This is currently the main performance bottleneck of the app.
Since the support for richer text, the cost to draw text on the canvas skyrocketed.

There is a lot of room for improvement here, for example:

• Instead of tokenizing text per word, it could be tokenized based on tag.
• Font properties can be calculated once and re-used.
• When moving stuff around in the canvas, if the text did not move, don't redraw it. (i.e. slow first draw, fast subsequent draws)