Garry, commenting on the Hatto business, wonders if record producers will eventually eliminate the live performer altogether and just splice notes into a completely synthetic performance. Let me take that straight and see where it goes. I guess there are approximations already in those eerily perfect studio recordings based on numerous takes and lots of editing. These don’t correspond to any one performance. But yes, eliminating the original performer altogether – and calling the product a performance – that would be a step beyond.
Would it work? Would people respond to such ‘recordings’ as if they were recordings? Maybe not. In the very early days of CD some people complained about the complete silence between tracks. It sounded weird to them. In response, the studios mixed in a little barely perceptible dirt, and suddenly the CD sounded normal, i.e. more like an LP or a live venue. To get away with a synthetic performance, the studio would have to build in similar marks of plausibility, concessions to psychoacoustics, and these can be subtle indeed.
Two pianists who record the same passage of rapid, apparently even semi-quavers from a Beethoven sonata will show up as different in micro-analysis of wave patterns. The irregularities are not a failure of technique but a sonic signature, a bit like a painter’s brush-strokes. On Gestalt principles, listeners smooth them out into what they hear as notes of exactly the same length. Stravinsky complained that nobody played triplets exactly in time: everybody played the last of the three too short and prolonged the first two. But maybe everybody somehow got something right. The commonest complaint people made about Stravinsky’s conducting was that the music sounded lifeless and metronomic.
So (I suspect) our machine-recordings would have to be cannily programmed. All in all, it might be cheaper to hire a piano-player.