I want to help an older person with severe macular degeneration from
thousands of miles afar where she could use an AUDIO file of long text
(like of a book or short story, normally, as in Great Books stories).

I'm on Windows 10 and Android 12. She's on the iPhone and iPads.

I can easily save a web page (like a news story) to text and sometimes I
can convert a PDF of a book to text (depending on how the PDF was made).

But how can I turn the text into a "podcast" for her to listen to?
And would it be too large to email to her in a typical email message?

Peter <occassionally-confused@nospam.co.uk> wrote:
> I want to help an older person with severe macular degeneration from
> thousands of miles afar where she could use an AUDIO file of long text
> (like of a book or short story, normally, as in Great Books stories).
>
> I'm on Windows 10 and Android 12. She's on the iPhone and iPads.
>
> I can easily save a web page (like a news story) to text and sometimes I
> can convert a PDF of a book to text (depending on how the PDF was made).
>
> But how can I turn the text into a "podcast" for her to listen to?
> And would it be too large to email to her in a typical email message?

iOS has very good accessibility features. I suspect it already can read any
pdf you provide.

https://support.apple.com/en-gb/guide/iphone/iph3e2e415f/ios

On 6/12/2023 9:13 AM, Peter wrote:
> I want to help an older person with severe macular degeneration from
> thousands of miles afar where she could use an AUDIO file of long text
> (like of a book or short story, normally, as in Great Books stories).
>
> I'm on Windows 10 and Android 12. She's on the iPhone and iPads.
>
> I can easily save a web page (like a news story) to text and sometimes I
> can convert a PDF of a book to text (depending on how the PDF was made).
>
> But how can I turn the text into a "podcast" for her to listen to?
> And would it be too large to email to her in a typical email message?
>

You will need:

1) Copy of Audacity or a similar audio recorder.
2) Male to male audio cable. Connect Line Out to Line In.
This step, is to avoid fighting with "What You Hear" setup,
which could waste half the afternoon.
3) Run Firefox.
4) In Windows Mixer, set Windows Sounds (beeps and boops) to zero.
This is to avoid the Windows Defender toast notification or a
You Have Mail ding, from ruining your recording.

Now, use Firefox Reader View, open your text file.

[Picture]

https://i.postimg.cc/DzhXyywq/firefox-reader-view-TTS.gif

After listening to a half dozen other TTS implementations
that suck, that one is almost bearable.

With some programming, as Newyana2 describes, you can do better
than that, but then it's a real long project. This method
may not be efficient, but you also don't have to be a programmer.

The Firefox page should auto-scroll, as it plays the text. I do not
know how long a web page (or text file) can be, and whether your attempt to
convert War & Peace to sound, is going to work.

I may have recorded for 16 hours before with Audacity, but that
was on the other, smaller machine. Make sure all the steps of
recording and saving are working on Audacity, before your run.
You would not want to generate all those tiny sound files,
and have it croak before saving.

I think Windows also has an "improved" sound recorder, which
you might want to research. That's if you can't get Audacity
to work (which... happens sometimes). The namespace for audio
is a bitch. An "improved" sound recorder, is one for which
you don't have to "define a fixed size audio file" before
you begin recording. Some of the Windows demo apps have
been absurd that way. The damn thing should just record,
and then you do the Save As, and done. Check that it works
that way.

Paul

On 06/12/2023 9:13 AM, Peter wrote:
> I want to help an older person with severe macular degeneration from
> thousands of miles afar where she could use an AUDIO file of long text
> (like of a book or short story, normally, as in Great Books stories).
>
> I'm on Windows 10 and Android 12. She's on the iPhone and iPads.
>
> I can easily save a web page (like a news story) to text and sometimes I
> can convert a PDF of a book to text (depending on how the PDF was made).
>
> But how can I turn the text into a "podcast" for her to listen to?
> And would it be too large to email to her in a typical email message?
This is a feature of Adobe Reader that I have never used but know it
exist.

Open the document that you want in the Adobe Reader (including the Free
version.) In View, Select "Read Out Aloud". Activate the Read Aloud
function and select the option you want.

