Two ways, one rule: the words never move and every tap is a note in the track rather than a click on top of it. Pick a word by striking its row on the beat, or drop the clock entirely and program the word’s rhythm by hand.
A grid earns its second axis only when the pattern is arbitrary — which is true when you are programming syllables, and false when you are just ordering four words. So the word mode is rows alone, and only the programming mode keeps the grid.
Your taps are the instrument. Every hit is a note from the track's pentatonic scale, walking upward as your combo builds — so a streak resolves into a melody and no interval can ever sound wrong. The instrument changes as the mix grows: woodblock, then pluck, then bell, then lead stab. Accuracy changes the timbre, not just the score — perfect hits ring bright and wet, late ones come out choked and dry. A miss doesn't buzz at you; it just leaves a hole where a note should have been.
Still synthesised, because an artifact can't fetch audio. It is built as buses — drums, bass, pads, lead, taps, with reverb and delay sends and the pads sidechained to the kick — so real stems drop straight into the same structure.