recognition_test.dart) cover MediaExtensions.isMotionPhotoExtension() across recognized extensions (.mp, .mv, .cover, .mp~1/.mp~2/.mp~12, mixed case, full paths) and rejections (.jpg, .mp4, .mpx, no extension, .mp~ without digits, .mp~abc, .cover.jpg); extended test/unit/invalid_extensions_test.dart verifies Step 1's skip guard does not rename .cover/.mp~2/.mp~12/.MP/.MV files to .mp4 (deferring to Step 6); an e2e test (test/e2e/e2e_issue_138_cover_and_mp_tilde_test.dart) drives the full pipeline with --transform-pixel-mp mp4 and default extension fixing, confirming .cover, .mp~2, and a .MP control are all discovered, transformed to .mp4, and placed in ALL_PHOTOS with metadata applied.
  • Regression test suites for issue #139 at three levels — Unit tests (test/unit/issue_139_cross_photo_gps_test.dart, test/unit/metadata_matcher_confidence_test.dart) cover the single-file gate and the isOwnSidecar confidence flag across own-sidecar cases (exact, shortened, bracket-swap, no-extension, Pixel .MP, numbered same-file) and not-own cases (-edited removal, cross-extension MP4→HEIC/JPG, numbered cross-extension, partial -ed). Integration tests (test/integration/issue_139_cross_photo_contamination_test.dart) cover multi-file folder layouts where own-sidecar and cross-photo candidates coexist, numbered duplicates each resolving to their own sidecar, tryhard-escalation guards, and GPS value round-trip correctness. An e2e test (test/e2e/e2e_issue_139_cross_photo_gps_test.dart) drives the full pipeline with --write-exif and reads back the output EXIF via ExifTool, confirming an MP4 motion companion does not inherit the JPG still's GPS/DateTime while the still itself keeps its own.
  • Regression test suites for issue #140 at two levels — Unit tests (test/unit/issue_140_video_batch_date_contamination_test.dart) capture the per-file tags emitted by writeExifDataSingle (the mock deliberately does not override writeExifDataBatch, so the real isolation logic runs) and assert two videos with different dates each receive their own XMP/EXIF date tags, plus that each file's queued tag map is a distinct instance (the defensive-copy snapshot property). An e2e test (test/e2e/e2e_issue_140_video_batch_date_test.dart) drives the full pipeline with --write-exif using the issue's exact scenario (two .MOV files, each with its own JSON sidecar carrying a distinct date) and reads back the XMP date tags via real ExifTool, confirming each video keeps its own date rather than inheriting the sibling's.

Full Changelog: v6.1.9...v6.1.10

v6.1.8

Choose a tag to compare

@github-actions github-actions released this 09 Jul 17:04

🛠️ Maintenance

  • Vendored the motion_photos package locally to resolve a dependency conflict — The motion_photos package (v1.0.0, last published 19 months ago) pins xml ^6.2.2, which is incompatible with image ^4.9.1 (which requires xml ^7.0.1), causing dart pub get to fail with a version-solving error. Rather than downgrade image (losing fixes and features) or wait for an upstream release that does not exist, the small surface area of motion_photos that GPTH uses (MotionPhotos, isMotionPhoto(), getMotionVideoIndex(), VideoIndex) is now re-implemented as a pure-Dart module at lib/common/services/media_services/motion_photos.dart. The implementation mirrors the package's algorithm verbatim (Boyer-Moore search for the MP4 ftyp mp42 header, then XMP fallback parsing GCamera:MicroVideoOffset / Item:Length), with the only difference being that XMP parsing uses regex instead of the xml package, dropping the transitive xml dependency entirely. The motion_photos dependency was removed from pubspec.yaml, and all imports in lib and test were updated to point at the local module.
  • Bumped the Dart SDK constraint to >=3.11.0 and pinned the CI Dart SDK to 3.11 — image 4.9.1 depends on xml ^7.0.1, which itself requires SDK ^3.11.0. The previous constraint (>=3.9.0) allowed Dart 3.9/3.10 where xml 7.0.1 cannot run, so dart pub get failed on CI runners that installed an older stable Dart. The SDK floor is now 3.11.0, and all three GitHub Actions workflows (build-nightly, new-release, build-and-create-release) pass sdk: 3.11 to dart-lang/setup-dart so the correct Dart version is installed on every build runner.

What's Changed

  • docs: add winget installation instructions for exiftool and 7zip by @mian196 in #135

New Contributors

Full Changelog: v6.1.6...v6.1.8

v6.1.6

Choose a tag to compare

@github-actions github-actions released this 06 Jul 16:10

✨ New Features

  • New --no-resume flag — GPTH automatically resumes a previous run when the output folder contains a progress.json. Pass --no-resume to discard that saved progress and always start fresh (--resume remains the default). When resume is disabled, the step-resume state is wiped at pipeline start so later runs cannot pick up a half-stale mixture of old and new step records.
  • Interactive mode now asks before resuming a previous run — When the selected output folder contains saved progress from an earlier run, interactive mode shows which steps were already completed and asks whether to resume or start fresh, instead of resuming silently. Part of the fix for issue #131.