There should be a way to capture the verbal document to an audio file.
Possibly by using the record feature of the Cellphone.

I may not be the best way, but it could be used to convert the write
document to an audio documents. May take some Googling.

Chris <ithinkiam@gmail.com> wrote:
>> But how can I turn the text into a "podcast" for her to listen to?
>> And would it be too large to email to her in a typical email message?
>
> iOS has very good accessibility features. I suspect it already can read any
> pdf you provide.

Thanks for that suggestion where the main goal is one way for everything.
She will get super confused if there isn't one way for everything.

Not one way for this kind of PDF. Another for that kind of PDF.
And yet another for text. And another for HTML. And so on.

It has to be ONE way for EVERYTHING, where there's no way any program, no
matter how smart it is, can read out ALL PDFs (since many are bitmaps).

For the bitmaps, I was thinking of looking up OCR or, if it's short, just
reading it aloud to myself and recording it on my phone's voice recorder.

> https://support.apple.com/en-gb/guide/iphone/iph3e2e415f/ios

Thanks for that suggestion since she only has iOS iPhones & iPads.
And I don't.

I looked at that page which seems to be much like what Windows does.
Which means it will drive any normal person nuts with extraneous talking.

It has to be an audio of the text only and nothing else for it to work.
The firefox text-only speaking suggestion from Paul is along those lines.
--
Removed android and replaced it with Firefox since the solution is there.

Paul <nospam@needed.invalid> wrote:
> You will need:
>
> 1) Copy of Audacity or a similar audio recorder.

Absolutely no problem. I have Audacity on both Android & Windows.

> 2) Male to male audio cable. Connect Line Out to Line In.
> This step, is to avoid fighting with "What You Hear" setup,
> which could waste half the afternoon.

I can make one of those.

> 3) Run Firefox.

I can run that.

> 4) In Windows Mixer, set Windows Sounds (beeps and boops) to zero.
> This is to avoid the Windows Defender toast notification or a
> You Have Mail ding, from ruining your recording.

Windows Mixer? https://www.google.com/search?q=what+is+windows+mixer

OK. It's explained here for both Windows 10 and Windows 11.
https://www.asus.com/support/FAQ/1044702/
>
> Now, use Firefox Reader View, open your text file.
>
> [Picture]
>
> https://i.postimg.cc/DzhXyywq/firefox-reader-view-TTS.gif

Firefox reader view?
https://www.google.com/search?q=what+is+firefox+reader+view

OK. It's explained here.
https://support.mozilla.org/en-US/kb/firefox-reader-view-clutter-free-web-page
"Reader View is a Firefox feature that strips away clutter like buttons,
ads, background images, and videos while also letting you customize the
layout and theme to fit your reading preferences."
> After listening to a half dozen other TTS implementations
> that suck, that one is almost bearable.

Whatever Windows sound plays when I am in MerriamWebster seems fine to me.
https://www.merriam-webster.com/dictionary/otorhinolaryngologist

> With some programming, as Newyana2 describes, you can do better
> than that, but then it's a real long project. This method
> may not be efficient, but you also don't have to be a programmer.

I don't want to be a programmer since I can already convert most "things"
to text so that's why I wanted to turn text into something she can play.

> The Firefox page should auto-scroll, as it plays the text. I do not
> know how long a web page (or text file) can be, and whether your attempt to
> convert War & Peace to sound, is going to work.

Most things are about twenty pages as I've been converting them to text
already for her where she selects the text inside an email and plays it.

> I may have recorded for 16 hours before with Audacity, but that
> was on the other, smaller machine. Make sure all the steps of
> recording and saving are working on Audacity, before your run.
> You would not want to generate all those tiny sound files,
> and have it croak before saving.

I'm familiar with audacity. It can output MP3 which I assume is the
standard file format for all computers as it is for my phone & PC.
> I think Windows also has an "improved" sound recorder, which
> you might want to research. That's if you can't get Audacity
> to work (which... happens sometimes).

I can handle audacity. It's on both Android & Windows.
If it's on the iPad, maybe she can play the file in iOS Audacity?

