Original-preserving optimization
Images up to a bounded 25 MB can be resized into a maximum-3200px working copy. Original bytes, optimized derivatives and their hashes remain separately linked.
DOCUMENT PROCESSING
Document upload, optimization and bilingual OCR assistance.
Normal high-resolution documents are checked, previewed and stored as encrypted originals. A separate optimized working copy can support processing while the original evidence remains unchanged and accessible only to authorized users.
Capabilities built around the actual customer journey.
Implementation and testing status is stated separately from commercial activation, so your team can evaluate what is genuinely available.
Local non-generative OCR
Tesseract runs locally with English and Urdu trained data. No paid generative AI or per-image AI processing is used for this path.
Confidence and manual correction
Extracted fields, original-script text, confidence, source hashes and correction status are returned for review. Missing or uncertain fields require attention.
Authorized OCR search
Customers can search their own extracted text, and explicitly authorized Admin reviewers can search within their permission scope. Other tenants’ text is not exposed.
Where this can be used.
- Identity document extraction
- English and Urdu text review
- High-resolution document upload
- Manual review and correction
- Secure original evidence retrieval
A controlled implementation path.
- 1Select and preview the document
- 2Optimize a working copy without replacing the original
- 3Run local English/Urdu OCR
- 4Correct low-confidence or missing fields
- 5Store and retrieve only through authorized access
Current availability and acceptance limits
Real local English OCR extraction has passed a high-resolution end-to-end test. Urdu script was returned in testing but field-level accuracy was insufficient for acceptance, so Urdu fields remain review-required. OCR never marks a document verified by itself.
Frequently asked questions
Is the original document overwritten?
No. The original is preserved in encrypted storage and an optimized derivative is stored separately with its own reference and hash.
Is Urdu OCR fully accepted?
Not yet. Urdu trained data is enabled, but current synthetic field-level evidence is insufficient. Keep manual correction and review enabled.
Are uploads unlimited?
No. Security limits remain explicit: the current original image limit is 25 MB, with format, dimension and post-processing checks. This is not advertised as unlimited.
Evaluate the exact workflow before you go live.
Review the documentation, test the actual document and provider path, or ask our integration team for a scoped technical assessment.