Currently there isn't such feature and it's unlikely it will be
published.
This option was tested when this tool was being researched and it was
abandoned because it was unworthy. The main reasons are as below:
1) the renderer needs to choose the better letter depending on the
math properties of the original photo at the relative region. It would
generate a very worse final image if each individual drawing region
was filled using an arbitrary (unrelated) letter. This kind of tool is
unsuitable to be used with pre-selected text because this kind of text
usually generates poor results.
2) letters can be used in final image with different sizes, what
distorts the original text too, since it would break the orinal text
lines (break the original characters sequence).