When I looked up Audacity on the iPad I realized it's not on it.
But it's not on Android either.
https://www.audacityteam.org/download/

The app called Audacity on Android doesn't seem to be the "real" one.
https://play.google.com/store/search?q=audacity&c=apps

> The namespace for audio is a bitch.

Can you clarify what "namespace" means in this context?

> An "improved" sound recorder, is one for which
> you don't have to "define a fixed size audio file" before
> you begin recording.

Absolutely. I can't imagine defining the size ahead of time.
What program requires that? I've never run into it myself.

> Some of the Windows demo apps have
> been absurd that way. The damn thing should just record,
> and then you do the Save As, and done. Check that it works
> that way.

Yes. Agree. If it requires me to guess, I don't want to use it.
--
Removing Android & replacing with Firefox as Audacity isn't there.

On 6/12/2023 11:25 PM, Peter wrote:
> Paul <nospam@needed.invalid> wrote:
>> You will need:
>>
>> 1) Copy of Audacity or a similar audio recorder.
>
> Absolutely no problem. I have Audacity on both Android & Windows.
>
>> 2) Male to male audio cable. Connect Line Out to Line In.
>> This step, is to avoid fighting with "What You Hear" setup,
>> which could waste half the afternoon.
>
> I can make one of those.
>
>> 3) Run Firefox.
>
> I can run that.
>
>> 4) In Windows Mixer, set Windows Sounds (beeps and boops) to zero.
>> This is to avoid the Windows Defender toast notification or a
>> You Have Mail ding, from ruining your recording.
>
> Windows Mixer? https://www.google.com/search?q=what+is+windows+mixer
>
> OK. It's explained here for both Windows 10 and Windows 11.
> https://www.asus.com/support/FAQ/1044702/
>>
>> Now, use Firefox Reader View, open your text file.
>>
>> [Picture]
>>
>> https://i.postimg.cc/DzhXyywq/firefox-reader-view-TTS.gif
>
> Firefox reader view?
> https://www.google.com/search?q=what+is+firefox+reader+view
>
> OK. It's explained here.
> https://support.mozilla.org/en-US/kb/firefox-reader-view-clutter-free-web-page
> "Reader View is a Firefox feature that strips away clutter like buttons,
> ads, background images, and videos while also letting you customize the
> layout and theme to fit your reading preferences."
>
>> After listening to a half dozen other TTS implementations
>> that suck, that one is almost bearable.
>
> Whatever Windows sound plays when I am in MerriamWebster seems fine to me.
> https://www.merriam-webster.com/dictionary/otorhinolaryngologist
>
>> With some programming, as Newyana2 describes, you can do better
>> than that, but then it's a real long project. This method
>> may not be efficient, but you also don't have to be a programmer.
>
> I don't want to be a programmer since I can already convert most "things"
> to text so that's why I wanted to turn text into something she can play.
>
>> The Firefox page should auto-scroll, as it plays the text. I do not
>> know how long a web page (or text file) can be, and whether your attempt to
>> convert War & Peace to sound, is going to work.
>
> Most things are about twenty pages as I've been converting them to text
> already for her where she selects the text inside an email and plays it.
>
>> I may have recorded for 16 hours before with Audacity, but that
>> was on the other, smaller machine. Make sure all the steps of
>> recording and saving are working on Audacity, before your run.
>> You would not want to generate all those tiny sound files,
>> and have it croak before saving.
>
> I'm familiar with audacity. It can output MP3 which I assume is the
> standard file format for all computers as it is for my phone & PC.
>
>> I think Windows also has an "improved" sound recorder, which
>> you might want to research. That's if you can't get Audacity
>> to work (which... happens sometimes).
>
> I can handle audacity. It's on both Android & Windows.
> If it's on the iPad, maybe she can play the file in iOS Audacity?
>
> When I looked up Audacity on the iPad I realized it's not on it.
> But it's not on Android either.
> https://www.audacityteam.org/download/
>
> The app called Audacity on Android doesn't seem to be the "real" one.
> https://play.google.com/store/search?q=audacity&c=apps
>
>> The namespace for audio is a bitch.
>
> Can you clarify what "namespace" means in this context?
>
>> An "improved" sound recorder, is one for which
>> you don't have to "define a fixed size audio file" before
>> you begin recording.
>
> Absolutely. I can't imagine defining the size ahead of time.
> What program requires that? I've never run into it myself.
>
>> Some of the Windows demo apps have
>> been absurd that way. The damn thing should just record,
>> and then you do the Save As, and done. Check that it works
>> that way.
>

I haven't tried any of them, but I did a web search for "convert pdf to
mp3" and there are a whole bunch of free on line converters listed.
Have you explored that option?
> Yes. Agree. If it requires me to guess, I don't want to use it.

