Scanned and image-only CVs
Send a scanned CV exactly like any other PDF. VerbatimCV notices pages with no usable text, runs OCR on them, and tells you in the result when it did and how reliable it was. You don't need a separate OCR step.
What counts as "no usable text"
- Scans and photos saved as PDF: no text layer at all (
document.text_layer: "absent"). - Exports from design tools where text became shapes, or extracts as garbage because of a broken font encoding (
"corrupt").
What happens
- OCR runs with Tesseract on our server, choosing languages from the script on the page (Latin, Cyrillic, Greek, Arabic, Hebrew, Devanagari, Chinese, Japanese, Korean).
- If the OCR text is still unreliable, an image of the page is sent to the model alongside it.
- Low word confidence adds the warning
OCR_LOW_CONFIDENCE, anddocument.ocr_appliedtells you OCR was used.
Limits
Inputs must be PDF or DOCX; image files (JPG, PNG) and legacy .doc files aren't accepted
directly, so wrap a photo in a PDF first. Scans take longer than text PDFs, typically 10–25 seconds.
Allow at least 60 seconds before timing out.
Try it
Get a free API key (50 parses a month, no card), then:
curl https://api.verbatimcv.com/v1/parse \
-H "Authorization: Bearer $VERBATIMCV_API_KEY" \
-F file=@cv.pdf
Agents can use the parse_resume tool instead: uvx verbatimcv-mcp. See the
quickstart for Python, JavaScript and MCP setup.