13/09/2026
I'm heading to the RAI in Amsterdam to attend International Broadcast Convention, not just to keep up-to-date with technical developments and the latest in broadcast trends, but I’m particularly interested in how AI is impacting that market.
AI is rapidly finding its way into almost everything in AV, but is it impacting broadcast and recording to the same extent? How far could it pe*****te the recording journey?
I’m not talking about a clean studio sound - think of a Cathedral with its unique acoustic signature and multiple mics placed around the building.
Imagine giving an AI a complete multitrack recording with several takes of the same piece.
Could it listen through everything and decide:
• which take has the best performance?
• where the strongest phrases are?
• which sections should be edited together?
• and ultimately mix and master it into a finished CD?
Technically, AI is already capable of analysing pitch, timing, dynamics, noise and many other aspects of a recording, but that's not really my question. Spending hours listening to takes before starting an edit, I'm not simply looking for the most accurate performance - rarely is the technically perfect take the one I want.
There may be something about another take - a phrase, a dynamic change, a moment of expression, or something I can't put my finger on that makes it feel right. Music has feel, it speaks to you and you just know when it's right.
That judgement comes from experience, knowing the music and, perhaps most importantly, having an idea of what the performance is trying to say.
Can AI do that?
Can it genuinely distinguish between the technically correct and musically compelling? Can musically compelling be clearly defined?
I think there is something about the human ear, the human musical experience that AI can't yet replicate, and whilst the future may not see AI replace the recording engineer, I think it will become (probably already is) an extraordinarily useful assistant.
I've yet to experiment with this, but one simple task would be to have it listen to all the takes and tell me the places where there is a meaningful choice.
But I can't bring myself to give it blind trust, and I would find myself listening to the takes anyway. Maybe I'm not alone?
I'm curious to hear from musicians, producers and engineers. Would you trust AI to create a performance from multiple takes?
What decisions would you happily hand over to it — and which ones would you never give up?
Meanwhile, a picture from earlier today - a Byrd treat by the girl choristers and lay clerks at St Paul’s this morning.