In article <news:u69qne$3jmg7$1@dont-email.me>, Retirednoguilt
<HapilyRetired@fakeaddress.com> says...

>>> You will need:
>>>
>>> 1) Copy of Audacity or a similar audio recorder.
>>
>> Absolutely no problem. I have Audacity on both Android & Windows.
>>
>>> 2) Male to male audio cable. Connect Line Out to Line In.
>>> This step, is to avoid fighting with "What You Hear" setup,
>>> which could waste half the afternoon.
>>
>> I can make one of those.
>>
>>> 3) Run Firefox.
>>
>> I can run that.
>>
>>> 4) In Windows Mixer, set Windows Sounds (beeps and boops) to zero.
>>> This is to avoid the Windows Defender toast notification or a
>>> You Have Mail ding, from ruining your recording.
>>
>> Windows Mixer? https://www.google.com/search?q=what+is+windows+mixer
>>
>> OK. It's explained here for both Windows 10 and Windows 11.
>> https://www.asus.com/support/FAQ/1044702/
>>>
>>> Now, use Firefox Reader View, open your text file.
>>>
>>> [Picture]
>>>
>>> https://i.postimg.cc/DzhXyywq/firefox-reader-view-TTS.gif
>>
>> Firefox reader view?
>> https://www.google.com/search?q=what+is+firefox+reader+view
>>
>> OK. It's explained here.
>> https://support.mozilla.org/en-US/kb/firefox-reader-view-clutter-free-web-page
>> "Reader View is a Firefox feature that strips away clutter like buttons,
>> ads, background images, and videos while also letting you customize the
>> layout and theme to fit your reading preferences."
>>
>>> After listening to a half dozen other TTS implementations
>>> that suck, that one is almost bearable.
>>
>> Whatever Windows sound plays when I am in MerriamWebster seems fine to me.
>> https://www.merriam-webster.com/dictionary/otorhinolaryngologist
>>
>>> With some programming, as Newyana2 describes, you can do better
>>> than that, but then it's a real long project. This method
>>> may not be efficient, but you also don't have to be a programmer.
>>
>> I don't want to be a programmer since I can already convert most "things"
>> to text so that's why I wanted to turn text into something she can play.
>>
>>> The Firefox page should auto-scroll, as it plays the text. I do not
>>> know how long a web page (or text file) can be, and whether your attempt to
>>> convert War & Peace to sound, is going to work.
>>
>> Most things are about twenty pages as I've been converting them to text
>> already for her where she selects the text inside an email and plays it.
>>
>>> I may have recorded for 16 hours before with Audacity, but that
>>> was on the other, smaller machine. Make sure all the steps of
>>> recording and saving are working on Audacity, before your run.
>>> You would not want to generate all those tiny sound files,
>>> and have it croak before saving.
>>
>> I'm familiar with audacity. It can output MP3 which I assume is the
>> standard file format for all computers as it is for my phone & PC.
>>
>>> I think Windows also has an "improved" sound recorder, which
>>> you might want to research. That's if you can't get Audacity
>>> to work (which... happens sometimes).
>>
>> I can handle audacity. It's on both Android & Windows.
>> If it's on the iPad, maybe she can play the file in iOS Audacity?
>>
>> When I looked up Audacity on the iPad I realized it's not on it.
>> But it's not on Android either.
>> https://www.audacityteam.org/download/
>>
>> The app called Audacity on Android doesn't seem to be the "real" one.
>> https://play.google.com/store/search?q=audacity&c=apps
>>
>>> The namespace for audio is a bitch.
>>
>> Can you clarify what "namespace" means in this context?
>>
>>> An "improved" sound recorder, is one for which
>>> you don't have to "define a fixed size audio file" before
>>> you begin recording.
>>
>> Absolutely. I can't imagine defining the size ahead of time.
>> What program requires that? I've never run into it myself.
>>
>>> Some of the Windows demo apps have
>>> been absurd that way. The damn thing should just record,
>>> and then you do the Save As, and done. Check that it works
>>> that way.
>>
>
> I haven't tried any of them, but I did a web search for "convert pdf to
> mp3" and there are a whole bunch of free on line converters listed.
> Have you explored that option?
>> Yes. Agree. If it requires me to guess, I don't want to use it.

