They are compressed AFAICT but not optimized but it's trivial to decompress them. If we have an LD_PRELOAD that can get access to the actual QML data (I don't yet know how) then we can capture the data, modify it and return the modified data to the QML "compiler". It's not resource intensive. I have done something similar in a Qt4 app of mine (replace certain strings during runtime in the embedded QML files). Don't ask me why I did that EDIT: One advantage for patching during runtime is we can tie the patches to a certain OS release easily.