Restore an Urdu scan without pretending pixels can invent missing ink
This page uses a Nastaliq-aware cleanup, then the same SeedVR2 upscaler. If the source is already out of focus, we extract the text for print instead of selling a larger blur.
This is not a language detector
Nastaliq hairlines that are already out of focus cannot become stamp-sharp. We restore remaining ink and try to extract the Urdu for print.
Start at 2K. 4K will not invent missing Nastaliq strokes.
Preview
Source scan
Extracted Urdu for print
Copy this text if the image is still too soft to print. Empty or unavailable means the source did not leave enough ink.
Text appears after a restore starts.
Newsletter
Join the community
Subscribe to our newsletter for the latest news and updates
Built for Salman's English-versus-Urdu test
A generic photo upscaler treats letters as texture. Nastaliq is thinner, connected, and loses dots first.
Print is a different job
If the Urdu result still looks out of focus at 2K, a 4K run will not invent ink. Use the extracted text for a print PDF.
Nastaliq-aware cleanup
The document pass skips the Latin median filter that erases 1-pixel dots, then sharpens grayscale ink before upscaling.
Extract text for print
When strokes are gone, we transcribe the Urdu so you can copy and re-typeset. Transcription can fail on a ruined source.
Same image credits
2K, 4K, and 8K use the existing SeedVR2 image prices. Text extraction is included and does not add a second charge.
How to get a usable Urdu page
Capture quality still beats any model. Start with a flat scan when you can.
- Step 1
Scan at 400 DPI if you can
Use a flatbed or Adobe Scan. Phone photos of notices are often already too soft for print.
- Step 2
Upload the same source
JPG, PNG, or WebP up to 20 MB. Sign in before the paid run.
- Step 3
Run the Nastaliq restore
Start at 2K. Inspect the strokes at 100 percent zoom before spending on 4K.
- Step 4
Copy the text if the image stays soft
Use the transcribed Urdu for a print document. The image remains a scan replica, not a new typeset page.
Before you spend credits
SeedVR2 does not read English or Urdu. It sharpens ink that is still in the pixels. English letters are thicker, so the same blur leaves more to recover.
A 4K or 8K run enlarges whatever remains. It cannot reconstruct Nastaliq hairlines that the camera already lost.
There is no native Android app yet. Capture with Adobe Scan, then open this page in the phone browser.
Failed Urdu tests that stay unreadable are a support credit case, not a reason to rerun 4K.
Urdu document FAQ
Short answers that match the English-versus-Urdu side-by-side test.
Why did English look print-ready and Urdu still look out of focus?
The upscaler does not recognize languages. English glyphs are thicker and separated. Urdu Nastaliq is thinner and connected. Once those strokes are gone, enlarging the image only enlarges the blur.
Can this make a blurry Urdu notice suitable for printing?
Only if enough ink remains in the scan. If 2K is still soft, copy the extracted text and re-typeset. That is the print path, not another upscale.
Is there an Android app?
Not yet. This page works in the mobile browser. For capture quality, use a scanning app such as Adobe Scan, then upload here.
What if the Urdu result is still not printable?
Email support with the job. We can restore the credits for that Urdu test instead of charging another 4K run.
Related workflows
Use the general upscaler for English photos. Use this page for Urdu scans.