feat(ocr): use Apple Vision for text recognition on macOS

docs/ENGINES.md

@@ -33,6 +33,7 @@ Filee chooses engines automatically. Settings → *Engines* shows their status,
 | Pandoc | **optional download** (~42 MB, 240 MB on disk) | Markdown ↔ DOCX/ODT/RTF, DOCX/ODT/HTML/RTF → Markdown, HTML ↔ DOCX/ODT, markup languages (reStructuredText, LaTeX, Org, AsciiDoc, Typst, MediaWiki, DocBook, Jupyter, Textile, man, OPML) ↔ Markdown/HTML/DOCX/ODT and each other (and → RTF, EPUB, TXT), Markdown and markup → PPTX slides |
 | Ghostscript | **optional download** (~20 MB, 31 MB on disk) | EPS/PS → PDF, PDF → EPS/PS, PostScript-based AI → PDF (images through PDF and PDFium); smaller PDFs when a preset asks for it |
 | Tesseract | **optional download** (~50 MB, 160 MB on disk; each language 0.4–10 MB when first used) | **text recognition (OCR)**: scans and pictures → searchable PDF, and their text into DOCX, TXT, Markdown and HWPX ("Recognize text" in the preset, see below) |
+| Apple Vision | part of macOS (Mac only) | **text recognition (OCR) on a Mac**, the same step as Tesseract without a download; Tesseract reads the languages Vision doesn't (see "Apple Vision on macOS") |
 | FFmpeg | **optional download** (~100 MB, 272 MB on disk) | **all video and audio**: video ↔ video (H.264, HEVC, AV1), video → animated GIF / APNG, a still frame or every frame, audio and subtitle extraction, audio ↔ audio (also APE, WavPack, TTA, DSD, DTS, ALAC, …), GIF, APNG and photos → MP4/WEBM/MOV, Live / Motion Photos → their video, a clip of a video → an iPhone Live Photo or Android Motion Photo (experimental), images → HEIC, clips or audio files joined into one, pictures → a slideshow MP4, copies without metadata (see "Media tools") |
 | Calibre | **optional download** (~216 MB, 660 MB on disk) | rare e-book formats: LIT, LRF, CHM, PDB, PML, RB, SNB, TCR, OEB → EPUB; EPUB → MOBI, AZW3, LIT, LRF, PDB, PML, RB, SNB, TCR; e-books → PDF (ranked above the HWPX writer, see below) |
 
@@ -1557,3 +1558,32 @@ scans and pictures with Tesseract (`Ocr/TesseractConverter.cs`), in the language
 - The language data is copied into the job's folder and passed with `--tessdata-dir`, so a system Tesseract reads
   only Filee's pinned files; `-c tessedit_create_pdf=1` replaces the `pdf` config file that lives in Tesseract's own
   tessdata folder.
+
+### Apple Vision on macOS
+
+On a Mac the same step runs with Apple Vision (`Ocr/VisionConverter.cs`, `VNRecognizeTextRequest` at the accurate
+level and the newest revision the Mac has), part of macOS, so nothing is downloaded. It has the same OCR edges as
+Tesseract and is registered before it, so the route planner takes it; on Windows and Linux it is unavailable
+("Not available on this platform").
+
+- The preset's language (a Tesseract code) is mapped to Vision's identifiers (`VisionLanguages`) and asked for with
+  English next to it; if Vision refuses the pair it reads with the language alone. Which languages Vision reads comes
+  from the Mac itself (`supportedRecognitionLanguagesAndReturnError:`, more with newer macOS versions); a language it
+  doesn't read (Hindi, Nepali, Greek, …) goes to Tesseract when that is installed, and is refused otherwise with
+  "Apple Vision can't read this language. Install Tesseract …". The preset editor shows no download size for the
+  languages Vision reads, and its hint names Apple Vision.
+- PDFs go page by page exactly like Tesseract's (`TesseractConverter.RecognizePdfAsync`, at most two pages at a time).
+  A JPEG picture goes into the PDF as it is, other pictures (and a JPEG that EXIF turns) as an upright PNG; every page
+  of a TIFF becomes a page. The page size comes from the picture's resolution, 300 dpi when it has none.
+- Each page is the picture with the recognized words as invisible text (`OcrPdfWriter`, render mode 3): one font
+  size per line (its height), the baseline at the line's bottom, each word stretched (`Tz`) to where the next one
+  starts with its space, so search, selection and copying follow the picture. Vision gives a box per word
+  (`boundingBoxForRange:error:`); without one the words are spread over the line by their length.
+- The text is set in a TrueType font with only empty glyphs, built in code (`GlyphlessFont`): every UTF-16 unit is
+  one character code (`Identity-H`), shown as glyph 1, and the `ToUnicode` map gives it back unchanged, so any
+  script works without a font in the file. `PdfDocumentReader` and `PagesWithText` read the layer like Tesseract's.
+- Vision is called through the Objective-C runtime (`Ocr/AppleVision.cs`), without a bridging library. A `CGRect`
+  result needs `objc_msgSend_stret` on Intel Macs; objects made with `alloc`/`init` are released, the rest go with an
+  autorelease pool per page. A cancelled job cancels the running request.
+- Tests: `VisionLanguageTests`, `OcrTextLayerTests` and `VisionConverterTests` (a stand-in recognizer) run everywhere;
+  `AppleVisionTests` read real pictures on macOS (CI's Apple silicon and Intel jobs).

