Ok I see. Maybe I should dust off my code then anyhow.
My code works for the trade window currently, (except for "alferix" for some reason), by comparing font samples taken from screen shots to extract the text.
also has character withs, and tries with different offsets, to cope with un-even kerning.
It should be possible to adapt for the skill view by adding a filter function to filter out only the pixels with high luminosity (text is pretty bright), and a new font sample table...
Do yours work along the same lines?
My code is available in C#, easily portable to C/C++, and an old version in plain C that probably works for the chat still
