lion universal subtitle reader and more NVDA addon


Josh Kennedy
 

Hi,

Here pasted below is the lion subtitle reader and more documentation. My question though is, what is the very first combo box used for in lion settings never mind I figured it out it changes how often OCR is performed.

What is it

LION is an addon which does Automatic OCr of certain screen portions, at a certain interval. Why do I say that it is intelligent? No, not because I've written it, nor because the "i" makes for a nice acronym. Since it does OCR of the same screen portion multiple times, normally it would read the same text multiple times, which is not very nice. So, I implemented a mechanism to skip speaking of the text if it seems similar to the previously recognised text.

What can I do with it?

The main reason I wrote it was to read subtitles. Due to the way it works, it can read any subtitles on the screen, including Youtube, Netflix and other online video sites, subtitles embedded in AVI files, and even live tV!

Make sure to set the video to full screen when using it, as it works like sighted eyes do. It needs big text for better performance. The quality won't be perfect, try to set the font of the subtitles as large as possible, if you have the option, and use a high resolution screen. The OCR engine it uses is not perfect. It may have inferior results on certain images.

Other than that, It can be used to monitor any inaccessible text on the screen, like video game menus. It won't tell you highlighted text, unfortunately.

How do I use it?

To launch it with default options, just press NVDA+alt+l. LION will start OCR of the whole screen at a 1 second interval, speaking only when the text changes. Hopefully. :) If you want to customize its behaviour, go to the NVDA Menu, Preferences, LION settings. For example, a video file might have a logo in the upper left corner that gets read along with the subtitle, and disrupts the experience. We'll see how to fix this in the next section.

LION has the following settings:

1. OCR interval: how often the program does OCR. Takes values from 0.1 second to 10 seconds.

2. OCR target: specifies the screen portion to OCR. Options are: current control, current window, navigator object, and full screen

3. Crop pixels from above, below, right, left. In full screen mode, those four fields allow you to crop sections from the screen from being scanned. Those settings work only in full screen and current window modes.

Why is this setting useful? Let's remember the logo example above. Just crop 10% or so from above to skip the logo, and you won't hear it. Actually, to make the recognition faster and less resource intensive, you can crop like 70% from above, since subtitles are usually found in the lower third of the screen.

what's new

version 1.11

Fixed a bug, mainly observed in youtube fullscreen.

version 1.1

1. the crop settings affect the current window mode, too

2. implemented a similarity threshold setting, mainlly useful in games. LION always compares the current text with the one which has been spoken previously. If it thinks the text hasn't changed, it doesn't speak. This setting controls how aggressive LION is in comparing texts. 0 means all texts are identical, practically rendering the addon unusable. 1 will read all texts, even if they are the same. The default is 0.5.

version 1.0

initial version

Who wrote this, how often will I see updates?

My name is Stefan Moisei, a blind programmer trying to survive in an East European country, Romania, who has a lot to catch up to get to an occidental level. LION, and other blind specific software are just a hobby for me, developed in my spare time. So, I'm afraid updates won't come too often. But if you can code, this is NVDA, so the source is right there. :)

 

Sent from Mail for Windows 10

 

Join nvda@nvda.groups.io to automatically receive all group messages.