Why not record yourself speaking the dialogue as you imagine it, then writing what you recorded? Then when you've written it, record yourself reading it out loud and listen to the recording to see if it's authentic.
After that, very little use of "s/he said" is needed. Action will tell who's speaking, as well as reveal things about their character, the location, etc., in other words, showing rather than telling.
Here's an example off the top of my head:
While he prattled on, still looking at the scene beyond the screen door, she rolled up a newspaper and smacked the table with a thwack that made him startle around to see her grimace.
"Go on then, love," she said as she slouched forward. She pursed her wrinkled lips and puffed a dead fly off the table.
He cleared his throat, readjusted his posture to a more square-shouldered erectness than when he'd started lecturing her, and began pacing the strip of linoleum between the sink and the door, the soles of his just-shined shoes setting a brisk cadence for his speech, each pass a sentence.
"My good dear, it makes no difference what that brainless twat said to you. One only has to consider..."