src/Filee.App/Assets/i18n/ar.json

@@ -80,6 +80,7 @@
   "error.cannot_turn": "التدوير والانعكاس متاحان للصور فقط",
   "error.no_transparency": "هذا التنسيق لا يدعم الشفافية، لذا لا يوجد ما يُقص",
   "error.ocr_language_download": "تعذّر تنزيل لغة التعرّف الضوئي OCR",
+  "error.ocr_needs_tesseract": "لا يستطيع Apple Vision قراءة هذه اللغة. ثبّت Tesseract من الإعدادات ← المحركات",
   "error.too_many_frames": "عدد الإطارات كبير جدًا لحفظها كصور. زِد قيمة «كل إطار رقم N»",
   "error.too_many_pictures": "عدد الصور كبير جدًا لفيديو واحد، قسّمها على عدة فيديوهات",
   "error.not_enough_space": "لا توجد مساحة كافية على القرص لحفظ الصور",
@@ -147,6 +148,7 @@
   "engine.pandoc.description": "التحويل بين Markdown وHTML وبين ODT وRTF، ومن DOCX إلى Markdown\nالتحويل بين reStructuredText وLaTeX وOrg وAsciiDoc وTypst وMediaWiki وDocBook وJupyter وغيرها وبين Markdown وHTML وDOCX وODT\nتحويل Markdown وOrg إلى شرائح PPTX",
   "engine.ghostscript.description": "تحويل EPS وPostScript وملفات Illustrator القديمة إلى PDF وصور؛ وتحويل PDF إلى EPS وPS.",
   "engine.tesseract.description": "يتعرّف على نص الصور والمستندات الممسوحة (OCR): PDF قابل للبحث وDOCX وTXT.",
+  "engine.vision.description": "يتعرّف على نص الصور والمستندات الممسوحة بأداة التعرّف على النص المدمجة في macOS (OCR): PDF قابل للبحث وDOCX وTXT. لا حاجة إلى أي تنزيل.",
   "engine.calibre.description": "صيغ الكتب الإلكترونية النادرة (LIT وLRF وCHM وPDB وRB…) وإخراج Kindle ‏(MOBI وAZW3).\nعند وضعه فوق محرك كتابة HWPX، يُنشئ أيضًا PDF من الكتب الإلكترونية (بسرعة أكبر بكثير للكتب الطويلة).",
   "engine.libreoffice.description": "قراءة الصيغ القديمة والأقل شيوعًا: DOC وPPT وPages وKeynote وNumbers وWordPerfect وClarisWorks وPublisher وVisio…\nوالحفظ بصيغ DOC وXLS وPPT وODT وODP وقوالب OpenDocument وEMF.",
   "profile.images": "الصور",
@@ -537,6 +539,7 @@
   "presets.pdf_ghostscript_hint": "غالبًا ما يُنتج حجمًا أصغر، لكنه يعيد كتابة الملف بالكامل.\nتُحفَظ النتيجة الأصغر.",
   "presets.ocr": "التعرّف على النص (OCR)",
   "presets.ocr_hint": "يصبح نص المستندات الممسوحة والصور قابلًا للبحث والتحرير. يتطلب Tesseract (الإعدادات ← المحركات).\nتبقى الصفحات التي تحتوي على نص كما هي.",
+  "presets.ocr_hint_macos": "يصبح نص المستندات الممسوحة والصور قابلًا للبحث والتحرير بأداة التعرّف على النص المدمجة في macOS (Apple Vision). اللغات التي لا تقرؤها تتطلب Tesseract (الإعدادات ← المحركات).\nتبقى الصفحات التي تحتوي على نص كما هي.",
   "presets.ocr_language": "لغة النص",
   "presets.ocr_language_size": "تنزيل {0}",
   "presets.pdf_keep_metadata": "الاحتفاظ بالبيانات الوصفية للصفحات والصور (XMP)",

