One of my main points was about learning the code as actual signals at the desired speed. It really doesn´t matter whether it´s audio or visual. There are even PC programs to learn the visual signals (e. g. JustLearnMorse IIRC). So the daily practice session with properly formed signals would still be feasible for visual signals.
Making a list and trying to transmit will make learning the code much harder.
As a HAM I have no real use for visual morse signals, so the speed and actual learning method (audio files) were fit for my purpose. Visual signals would be too slow for my needs.