From Newsgroup: alt.comp.os.windows-11

I'm running Win 11 PRO on a desk top with a Canon 9000F Mark II scanner.
Canon provides an interface between the machines. I need to do an OCR
chore of about 15 computer printed pages of text that's not too
complicated - a few bullet numbered paragraphs. I'm looking for some OCR software to scan the document into editable text and I'll polish it
using either LibreOffice or LaTeX. I need a suggestion of a source of
either some free or not too expensive OCR software. I'm looking for
something robust in the sense that it doesn't blow multiple lines
because it can't tell a 1 from an I.

Canon used to provide OCR with this scanner. I haven't used it for
months but it's now disappeared from my machine along with Image Garden
that was the "home" for their OCR. I downloaded the current version and
it doesn't have OCR. It seems they snuck in an "update" that just
uninstalled the program. The original was one of the poorest excuses for
user software I've ever seen; it's only saving grace was its OCR.
--
Jeff Barnett

--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

Jeff Barnett wrote:

I need to do an OCR chore of about 15 computer printed pages of text
that's not too complicated

If you don't mind using an online service, there's Google Docs, but be
wary if the docs are private, I think OneNote can do it but have never
used it.
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On 2026-05-18 07:50, Andy Burns wrote:

Jeff Barnett wrote:

I need to do an OCR chore of about 15 computer printed pages of text
that's not too complicated

If you don't mind using an online service, there's Google Docs, but be
wary if the docs are private, I think OneNote can do it but have never
used it.

IME of scanners, often there's some OCR software supplied on a CD with
the machine or packaged up with the driver download. Was there really
nothing supplied with it that can do this? It might be implicitly
invoked amongst the output choices, for example:

* Scan to image
* Scan to PDF
* Scan to text <-- this one

Failing that, if you can scan to PDF, then many Acrobat alternatives,
perhaps even Acrobat itself, can OCR a document to give it searchable
text, and then you could perhaps get hold of the text layer separately.

Failing that as well, remembering Andy's warning about privacy above,
(there is a link to their privacy policy in the page footer; basically
they say they collect hardly anything, but check it out for yourself),
I've found this online service useful in the past:

https://www.onlineocr.net/

As long as you have only a few pages to do and each individual document
is less than 15MB, then you can use the service for free.
--

Fake news kills!

I may be contacted via the contact address given on my website: www.macfh.co.uk

--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On 18/05/2026 07:16, Jeff Barnett wrote:

I'm running Win 11 PRO on a desk top with a Canon 9000F Mark II scanner. Canon provides an interface between the machines. I need to do an OCR
chore of about 15 computer printed pages of text that's not too
complicated - a few bullet numbered paragraphs. I'm looking for some OCR software to scan the document into editable text and I'll polish it
using either LibreOffice or LaTeX. I need a suggestion of a source of
either some free or not too expensive OCR software. I'm looking for something robust in the sense that it doesn't blow multiple lines
because it can't tell a 1 from an I.

Canon used to provide OCR with this scanner. I haven't used it for
months but it's now disappeared from my machine along with Image Garden
that was the "home" for their OCR. I downloaded the current version and
it doesn't have OCR. It seems they snuck in an "update" that just uninstalled the program. The original was one of the poorest excuses for user software I've ever seen; it's only saving grace was its OCR.

There are plenty of freeware programs which do what you want. I use two
with OCR plugins: Irfanview for pdf or images, PDF XChange Viewer for
pdf. Both are accurate given reasonable quality text images to work with.
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On Mon, 18 May 2026 01:16:25 -0500, Jeff Barnett <jbb@notatt.com> wrote:

I'm running Win 11 PRO on a desk top with a Canon 9000F Mark II scanner. Canon provides an interface between the machines. I need to do an OCR
chore of about 15 computer printed pages of text that's not too
complicated - a few bullet numbered paragraphs. I'm looking for some OCR software to scan the document into editable text and I'll polish it
using either LibreOffice or LaTeX. I need a suggestion of a source of
either some free or not too expensive OCR software. I'm looking for
something robust in the sense that it doesn't blow multiple lines
because it can't tell a 1 from an I.

