ToolPrivasi sejak halaman pertama
File, gambar kerja, teks, dan PDF hasil tetap di perangkat Anda. Hanya mesin dan model bahasa pilihan yang diunduh dari RecodeX.
WORDS WAITING TO BE FOUND.
PDF OCR.
Make scanned PDFs searchable, selectable, and easy to copy.
100% Private — OCR runs on your device.Free · No file-size quota
Scan yang bisa dicari
OCR Indonesia dan Inggris membaca isi gambar sebenarnya. Lapisan teks baru ditempatkan pada halaman asli tanpa kompresi ulang gambar.
Kendali pada setiap halaman
Pilih halaman, bahasa, resolusi, dan peningkatan gambar kerja. Preview teks dan confidence membantu memeriksa hasil; pengenalan teks tetap perlu ditinjau.
Cara menggunakan PDF OCR
PDF OCR membaca huruf pada gambar scan dengan model bahasa lokal. Pilih halaman dan bahasa, periksa teks yang dikenali, lalu simpan PDF dengan lapisan teks atau unduh TXT. Indonesia dan Inggris dapat dipakai bersama; model tambahan dimuat hanya saat dipilih.
- Select your scanned PDF. Halaman dan pengaturan dapat langsung dipakai sementara mesin PDF dan model bahasa dimuat di background.
- Choose Indonesian, English, or a supported additional language. Pilih semua halaman, halaman tertentu, atau rentang. Dokumen dengan teks yang sudah dapat dipilih dilewati secara default.
- Start OCR. Gambar kerja dirender sesuai DPI dan batas memori. Enhanced OCR menyediakan grayscale, contrast, smoothing, dan threshold pada gambar kerja; gambar PDF sumber tidak diganti.
- Review the actual recognized words, confidence, and flagged pages. Unduh Searchable PDF, TXT, atau JSON sesuai pilihan. Periksa nama, angka, tabel, dan informasi penting terhadap scan asli.
Pertanyaan tentang PDF OCR
What is PDF OCR?
Optical character recognition reads text from images. A scanned PDF often contains only pictures of pages; OCR can add an invisible text layer so recognized words can be searched, selected, and copied. Existing selectable text can be extracted without recognition.
Can OCR read scanned PDFs?
Yes. This tool renders selected pages locally and runs Tesseract recognition on the actual images. Clean, upright scans with sufficient resolution usually work better. Handwriting, complex tables, unusual fonts, low contrast, and severe skew may require review.
Can I copy text after OCR?
Yes. Review recognized text per page, copy it, or download UTF-8 TXT. Searchable PDF output preserves source page appearance and adds real invisible text at recognized word positions. It does not create an editable Word document or guarantee table reconstruction.
Does OCR work with Indonesian and English?
Yes. Indonesian + English is the default language preset. Single-language Indonesian and English are also available, with French, Spanish, and German as additional installed packs. Only selected packs are downloaded from RecodeX. The preset does not claim automatic language identification.
Can I OCR selected pages?
Yes. Select actual thumbnails, use odd/even pages, or enter a custom range such as 1-5, 8. Only those pages are recognized; other PDF pages are retained unchanged and may remain image-only.
How can I improve OCR accuracy?
Choose the correct language and page segmentation, use an upright working image, and try 300 DPI when device memory allows. Filters affect only recognition images. Larger pages may be rendered below the requested DPI to stay within the pixel budget; the report shows actual DPI. Confidence is not a guarantee of accuracy.
Is my PDF uploaded?
No. Reading, rendering, OCR, text positioning, validation, preview, and downloads happen in your browser. Only fixed engine assets and selected language packs download from RecodeX. Documents, filenames, text, and metadata are not sent to servers or analytics. Processing can continue with loaded dependencies in the open page; a first offline visit is not promised.
What are the limitations of searchable PDF output?
Source pages are copied without rasterizing or recompressing their images. Supported links, annotations, and bookmarks can be preserved. Forms require flattening consent. Signature validity, accessibility tags, document attachments, and XMP cannot be preserved by the page-copy pipeline; detected features require confirmation. Encrypted, XFA, and unsupported layered PDFs are rejected. Auto deskew/orientation and editable table reconstruction are visibly unavailable.