Problem: words and sentences written in CAPITAL letters are read letter-by-letter

Actually the way the text with all capital letters is read, depends on the Text-to-Speech voice software you are using, not my app. Some voices may read such text normally, others may read it letter by letter. Somewhere in the voice settings they may have an option to read upper case letter text one or the other way.
 
There is a way in @Voice to convert uppercase letter to lower case, it’s slightly technical and could cause other problems when reading text, where upper case is important for some reason. Here is how you could do this:
 
-          Start @Voice app and open some text in the language you want to read in, e.g. in Spanish.
-          Press the menu button – Settings – Edit speech.
-          Press the button with + sign at the top to add a new speech replacement.
-          In the “Type” field, select “Regular Expression (RegEx)
-          In the Pattern field, type exactly this, also make sure that there are no spaces before and after this text:

\b[:upper:]{2,}\b
 
-          In the Replace field, type exactly:

A->a
 
-          If the keyboard is still showing on the screen after this typing, press the Back button once on your device to hide the keyboard, then press the “Save” button.
 
Now the program will change any “word” that consists of 2 or more uppercase letter to lower case and should read it aloud normally. Of course, this may not always be what you want, e.g. it would read USA as "usa" etc. No perfect solution.
 
Greg

first, if you have added she speech replacement listed at the top of this thread, it should be disabled, then test again.
If the TTS voice you are using still reads them as nonsensical words afterwards, you could create the following replacement:
 
Type: RegEx
Pattern: ([:upper:])([:upper:])
Replace: $1 $2                     <-- there should be 1 space after $2 as well
 
It will change text like "USA" to "U S A "

Indeed seems like you would need to create separate replacements for the longest words first, then going down to shortest... I thought that the replacement for just two letters would work reclusively, but apparently this is not how I programmed it. Sorry...
 
Greg