It’s like a fingerprint in space. Odd syntactical constructions, catch-phrases, tending to “like” rather than “you know” – they never just happen once. Of course it can be controlled – hence the idea of formal register – but even then certain heuristic traces will remain. Even if you give two people the exact same words, the sounds will be different, and not just when comparing between Singaporeans and South Africans. It’s relatively easy to capture the uniqueness of diction, but tone and inflection are not as acquiescent.
There is no truly lossless format. What has said cannot be retracted, but will start to change shape if we pay it enough attention. It sloughs off its marginalia, recomposes itself into something more refined and polished but with the same meaning – or with a shift so slight that we might never notice. When you try to capture it with paper, it behaves like a cat in a box. If I omit a cough while transcribing from tape, does it still sound like a falling tree? Too many “um”s can make the erudite look dense; when caught in writing, a trifecta of “well”, “you know” and “I mean” spells obfuscation. There’s nothing self-explanatory about he said/she said.
My approach is being true to one another. Just like with cats in boxes, the relativity of a goal doesn’t mean it’s not worth doing. The hardest part is when your fingerprints are coming loose from too much backspacing, and all you want to do is a vague sketch instead of careful tracing. But like truth, karma is chameleonic: misquote someone, and you’ll never know what fate your words will meet in a dark study or when. The main thing is remembering that neither sound nor its absence ever signifies nothing.
Only after many hours of close listening will you realise how dull most people sound. Then you have two choices: change the station, or keep listening. When doing the first, remember both the best and worst fictive dialogue sounds exactly the opposite to how real people do. Only do the second if it begins to sound like song.