Canon used to provide OCR with this scanner. I haven't used it for
months but it's now disappeared from my machine along with Image Garden
that was the "home" for their OCR. I downloaded the current version and
it doesn't have OCR. It seems they snuck in an "update" that just
uninstalled the program. The original was one of the poorest excuses for
user software I've ever seen; it's only saving grace was its OCR.

I'm not experienced with this type of software, but I found out about Text Grab on the Windows 10 newsgroup a couple of years ago.
https://textgrab.net/

The freeware version can be downloaded from GitHub at https://github.com/TheJoeFin/Text-Grab/releases/latest

It doesn't include everything that the Microsoft Store version does.
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

In article <n6vr63FtnflU1@mid.individual.net>, usenet@andyburns.uk
says...

Jeff Barnett wrote:

I need to do an OCR chore of about 15 computer printed pages of text
that's not too complicated

If you don't mind using an online service, there's Google Docs, but be
wary if the docs are private, I think OneNote can do it but have never
used it.

Yes, OneNote can do OCR. Typically used for extracting text from an
image on a page, rather than in bulk.
--
--
Phil, London
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On 5/18/2026 8:16 AM, Jeff Barnett wrote:

I need to do an OCR
chore of about 15 computer printed pages of text that's not too
complicated

Maybe it is easier to use the Snipping Tool instead of installing a
software.

Scan the pages as pictures, right click on a picture -> open with ->
Snipping Tool. Then click on the OCR button.
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

Jeff Barnett <jbb@notatt.com> wrote:

As Herbert suggested, I also advise to use the bundled Windows 11
'Snipping Tool' to do the job.

That said, back to this:

I'm running Win 11 PRO on a desk top with a Canon 9000F Mark II scanner.

[...]

Canon used to provide OCR with this scanner. I haven't used it for
months but it's now disappeared from my machine along with Image Garden
that was the "home" for their OCR. I downloaded the current version and
it doesn't have OCR. It seems they snuck in an "update" that just uninstalled the program. The original was one of the poorest excuses for user software I've ever seen; it's only saving grace was its OCR.

Can't you just re-install the OCR part from the supplied 'Canon
CanoScan 9000F Mark II Setup CD-ROM'?

If the "update" keeps removing it, perhaps you can install it in a non-standard folder.
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On 18 May 2026, Andy Burns <usenet@andyburns.uk> wrote in alt.comp.os.windows-11:

If you don't mind using an online service, there's Google Docs,
but be wary if the docs are private, I think OneNote can do it but
have never used it.

You're right about the privacy concern, but otherwise, Google Docs does
a *FAR* better job at OCR than any OCR program I've ever used. I
supposed it's an AI thing.
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On Mon, 5/18/2026 3:14 PM, Nil wrote:

On 18 May 2026, Andy Burns <usenet@andyburns.uk> wrote in alt.comp.os.windows-11:

If you don't mind using an online service, there's Google Docs,
but be wary if the docs are private, I think OneNote can do it but
have never used it.

You're right about the privacy concern, but otherwise, Google Docs does
a *FAR* better job at OCR than any OCR program I've ever used. I
supposed it's an AI thing.

There is the Tesseract level of OCR, with the 0/o/O problem.

What LLM-AI potentially adds to that, is the ability to use
the grammar and syntax to aid conversion and decide which of
multiple probabilities makes sense.

And they were doing that before, before LLM-AI, using regular
procedural code (using syntax and grammer to buttress bad conversion).
The LLM-AI just has a larger database to work with.

The most impressive demo I've had of this so far, involved
UEFI Secure Boot messages on an LCD screen. The Break/Pause key
does not work to halt the screen when a Secure Boot error appears.
Only shooting video of it, offers a chance to take note of
some PCR issue.