src/Filee.App/Assets/i18n/en.json

@@ -86,6 +86,7 @@
   "error.cannot_turn": "Rotate and flip work on images only",
   "error.no_transparency": "This format has no transparency, so there's nothing to trim",
   "error.ocr_language_download": "Couldn't download the OCR language",
+  "error.ocr_needs_tesseract": "Apple Vision can't read this language. Install Tesseract in Settings → Engines",
   "error.too_many_frames": "Too many frames to save as pictures, raise \"Every Nth frame\"",
   "error.too_many_pictures": "Too many pictures for one video, split them into several",
   "error.not_enough_space": "Not enough free disk space for the pictures",
@@ -154,6 +155,7 @@
   "engine.pandoc.description": "Markdown and HTML ↔ ODT and RTF, DOCX → Markdown\nreStructuredText, LaTeX, Org, AsciiDoc, Typst, MediaWiki, DocBook, Jupyter and more ↔ Markdown, HTML, DOCX and ODT\nMarkdown and Org → PPTX slides",
   "engine.ghostscript.description": "EPS, PostScript and older Illustrator files to PDF and images; PDF to EPS and PS.",
   "engine.tesseract.description": "Recognizes the text of scans and pictures (OCR): searchable PDF, DOCX, TXT.",
+  "engine.vision.description": "Recognizes the text of scans and pictures with macOS's own text recognition (OCR): searchable PDF, DOCX, TXT. Nothing to download.",
   "engine.calibre.description": "Rare e-book formats (LIT, LRF, CHM, PDB, RB…) and Kindle output (MOBI, AZW3).\nAbove the HWPX writer, it also makes the PDFs of e-books (much faster for long books).",
   "engine.libreoffice.description": "Reads older and less common formats: DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nAlso saves as DOC, XLS, PPT, ODT, ODP, OpenDocument templates and EMF.",
 
@@ -552,6 +554,7 @@
   "presets.pdf_ghostscript_hint": "Often smaller, but it rewrites the whole file.\nThe smallest result is kept.",
   "presets.ocr": "Recognize text (OCR)",
   "presets.ocr_hint": "Scans and pictures get text you can search and edit. Needs Tesseract (Settings → Engines).\nPages that already have text stay as they are.",
+  "presets.ocr_hint_macos": "Scans and pictures get text you can search and edit, with macOS's own text recognition (Apple Vision). Languages it can't read need Tesseract (Settings → Engines).\nPages that already have text stay as they are.",
   "presets.ocr_language": "Language of the text",
   "presets.ocr_language_size": "{0} download",
   "presets.pdf_keep_metadata": "Keep page and picture metadata (XMP)",

src/Filee.App/Assets/i18n/es.json

@@ -80,6 +80,7 @@
   "error.cannot_turn": "Rotar y voltear solo funcionan con imágenes",
   "error.no_transparency": "Este formato no tiene transparencia, así que no hay nada que recortar",
   "error.ocr_language_download": "No se pudo descargar el idioma de OCR",
+  "error.ocr_needs_tesseract": "Apple Vision no puede leer este idioma. Instala Tesseract en Ajustes → Motores",
   "error.too_many_frames": "Demasiados fotogramas para guardar como imágenes; aumenta «Cada N fotogramas»",
   "error.too_many_pictures": "Demasiadas imágenes para un vídeo; divídelas en varios",
   "error.not_enough_space": "Espacio libre en disco insuficiente para las imágenes",
@@ -147,6 +148,7 @@
   "engine.pandoc.description": "Markdown y HTML ↔ ODT y RTF, DOCX → Markdown\nreStructuredText, LaTeX, Org, AsciiDoc, Typst, MediaWiki, DocBook, Jupyter y más ↔ Markdown, HTML, DOCX y ODT\nMarkdown y Org → diapositivas PPTX",
   "engine.ghostscript.description": "EPS, PostScript y archivos antiguos de Illustrator a PDF e imágenes; PDF a EPS y PS.",
   "engine.tesseract.description": "Reconoce el texto de escaneos e imágenes (OCR): PDF con búsqueda, DOCX, TXT.",
+  "engine.vision.description": "Reconoce el texto de escaneos e imágenes con el reconocimiento de texto propio de macOS (OCR): PDF con búsqueda, DOCX, TXT. No hay nada que descargar.",
   "engine.calibre.description": "Formatos raros de libros electrónicos (LIT, LRF, CHM, PDB, RB…) y salida Kindle (MOBI, AZW3).\nPor encima del escritor HWPX, también crea los PDF de libros electrónicos (mucho más rápido para libros largos).",
   "engine.libreoffice.description": "Lee formatos antiguos y poco habituales: DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nTambién guarda como DOC, XLS, PPT, ODT, ODP, plantillas OpenDocument y EMF.",
   "profile.images": "Imágenes",
@@ -537,6 +539,7 @@
   "presets.pdf_ghostscript_hint": "A menudo es más pequeño, pero reescribe todo el archivo.\nSe conserva el resultado más pequeño.",
   "presets.ocr": "Reconocer texto (OCR)",
   "presets.ocr_hint": "Los escaneos e imágenes tendrán texto que se puede buscar y editar. Requiere Tesseract (Ajustes → Motores).\nLas páginas que ya tienen texto no cambian.",
+  "presets.ocr_hint_macos": "Los escaneos e imágenes tendrán texto que se puede buscar y editar, con el reconocimiento de texto propio de macOS (Apple Vision). Los idiomas que no lee requieren Tesseract (Ajustes → Motores).\nLas páginas que ya tienen texto no cambian.",
   "presets.ocr_language": "Idioma del texto",
   "presets.ocr_language_size": "Descarga de {0}",
   "presets.pdf_keep_metadata": "Conservar metadatos de página e imagen (XMP)",