I have tried them in the past but most PDFs don't contain any text.
The text looks like text but try selecting it and you can't.

Is there an electronic OCR for PDFs that can reveal the true text?

Peter <occassionally-confused@nospam.co.uk> wrote:
> Chris <ithinkiam@gmail.com> wrote:
>>> But how can I turn the text into a "podcast" for her to listen to?
>>> And would it be too large to email to her in a typical email message?
>>
>> iOS has very good accessibility features. I suspect it already can read any
>> pdf you provide.
>
> Thanks for that suggestion where the main goal is one way for everything.
> She will get super confused if there isn't one way for everything.

Don't think that's possible. There will always be exceptions, but I guess
you/her need to decide which option has the fewest exceptions.

> Not one way for this kind of PDF. Another for that kind of PDF.
> And yet another for text. And another for HTML. And so on.
>
> It has to be ONE way for EVERYTHING, where there's no way any program, no
> matter how smart it is, can read out ALL PDFs (since many are bitmaps).

Unless it's very old pdfs the usual is certainly not bitmap pdfs

> For the bitmaps, I was thinking of looking up OCR or, if it's short, just
> reading it aloud to myself and recording it on my phone's voice recorder.

How is that "one way for everything"? A screen reader would be far more
consistent and quicker for you.

>> https://support.apple.com/en-gb/guide/iphone/iph3e2e415f/ios
>
> Thanks for that suggestion since she only has iOS iPhones & iPads.
> And I don't.
>
> I looked at that page which seems to be much like what Windows does.
> Which means it will drive any normal person nuts with extraneous talking.

People with visual impairment need that extraneous talking.

> It has to be an audio of the text only and nothing else for it to work.
> The firefox text-only speaking suggestion from Paul is along those lines.

Chris <ithinkiam@gmail.com> wrote:
>> It has to be ONE way for EVERYTHING, where there's no way any program, no
>> matter how smart it is, can read out ALL PDFs (since many are bitmaps).
>
> Unless it's very old pdfs the usual is certainly not bitmap pdfs

There is probably a program to automatically OCR a file that is really an
image of the text. It's not really a PDF. It's more like an image of a PDF.

On 06/15/2023 12:19 PM, Peter wrote:
> Chris <ithinkiam@gmail.com> wrote:
>>> It has to be ONE way for EVERYTHING, where there's no way any program, no
>>> matter how smart it is, can read out ALL PDFs (since many are bitmaps).
>>
>> Unless it's very old pdfs the usual is certainly not bitmap pdfs
>
> There is probably a program to automatically OCR a file that is really an
> image of the text. It's not really a PDF. It's more like an image of a PDF.
A quick Google, seems to indicate that OCR is part of the professional
reader, ot a paid subscription

On 6/15/2023 12:34 PM, knuttle wrote:
> On 06/15/2023 12:19 PM, Peter wrote:
>> Chris <ithinkiam@gmail.com> wrote:
>>>> It has to be ONE way for EVERYTHING, where there's no way any program, no
>>>> matter how smart it is, can read out ALL PDFs (since many are bitmaps).
>>>
>>> Unless it's very old pdfs the usual is certainly not bitmap pdfs
>>
>> There is probably a program to automatically OCR a file that is really an
>> image of the text. It's not really a PDF. It's more like an image of a PDF.
>
> A quick Google, seems to indicate that OCR is part of the professional reader, ot a paid subscription

A PDF can be OCR-processed and have a "character overlay" done
by the OCR software. You can for example, wipe over the
new overlay with a mouse, then copy and paste the text, elsewhere.
This could also cause a font to be defined in the PDF, which will
show in the Properties of the document, where previously the
document had absolutely no fonts referenced.

