Do you do the conversion (to images) on the fly as you read? Or do you come back to the text later and create your images?
Both. Sometimes,some lines of a paragraph get converted to images automatically/on the fly. Say,for example,this line: the universe is flat
When I read this line on universe being flat,I automatically see the the image of an universe flat like a plate!!! I don’t need to make myself visualize it intentionally.
However,for some paragraphs,I do come back to the text after going through the paragraph,and convert those paragraphs to images consciously/intentionally. I do this only if I want to ‘remember or store’ the main ideas from the paragraph in specific ‘loci/peg’ for long term memory! Usually,one or two images from one paragraph!
Also, what kind of topics have you had success with? Some light reading (i.e. blogs, novels etc.) or something more complex (scientific or technical articles for ex.)?
I am mostly interested in ‘non-fiction’ books. Basically,my interest in memory enhancement techniques took place,because I wanted to remember the books that I read. And I am a voluptuous reader of ‘non-fiction’ books. I am also a computer programmer. So,I needed ‘mnemonics’ to remember ‘functions/methods’ from programming books. And I use ‘text to image conversion’ techniques to store all the stuffs in my ‘memory pegs/palaces’.
I however never tried ‘physics,chemistry,biology’ books. Those are not my area of expertise or interest at the moment. But I will slowly get involved into them in future(just for knowledge shake)! 
How long did it take for you to get proficient with this approach?
Actually,I took a course from Udemy known as “Become a SuperLearner” back in 2016. I had been practicing the ‘text to image’ conversion since then,as is recommended by the author of the course. I have always been a very visual person. That course taught me following important things(that I didn’t know before…and helped me tremendously to utilize my visualization skill):
- At first,have a pre-made ‘loci/placeholder/peg’ for storing what you are reading in a book/article.
- After reading a paragraph,pause for two,three seconds. Convert the main ideas of the paragraph into image.
- Put the converted images into your pre-made ‘loci/placeholder/peg’(of point 1).
- keep doing this for each page/paragraph,etc,that you think important for you to keep in memory.
- etc…
It has been around 3 years now that I am using ‘text to image conversion’ technique. So,things are easier for me now than when I first started. And I love this ‘skill’ of mine. It is so helpful not only to store textual informations,it also helps me to store and remember office meetings,conversations,etc…. 