src/Filee.App/Assets/i18n/fr.json

@@ -86,6 +86,7 @@
   "error.cannot_turn": "La rotation et le retournement ne s'appliquent qu'aux images",
   "error.no_transparency": "Ce format n'a pas de transparence : il n'y a rien à rogner",
   "error.ocr_language_download": "Impossible de télécharger la langue de l'OCR",
+  "error.ocr_needs_tesseract": "Apple Vision ne lit pas cette langue. Installez Tesseract dans Paramètres → Moteurs",
   "error.too_many_frames": "Trop d'images à enregistrer, augmentez « Une image sur N »",
   "error.too_many_pictures": "Trop d'images pour une seule vidéo, répartissez-les en plusieurs",
   "error.not_enough_space": "Pas assez d'espace disque libre pour les images",
@@ -154,6 +155,7 @@
   "engine.pandoc.description": "Markdown et HTML ↔ ODT et RTF, DOCX → Markdown\nreStructuredText, LaTeX, Org, AsciiDoc, Typst, MediaWiki, DocBook, Jupyter et plus ↔ Markdown, HTML, DOCX et ODT\nMarkdown et Org → diapositives PPTX",
   "engine.ghostscript.description": "EPS, PostScript et anciens fichiers Illustrator en PDF et images ; PDF en EPS et PS.",
   "engine.tesseract.description": "Reconnaît le texte des numérisations et des images (OCR) : PDF consultable, DOCX, TXT.",
+  "engine.vision.description": "Reconnaît le texte des numérisations et des images avec la reconnaissance de texte intégrée à macOS (OCR) : PDF consultable, DOCX, TXT. Rien à télécharger.",
   "engine.calibre.description": "Formats de livres numériques rares (LIT, LRF, CHM, PDB, RB…) et sortie Kindle (MOBI, AZW3).\nPlacé au-dessus du rédacteur HWPX, il produit aussi les PDF des livres numériques (bien plus vite pour les longs livres).",
   "engine.libreoffice.description": "Lit les formats anciens et moins courants : DOC, PPT, Pages, Keynote, Numbers, WordPerfect, ClarisWorks, Publisher, Visio…\nEnregistre aussi en DOC, XLS, PPT, ODT, ODP, modèles OpenDocument et EMF.",
 
@@ -552,6 +554,7 @@
   "presets.pdf_ghostscript_hint": "Souvent plus léger, mais il réécrit tout le fichier.\nLe résultat le plus léger est conservé.",
   "presets.ocr": "Reconnaître le texte (OCR)",
   "presets.ocr_hint": "Le texte des numérisations et des images devient consultable et modifiable. Nécessite Tesseract (Paramètres → Moteurs).\nLes pages qui ont déjà du texte restent telles quelles.",
+  "presets.ocr_hint_macos": "Le texte des numérisations et des images devient consultable et modifiable grâce à la reconnaissance de texte intégrée à macOS (Apple Vision). Les langues qu'elle ne lit pas nécessitent Tesseract (Paramètres → Moteurs).\nLes pages qui ont déjà du texte restent telles quelles.",
   "presets.ocr_language": "Langue du texte",
   "presets.ocr_language_size": "Téléchargement de {0}",
   "presets.pdf_keep_metadata": "Conserver les métadonnées des pages et des images (XMP)",