The other kind of OCR, would be runtime OCR of some sort.

Bitmap-only PDFs come from cheap scanners. A scanner with
more software capability, can do the text overlay thing. An Adobe PDF
product can do OCR. My Acrobat Reader 4 or so (ancient), had OCR, but
the resolution of the bitmap PDF had to be between 200DPI and 400 DPI
or so. In fact, it did not like higher resolution scans to be offered
to it, which was a major nuisance.

Linux has Tesseract, and that would be an example of a keyword to use
when researching various PDF softwares there.

*******

You can deconstruct a PDF various ways.

mutool.exe extract example.pdf

would draw out the bitmaps for you, and save as image files.

That's not particularly useful as some sort of workflow.
It's mainly for the user to have a better understanding
of the PDF type in front of them.

In this example, the PDF file has no fonts. The document is
ten pages. There happen to be ten output images.

[Picture]

https://i.postimg.cc/WbY3MkBR/bitmap-pdf-sample.gif

Now, I process it:

D:\>mutool extract sony_srs-t1_t1pc_sm.pdf
extracting image-0001.png
extracting image-0014.png
extracting image-0019.png
extracting image-0024.png
extracting image-0029.png
extracting image-0034.png
extracting image-0039.png
extracting image-0044.png
extracting image-0049.png
extracting image-0054.png

Properties show:

image-0044.png (document page 8)

dimensions 6600 x 4529 17"x11" est. 388 dpi
8 bit depth (likely grayscale in this case)
421 KB (some decent compression)

On Ubuntu, using the keyword, I can find:

ocrmypdf - add an OCR text layer to PDF files

We give it a try.

ocrmypdf -l eng --image-dpi 400 --output-type pdf image-0044.png image-0044.pdf
INFO - Input file is not a PDF, checking if it is an image...
INFO - Input file is an image
INFO - Image seems valid. Try converting to PDF...
INFO - Successfully converted to PDF, processing...
Scan: 100% 1/1 [00:00<00:00, 625.83page/s]
INFO - Using Tesseract OpenMP thread limit 3
OCR: 100% 1.0/1.0 [00:07<00:00, 7.01s/page]
JPEGs: 0image [00:00, ?image/s]
JBIG2: 0item [00:00, ?item/s]
INFO - Optimize ratio: 1.00 savings: 0.0%

Wiping over the output PDF file image-0044.pdf reveals text is there.

[Picture]

https://i.postimg.cc/76gvZFkP/bitmap-pdf-output.gif

The OCR step in this case, was done in bash shell in Win11.
So I didn't even need a VM like VBox to do it.

Paul

On 15/06/2023 17:19, Peter wrote:
> Chris <ithinkiam@gmail.com> wrote:
>>> It has to be ONE way for EVERYTHING, where there's no way any program, no
>>> matter how smart it is, can read out ALL PDFs (since many are bitmaps).
>>
>> Unless it's very old pdfs the usual is certainly not bitmap pdfs
>
> There is probably a program to automatically OCR a file that is really an
> image of the text. It's not really a PDF. It's more like an image of a PDF.
>
Actually its a pdf of an image (usually a scan of a printed page). There
are hundreds of programs to OCR text images in pdf or image files, many
of them freeware. Many pdf readers include an OCR function - I use
PDF-XChange Viewer.

On 6/15/2023 10:57 AM, Peter wrote:
> In article <news:u69qne$3jmg7$1@dont-email.me>, Retirednoguilt
> <HapilyRetired@fakeaddress.com> says...
>
>> I haven't tried any of them, but I did a web search for "convert pdf to
>> mp3" and there are a whole bunch of free on line converters listed.
>> Have you explored that option?
>>> Yes. Agree. If it requires me to guess, I don't want to use it.
>
> I have tried them in the past but most PDFs don't contain any text.
> The text looks like text but try selecting it and you can't.
>
> Is there an electronic OCR for PDFs that can reveal the true text?
>

In Win10, you can run Bash shell and use "ocrmypdf", a Ubuntu package.