So I shoot the video, my camera was on an angle, so pitch and
yaw and so on, not perfect. The contrast ratio was not good.
The OCR of that message ? Perfect. Even though no regular
(Tesseract style) conversion would have coughed up anything.
You could not follow the edge of the characters when the
image is degraded that badly. But the PCR issue, is something
the LLM-AI has seen as text before, so it can take a stab
at the message. And visually comparing the picture to the
OCR, it was perfect.

There is a claim that some of the LLM-AI can do OCR as
a native function. What's unclear, is how a person figures
out what input modes exist. The LLM-AI "needs to be a little
bit agentic" to be instructed to grab a file from the
file system, if you have an OCR task in a PNG for it.
I have seen some screenshots of LLM-AI that no longer
blow up when you ask then "what are your capabilities?" :-)
Previously, that was not a question you could ask.

If you "scan to PDF", using mutool (mupdf), you should be able
to extract the pixmap from each page if you want to
process it as a PNG or JPG.

But as for being easy to do, the SnippingTool function
(which accepts an image file as input), at least the
interface works. The processing of the columnar
appearance of a document, is pretty basic and needs work.
You either get a "reasonable" conversion, or you end up
with words "sprayed all over the place", depending on
the white-space pattern of the doc page. It still cannot
"process a table" properly. I have artificially created
tables for it, to "help it", and it made no difference
to the botched output. So it's "not an input quality" problem.

Paul
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: alt.comp.os.windows-11

On Mon, 5/18/2026 2:16 AM, Jeff Barnett wrote:

I'm running Win 11 PRO on a desk top with a Canon 9000F Mark II scanner. Canon provides an interface between the machines. I need to do an OCR chore of about 15 computer printed pages of text that's not too complicated - a few bullet numbered paragraphs. I'm looking for some OCR software to scan the document into editable text and I'll polish it using either LibreOffice or LaTeX. I need a suggestion of a source of either some free or not too expensive OCR software. I'm looking for something robust in the sense that it doesn't blow multiple lines because it can't tell a 1 from an I.

Canon used to provide OCR with this scanner. I haven't used it for months but it's now disappeared from my machine along with Image Garden that was the "home" for their OCR. I downloaded the current version and it doesn't have OCR. It seems they snuck in an "update" that just uninstalled the program. The original was one of the poorest excuses for user software I've ever seen; it's only saving grace was its OCR.

I finally found a later Tesseract, but it doesn't
seem to work any better than the other ones I've got.

*******

You can tick the two boxes for "scripts" and tessdata (language-specific training).

https://digi.bib.uni-mannheim.de/tesseract/

tesseract-ocr-w32-setup-v5.3.0.20221214.exe 2022-12-14 18:57 50M

tesseract-ocr-w64-setup-v5.3.0.20221214.exe 2022-12-14 19:29 51M

D:\mupdf-1.23.0-windows-tesseract> .\mutool extract recognition.pdf # 4 page document, we will use page 2 of it.
extracting image-0006.png
extracting image-0009.png <=== could be page 2... Transfer to VM Downloads directory (tesseract being tested in VM)
extracting image-0012.jpg
extracting image-0015.jpg
PS D:\mupdf-1.23.0-windows-tesseract> <=== (The Tesseract doesn't work in it, so the name is mis-leading)

C:\Users\bullwinkle\Downloads>"C:\Program Files\Tesseract-OCR\tesseract.exe" recognition.pdf recogout
Error in pixReadStream: Pdf reading is not supported
Error in pixRead: pix not read
Error during processing.

C:\Users\bullwinkle\Downloads>"C:\Program Files\Tesseract-OCR\tesseract.exe" image-0009.png recogout # It's not in %path%...

05/19/2026 12:46 AM 1,350 recogout.txt # The "usual Tesseract quality"
# Columnar format not maintained.
# Tables will need realignment.
# This is page 2.

You can play with the command, so it does all the pages in one run.
I didn't test that.

Paul
--- Synchronet 3.21d-Linux NewsLink 1.2