🐛 Bug Fixes

  • Reusing folders from a previous run no longer aborts interactive mode — Selecting an extraction folder that still contained data from an earlier run made GPTH exit with a fatal error right after the disk-space notice, before any of the processing questions were asked. Interactive mode now validates the extraction folder the same way CLI mode does: a completed previous extraction of the same ZIP set is reused (extraction is skipped entirely), and any unsafe state (leftover data, a different ZIP set, an interrupted extraction) is explained with a prompt to select a different folder. Interactive runs also record the extraction sentinel in progress.json now, so future runs can safely detect and reuse the extracted data. This resolves issue #131.
  • Stale resume state is detected and discarded — If a previous run's progress.json marked processing as completed but the recorded output files no longer exist (e.g. the output folder was emptied between runs), GPTH used to "resume" by skipping every step and reporting success within seconds while doing nothing. The saved state is now validated against the files on disk at pipeline start and discarded with a clear warning when it is stale, so the run processes everything fresh. This resolves issue #131.
  • --fix mode no longer exits with ERROR_CODE_13 — In fix mode the output directory is the input directory, so the non-empty-output safety check always triggered and refused to run. The check is now skipped when output equals input. This resolves issue #128.
  • Partner-shared companion videos and numbered files are now sorted into PARTNER_SHARED — MP4/MOV companion videos paired with HEIC/JPG stills, and numbered files like IMG_1976(1).MP4 (whose sidecar is IMG_1976.HEIC.supplemental-metadata(1).json) or x(1).jpg, were not flagged as partner-shared because the cross-extension and numbered JSON matching strategies were not applied during discovery. JSON sidecar matching now supports numbered suffixes and cross-extension pairing, using three-level matching (exact, dot-boundary, first match) to prevent false positives such as photo matching photograph. This resolves issue #123.
  • ZIP selection no longer crashes when the file picker returns files without paths — On certain system configurations the file picker returns ZIP entries with a null path, which crashed interactive ZIP selection with a null-check exception. Null-path entries are now filtered out with a visible warning (also when only some of the selected files are affected, which previously dropped them silently), and a clear error message guides the user to extract manually if no usable ZIPs remain. This resolves issue #129.

Full Changelog: v6.1.5...v6.1.6

v6.1.5

Choose a tag to compare

@github-actions github-actions released this 06 May 15:24

🐛 Bug Fixes

  • Fixed that existing DateTime values in Exif were overwritten with json values, even if they already existed. This caused local timestamps to be overwritten with UTC timestamps from the jsons. It introduces another read operation for all media files, but also it means that less exif data is being written, which should balance each other out performance wise.

Full Changelog: v6.1.3...v6.1.5

v6.1.3

Choose a tag to compare

@github-actions github-actions released this 16 Apr 16:38

🐛 Bug Fixes

  • Album symlinks for Pixel Motion Photos now use the correct extension — When a .MP/.MV file was transformed to .jpg or a still image, the moving step only updated the primary file's path. Secondary references to the same file (album copies used for shortcut/symlink creation) still referenced the old .MP path, so album symlinks ended up named PXL_….MP pointing at a .jpg file. All secondary references are now updated in-place immediately after each transform, ensuring album symlinks use the correct filename.
  • Apple Live Photo .jpg siblings no longer appear as orphaned MP4 companions — When a Google Storage-Saver HEIC was fixed to .jpg (e.g. PXL_20230101.heic → PXL_20230101.jpg), the companion .MP4 suppression logic only looked for a .heic sibling, not a .jpg sibling. The .MP4 was therefore not suppressed and ended up as a stray file in the output. The check now also looks for an existing .jpg sibling.
  • Pixel Motion Photo video-index lookup made more robust — The motion_photos package's getMotionVideoIndex() method searches for an ftyp mp42 MP4 header pattern and falls back to XMP parsing. For Pixel .MP.jpg files that use a different MP4 container brand or a slightly different XMP attribute format, both lookups returned null, causing extraction to fail. A pure-Dart fallback parser now reads the GCamera:MicroVideoOffset attribute directly from the JPEG XMP segment via regex, so extraction works reliably across all Pixel motion photo variants.
  • Still-mode output .jpg files are no longer detected as motion photos — The JPEG extracted from a Pixel .MP file contains a stale XMP segment with GCamera:MicroVideoOffset and MicroVideo markers. Because no MP4 is appended to the extracted still, the offset is invalid, but the motion_photos package does not bounds-check it — isMotionPhoto() returns true for the plain JPEG. GPTH now strips the entire XMP APP1 segment from the extracted bytes before writing the output file, so photo managers no longer misidentify the still as an unplayable motion photo.

Full Changelog: v6.1.2...v6.1.3