It uses Tesseract. Tesseract is a free OCR, quite old now. The
latest innovation, is it has a new module trained on a neural net
or the like, which offers a second conversion process, and potentially
improves the quality.

In the past, 0 and O were easily confused, and that's one
of the simpler tests of whether an OCR is any good or not.

One Windows OCR, used three converters, and used majority voter
logic and grammar checks, in an attempt to get closer to 99%
accuracy. Such accuracy still isn't really good enough. You would
not predict Tesseract would be at that level.

OCRs work best, if "font training" was applied, you get lucky,
and the document font is one that it was trained on.

OCR is generally an annoying topic. Don't drink too much coffee
and then OCR, or you could punch the LCD monitor on the computer :-)
I gave up long ago, on my Windows OCR options (I have two $99 packages
and the Adobe one, as representations of my frustration). The OCR that
comes with some scanner-to-PDF products now, is probably better
than anything I have in the room here.

OCR really excels on closed glyphs. If you had a film negative with
characters drawn on it, and the thing had "never seen a printing press",
the accuracy would be 100% (the grain size on film is tiny compared
to printing presses and paper). But real samples, and real letters, are
pock-marked along the edges of the characters, and this is what
drives the OCR nuts. It's getting 99% accuracy on "noisy" samples.

Newspaper text is done with little dots of ink. The edges of letters are
porous and not particularly suited to OCR. Again, you would not
expect even the highest resolution photo of newsprint, to yield
good text extraction.

For a person needing vision assistance, this is their world,
a world filled with noise that sighted people don't have to
deal with. Your diet is filled with 0's and O's, spoken
by robots.

And this is what inspires some people to write software. To fix that.

Paul

knuttle <keith_nuttle@yahoo.com> wrote
> A quick Google, seems to indicate that OCR is part of the professional
> reader, ot a paid subscription

I have Adobe Acrobat 6, which is a decade or more old but I tried that
after seeing your suggestion about the "professional reader" (which doesn't
really make sense to me as the professional is a writer and not a reader).

Adobe Acrobat 6 Standard has a line item menu for
"Document > Paper Capture > Start Capture" and then a choice of
All pages, Current page, and From Page x to y

The default "Paper Capture Settings" language is
"Primary OCR Language = English"
(although you can set any language in their long list).

The default "Paper Capture Settings" "PDF Output Style" is set to
"Searchable Image (Exact)"
But there were options of "Searchable Image (Compact)" & an option for
"Formatted Text and Graphics".

The default "Paper Capture Settings" "Downsample Images" dpi is set to
"None" but there were options for "Low (300 dpi)", "Medium (150 dpi)",
and "High (72 dpi)".
I thought I was heading toward the finish line when I tried it on a sample
full text page of a 200 page document whose text I couldn't select - but
the professional Adobe Acrobat 6 Paper Capture feature erred out saying
"Acrobat could not run Paper Capture on the page because
of the following error: This page contains renderable text."

I thought I'd be sneaky and run the Adobe Acrobat Professional feature of
"File > Reduce File Size > Compatible with Acrobat 4.0 and later"
But it erred out in trying to convert the PDF to Acrobat 4 saying
"This PDF cannot be made compatible with Acrobat 4.0 because it uses
transparency.", however it did convert to Acrobat 5.0 and 6.0 versions.

Unfortunately I wasn't sneaky enough as it still complained that
"This page contains renderable text."

Only did it slowly dawn on me that if the file contains "renderable text",
then maybe I could just save the entire file to plain text, which worked.

Which means there must be at least 3 kinds of PDF text documents at least
Real PDF (with selectable text which we can presume can be OCR'd)
Renderable text (without selectable text which can't be OCR'd apparently)
Bitmap text (without selectable text which we can presume can be OCR'd)

Being confused how to tell which is which, I dug up a PDF where I could
select the text, and THAT was able to be OCR'd by the professional Acrobat.
Paper Capture
"Rasterizing page and sending to Paper Capture..."
Performing page recognition
Converting to indexed color
Thresholding image
Deskewing image
Finding rules and frames
Reading characters
Forming words
Grouping characters and words
Writing ACP file
Do you want to save the changes to "file.pdf" before closing?

