AI Audio Salvage vs. ADR: When to Make the Final Call?

Hey everyone,

I'm producing a short film right now, and we had a scene with an absolutely incredible, raw performance from our lead, exactly what we needed. Problem is, we were shooting on location, and despite our best efforts with sound blankets and a 4017 on a boom, some unexpected street noise bled in pretty heavily. We were running an AMIRA with good glass, and the visuals are stunning, but the dialogue is compromised.

Now, in post, we're looking at options: trying to clean up the existing track using new AI-driven tools, or going straight to ADR. Both have their pros and cons. I've heard great things about what AI can do, but I'm also worried about losing the nuance of that original, spontaneous performance. ADR can fix it perfectly, but it's a whole other process.

For those of you who've faced this (a noisy but truly powerful on-set performance) what's your ultimate metric for deciding whether to try salvaging the original audio with AI tools or to just bite the bullet and replace it with ADR? What's your tipping point?