32-bit Float & AI NR: Has your on-set audio approach shifted?

Hey everyone,

I'm Simone, a 2nd AD, and I've been really curious about something on the audio side of things. We just wrapped a commercial where our sound mixer was using a MixPre-10 II with a few MKH 50s, and I know they've been experimenting with 32-bit float recording. On top of that, I keep hearing about how powerful things like iZotope RX have become for post-production clean-up.

Historically, the mantra has always been to get it as perfect as possible in-camera (or on-set for sound). But with 32-bit float offering such dynamic range protection and AI making impressive strides in noise reduction, I'm wondering if this has actually changed the practical on-set approach for sound mixers. Are you still stressing over every single dB with gain staging as much as before? Or is there a little more leeway, knowing what can be fixed later? When we're shooting on an ALEXA Mini, for example, the image latitude is incredible, and I'm wondering if audio is catching up in a similar way.

Has your fundamental approach to capturing 'clean' audio on set changed because of these advancements, and if so, how?