But where's the OCR results?
--
fup set to replace firefox with r.p.d instead.

On 6/15/2023 10:53 PM, Peter wrote:

> Adobe Acrobat Professional

> Paper Capture
> "Rasterizing page and sending to Paper Capture..."
> Performing page recognition
> Converting to indexed color
> Thresholding image
> Deskewing image
> Finding rules and frames
> Reading characters
> Forming words
> Grouping characters and words
> Writing ACP file
> Do you want to save the changes to "file.pdf" before closing?
>
> But where's the OCR results?
>
The text precisely overlays the scanned characters underneath.

If any of the bitmap below the letter "sticks out", that means
an OCR error has been made.

In Acrobat Free Reader, the File : Properties : Fonts,
is your friend. The richness and type of fonts listed
in there, tells you the kind of document you're dealing with.

The Sony service manual, which is just a scan of some paper and
stored in a PDF, the fonts list for that is: blank.

Finding no fonts, means you need to run OCR.

When I run the tool that uses Tesseract, well, Tesseract does
not add the PDF text, the thing running Tesseract on your
behalf, adds overlay text on the bitmap images. The Font
properties after it is finished ? Just one font is listed,
and it is a magical font which does not render in the reader.
You can only detect it is present, by doing a "select all",
and then a copy/paste.

When a document has a dozen fonts listed, then that has "rendered text"
in it, and you're not allowed by the tools, to be adding OCR overlay text
to a document that already has "rendered text".

Paul

In comp.mobile.android Peter <occassionally-confused@nospam.co.uk> wrote:
> But how can I turn the text into a "podcast" for her to listen to?
> And would it be too large to email to her in a typical email message?

AWS has a text to speech service called 'Amazon Polly'. There's a site for
testing it here:
https://ai-service-demos.go-aws.com/polly

You send Polly text and it generates you a URL with the speech, which you
could perhaps send to your correspondent.

AWS requires an account and payment over a certain number of characters.

Polly is an API rather than an app that you can use, but maybe somebody has
written a frontend for it that makes it easier to upload text (the demo is a
bit limited) and either pay them or use your own AWS account.

Theo

Theo <theom+news@chiark.greenend.org.uk> wrote:
> In comp.mobile.android Peter <occassionally-confused@nospam.co.uk> wrote:
>> But how can I turn the text into a "podcast" for her to listen to?
>> And would it be too large to email to her in a typical email message?
>
> AWS has a text to speech service called 'Amazon Polly'. There's a site for
> testing it here:
> https://ai-service-demos.go-aws.com/polly
>
> You send Polly text and it generates you a URL with the speech, which you
> could perhaps send to your correspondent.
>
> AWS requires an account and payment over a certain number of characters.
>
> Polly is an API rather than an app that you can use, but maybe somebody has
> written a frontend for it that makes it easier to upload text (the demo is a
> bit limited) and either pay them or use your own AWS account.

This thread has gone from the ridiculous to the sublime. All because the OP
doesn't want to use the in-built iOS functions due to his unfamiliarity and
perhaps unconscious ableism.

Peter a exprimé avec précision :
> I want to help an older person with severe macular degeneration from
> thousands of miles afar where she could use an AUDIO file of long text
> (like of a book or short story, normally, as in Great Books stories).

> I'm on Windows 10 and Android 12. She's on the iPhone and iPads.

> I can easily save a web page (like a news story) to text and sometimes I
> can convert a PDF of a book to text (depending on how the PDF was made).

> But how can I turn the text into a "podcast" for her to listen to?
> And would it be too large to email to her in a typical email message?

Try Balabolka.
http://www.cross-plus-a.com/balabolka.htm

> ...a program to automatically OCR a file that is really an
> image of the text.

PDF-XChange Editor Portable (free)
https://portableapps.com/apps/office/pdf-xchange-editor-portable

Convert -> OCR Page(s)-> All