How to make a text on picture undetectable by pcr

[al250]

hello

i'm looking for a way to put text on image and ocr cant read the text

any suggestion?

 

[moonshine7000]

Create your own font that their system don't recognize.

 

[4ry4n]

Just distort the text like a captcha

 

[al250]

Create your own font that their system don't recognize.

Thank you very much ! I will try

 

[wooden22]

Just distort the text like a captcha

View attachment 198825

Does this works?

 

[terrycody]

I am pretty sure there are some ways, I saw it, but can't remember.

It's like the OCR can't read them, even they are normal in human eyes.

 

[AlexAkimbo]

Distortion works pretty well but it will fail against advanced AI algorithms like Google Cloud Vision and Rossum. You should consider who's gonna try to "read" it and and then choose what type of "protection" you're going to need.

 

[4ry4n]

Does this works?

I think it should at least for some, that is the main intend of captchas, to not be able to be recognized by bots.

 

[Opps]

Is there anyway to type alphabets like captcha that is a way the alphabets can be distorted just like captcha is?

 

[RoyalDarkBlue]

Is there anyway to type alphabets like captcha that is a way the alphabets can be distorted just like captcha is?

If you mean fonts, I think you can make your own though it would be static.

 

[Opps]

Create your own font that their system don't recognize.

In general OCR does not recognizes text by identifying their ‘fonts’, instead they do it by analyzing the features and shapes of characters, means it looks for similarities in the figure open areas, shapes of the different texts, and letters in the file being scanned for conversion. (That’s why it can also recognizes handwritten documents which are not using any fonts for that matter).

This process of identifying text through their feature is knows as Intelligent Character Recognition

 

[moonshine7000]

In general OCR does not recognizes text by identifying their ‘fonts’, instead they do it by analyzing the features and shapes of characters, means it looks for similarities in the figure open areas, shapes of the different texts, and letters in the file being scanned for conversion. (That’s why it can also recognizes handwritten documents which are not using any fonts for that matter).

This process of identifying text through their feature is knows as Intelligent Character Recognition

But where you think those patterns stored in a DATABASE for a computer to compare against to determine which letters is which.

 

[Opps]

YOu have a point but there are limits to this besides almost all fonts you'll use have similar Pattern to the normal alphabets. So using diff fonts won't make much difference since ICR detects patterns and there's a high chance your fonts have similarities

 

[Opps]

But where you think those patterns stored in a DATABASE for a computer to compare against to determine which letters is which.

It doesn't use font to compare which is which tho. Even if it does having different fonts doesn't mean the patterns and shapes of the letters would be any different most of the time.

 

[moonshine7000]

It doesn't use font to compare which is which tho. Even if it does having different fonts doesn't mean the patterns and shapes of the letters would be any different most of the time.

I use special unicode characters in titles to get videos pass Youtube flagging system .Any system made by man can be broken don't fall for the AI hype.

 

[flprocess]

Have you tried with opacity?