The purpose of this can be, for example, drawing different vowel formants (AEOUI) for each column of image and then using image position (time) to jump or fade between them. This way, it woud be possible to synthesize fairly convincing backing choirs by automating vowel changes, especially if using images that include diphtong transitions.
And obviously it would enable other fun stuff as well
If this isn't possible in current version of Harmor, I hope you consider it as a feature for next one. Ideally so that the width of image gain's effect would be controllable, making it like a filter/resonance module with multiple spectrum contours to fade between. Or, alternatively, adding multiple "frames" to filter modules, sort of like in EQUO.