App reference
OCR reference: scanned PDFs and images
26.10.1001
Novaber · Content reviewed
On this page
Extract editable text and tables from images and scanned PDFs with local batch OCR. Review against originals with PaddleOCR, GLM-OCR, or OvisOCR2, then export Word, Excel, Markdown, TXT, or searchable PDF for document editing and archiving.
Capabilities and input / output
-
Batch recognition for images and PDFs
-
PaddleOCR, GLM-OCR, and OvisOCR2 model support
-
Text, layout, and table-structure extraction
-
Side-by-side review and export to five formats
Workflow
1. Input
Import scans, screenshots, or PDFs in batches and select pages
2. Recognition
Choose a recognition model to extract text and document structure
3. Review
Review text and table cells alongside the original pages
4. Output
Export editable text, documents, spreadsheets, and searchable PDF
Requirements
- GPU: NVIDIA GPU required (CUDA 12.6)
- VRAM: VRAM varies by model and page complexity
- Images and models: Backend/worker ~18.1 GB unpacked each; models extra
Review your first result
These checks are suggestions based on published inputs and outputs. Use a small sample to verify processing, preview and export together; no measured accuracy or speed is provided here.
Start with a few pages and confirm the model, page range and original pages. Review text, table cells and layout, then reopen the export in the relevant software. OCR needs human review; support for an export format does not guarantee exact reproduction of every complex layout.
Version, limitations and license
Current catalog version: 26.10.1001.
Third-party names and trademarks belong to their respective owners. Use and distribution are governed by each project's license and do not imply endorsement of Novaber.
Product details and installation
Choose an export by its next use
| Next use | Available format | Review focus |
|---|---|---|
| Edit body text | Word / DOCX, TXT or Markdown | Text, paragraphs, reading order and completeness |
| Work with table data | Excel / XLSX | Rows, columns, cells, numbers and headers |
| Archive scanned pages with text search | Searchable PDF | Original-page correspondence and actual search results |
Choose text editing, tabular work or searchable archiving before exporting. Format support does not promise exact complex-layout reproduction or automatic correction of recognition errors.
From source pages to usable documents
1. Import and verify the page range
Check source, page numbers and order after importing images or PDFs. Select the required pages and try a few representative text, table or complex-layout pages before enlarging the batch. Keep originals for comparison.
2. Select and prepare the model
The catalog lists PaddleOCR, GLM-OCR and OvisOCR2. Choose for the task and confirm service/model readiness. VRAM depends on model and page complexity; no universal figure applies to every page, and file count alone does not establish suitability.
3. Review against originals
Verify page correspondence. Check numbers, names, paragraphs and reading order in text; inspect headers, row/column structure and cells in tables. Record erroneous pages and the model. Completion alone is insufficient for delivery.
4. Reopen the export
Open Word, Excel, text or PDF in the intended software. Confirm readability, page range and content; search a known passage in searchable PDF. See OCR questions for recognition and layout issues.
Prepare the environment
For NVIDIA GPU workflows, check driver major >= 575 and Windows driver >= 576.57 for the project’s CUDA 12.9.1 baseline. See the driver and CUDA baseline for version comparison, unknown results and container checks. Check the selected app’s GPU model, VRAM and model requirements separately.