Extract text from images and PDFs with on-device OCR

OCR sites ask you to upload a passport scan or a whiteboard photo. The text you want is often the most sensitive part: names, amounts, medical print.

TextLens extracts text from images and PDFs with on-device OCR. The file stays in the tab. It lives on the document host, with signatures and certificates, not on the file-converter hub.

What “on-device OCR” means

The page may download a public model on first use. That download is not your scan. After the model is cached, recognition runs locally. A large PDF can be slow on a phone. That is expected.

Client-side means we do not run an OCR farm that stores pages “for quality.”

How to extract text

1. Open https://docs.sw-commercial.com/ocr.
2. Pick an image or PDF from your computer.
3. Wait for the text. Check it. OCR misreads numbers; you are the proofreader.
4. Copy or download. Close the tab if the machine is shared.

You do not create an account. You do not buy “pages per month” from us for the basic job.

Related

To merge the original scans, use Tool Hub. To hide text permanently in a PDF, use Redact — OCR is not redaction. To resume from a paper CV, paste into Rezimo after you check the text.

What we do not do

We do not store the scan. Ads, if shown, are labeled. PlayNest is not an OCR toy for homework photos of other children; keep school photos off random upload OCR.

The copy you take away is text you checked. There is no page pack to buy before the first page runs.

A worked example

Photograph a printed receipt in daylight. Run OCR. Check the total. If a 6 became a 5, you caught it. That is why the human stays in the loop.

Passports and bank letters are the worst uploads on the public web. Local OCR is the point. Still: a shared computer can keep the last download in the Downloads folder. Delete when you are done.

Language and layout

Multi-column magazines fail more than a single column form. Tables fail more than paragraphs. We do not promise a perfect pass on a handwritten doctor’s note.

A phone photo of a form works better when the page is flat, the light is even, and the camera is square to the paper. A glare stripe across a total is a miss you will not see until you read the line.

When to skip this tool

Skip it if you need certified translation. Skip it if you need a human transcriber. TextLens is machine OCR in the tab.

Merge the original images on Tool Hub if you need one PDF packet after you copied the text out.

FAQ

Do you upload the scan?

No. TextLens runs in your browser. We do not run an upload API for this job. If a model file downloads on first use, that is the recognition program, not your document.

Do I need an account?

No. Open the page and use it. Free use does not start with an email wall. A Pro key on the Chrome extension is not required to try TextLens.

What happens if I clear the site data?

The cached model may download again next time. Text you already copied stays wherever you pasted it. The scan you still have on disk is unchanged. Delete downloads on a shared computer when you are done.

Can I use this on a phone?

Yes, if the browser can pick a photo. A multi-page PDF is slower on a phone because the tab is doing the recognition. That is your device, not a queue on our side.

Do ads see the scan?

Labeled ads, if shown, do not receive the file. The SW Toolbox extension has 0 ads. TextLens is a page on docs.sw-commercial.com.

Handwriting?

Unreliable. Print and daylight work better. Read every amount before you trust it in a spreadsheet.

Certified translation?

No. This is OCR, not a translator. It does not turn a foreign form into a certified English copy.

After you have the text

Paste it where you needed it, then read the numbers again against the page. OCR is a draft of the print, not a second original. If a digit matters (a total, an account, a dose), compare it to the photo before you send the text on.

Close the tab when you finish on a shared computer. The privacy policy at sw-commercial.com/privacy is the full legal text. This guide is only TextLens.