If you have two different input resolutions, you'd set two different input gates, and either route them both to the same output or have different output paths. Or use the wildcard 0 in an input gate to catch a range of input widths/heights. The flow of data is always from input to output, and the filter chain with the most specific matching set of input/output conditions gets used. In the case of the output scaler, only the input gate matters, since the output resolution is dictated by the filter.
The text renderer is a new filter, generally as an internal text renderer for use by other filters, but also if you want to render some basic text yourself. You could also use the overlay for more advanced text display.