Issue link: https://digital.copcomm.com/i/1543070
C I N E M A A U D I O S O C I E T Y. O R G 43 reduction plugins employing what she calls "legalized AI, assistive AI," finding them effective to varying degrees. Brian explains how AI has revolutionized noise reduction by moving beyond static analysis to dynamic processing: "Now with AI, it can go in and do dynamic analysis of the noise as it's going along and that helps us." In restoration work, these same tools can handle noise from optical tracks and old mag tracks. "Instead of just swapping them out, we can actually get the dynamics [back, and it can] give you back something that's actually astonishingly good." The key theme, as Brent identifies, is analysis that remains assistive. "Show me where the problems are but let me make the decision about how we might solve the problem." But there's a darker side to this technological evolution. Brian's particularly alarmed by programs that "will actually do mixing for you." It goes in and "[hands] you a 5.1 mix." It's not just about job displacement—it's about the fundamental misunderstanding of what mixing is. Alexandra sees this playing out with newer filmmakers who "are less experienced," but knows "they're also given less resources, so they actually don't understand the value of a mix." The result? "We're going to end up with tons of films that have an overhead shot of a desert and one single eagle"—a devastating indictment of homogenized, AI-generated content. Brent shares a telling anecdote about asking an AI to generate "a 1946 Chevy Corvette engine start, rev, and away." The system proudly delivered results without mentioning that Corvettes weren't made until the 1950s. "It didn't say, 'I can't do that, but I can give you an American muscle car,' it just said, 'here you go."' If users don't know better, they'll move forward with fabricated reality as truth. The disruption is already happening in certain areas. Specifically with dialogue, ADR work is seeing the most immediate impact. As Brent notes, "AI noise reduction tools are actively reducing ADR hours now," though he's quick to add that "it's not changing how the story is being told, it's allowing us to tell the same story more efficiently." The danger lies when efficiency becomes the only metric. ElevenLabs, an ultra-realistic text-to-speech (TTS) and voice-cloning AI generator, is helping picture editors maintain pace—technically impressive, ethically murky. Alexandra faced this ethical quandary firsthand when asked to create AI ADR for an absent actor. When she questioned whether the actor consented, producers essentially told her other supervisors were already doing it. "We're put into a position right now where the AI is advancing a lot quicker than the opinions about how to use it and how to ethically use it." Brian sees an even more insidious threat in "cloud processes" that promise to generate all your deliverables from an Atmos mix. The price difference is significant enough that producers are "perfectly happy to just say hey, we signed this thing off on the mix stage, and hand it off to the cloud bot." But as Alexandra points out, "If AI does that, we can't guarantee that it's actually understanding emotionally what impact is happening at that moment and why we need the dynamics to change here." The panel did identify positive opportunities. Brian wants analysis tools that determine if dialogue is actually intelligible—crucial now that "we don't have the time to listen to our mixes in five different environments like we used to." Alexandra dreams of AI that shows frequency buildup or masking to help clear mixes. Brent envisions tools that account for listener demographics, recognizing that "the way music is perceived on a college radio station is very different compared to an assisted-care facility." Perhaps most concerning is what's happening on the reproduction side. Brian points out that TV's now include dialogue adjustment features because more viewers watch with subtitles. "The AI's built into the TV will now boost up the dialogue and all of a sudden that mix that we spent a lot of time on is no longer" what was intended, destroying the creative intent. The consensus seems clear: AI as an assistant [YES]. AI as a replacement [NO]. The technology should solve problems, not eliminate problem-solvers. Because at the end of the day, mixing is not just technical execution—it's about feeling your way through a story, discovering moments, and creating emotional dynamics that no algorithm can truly understand … yet.