Issue link: https://digital.copcomm.com/i/1546135
C I N E M A A U D I O S O C I E T Y. O R G 63 What Happens on the Dub Stage? All of these editorial advantages were interesting. But to me, the mix was where the workflow became most compelling. More Natural Reverb and Noise End Up in the Mix Any offline rendering—whether performed during editorial or on the mix stage—is ultimately an educated guess. Deciding in advance how much noise reduction or de-reverbing will serve a mix remains somewhat speculative since it is first rendered and only then heard in context. As a result, processed dialogue often trends toward the dry side. It is safer. Cleaner. More controlled. But it can also lose some of the qualities that made the original recording feel alive, which also happens to be a common feedback of production sound mixers when asked about mixes today. Maintaining voice, reverb, and noise as separate elements changes that relationship entirely. Instead of committing to a balance before the mix begins, mixers can explore it in real time. Dedicated VCA's allow the relationship between the three elements to become a creative decision at the mixer's fingertips. A single fader move can shift a scene from an auteur cinema sound reminiscent of Godard or Antonioni to a highly polished studio-style dialogue that leaves plenty of room for action sound effects and a fully filled Foley track. The reverb component becomes particularly powerful. The combination of today's perfectly phase-aligned lavalieres, aggressive de-noising, and modern cleanup tools has undoubtedly improved intelligibility. Yet it has also involuntarily created a tendency toward an increasingly dry and direct sound that can sometimes feel disconnected from performances. With dedicated control over production reverb, it becomes possible to move in the opposite direction when appropriate. Raising the reverb component above unity can often restore a sense of space and perspective that aligned lavaliers tend to mask. Sometimes that first reflection of the studio floor is just what is needed before extending the space further using additional multi-channel reverb. Perspective changes and the spatial depth of characters within a room can be designed and fine-tuned in detail. Those wide shots, where the boom could not get close enough, are easily dialed back into foreground. In many cases, the result feels less processed, less manufactured, and more truthful to the performances on set. Practical Limitations Of course, the workflow is far from perfect. The first challenge is simply visual. Anyone opening a permanently separated session without prior warning will likely be greeted with confusion. At first glance, the session appears unnecessarily complex. Tracks multiply rapidly, and the familiar relationship between production recordings and editorial organization becomes less obvious. Then there is the listening experience. Soloing any individual component can sound strange because none of them were ever intended to exist independently. Voice without noise feels unnaturally isolated. Noise without voice can sound compromised, but might pass in the context of an M&E. The system only truly makes sense when the entire crew is on board with the workflow. More importantly, today's separation tools make mistakes. With Hush Pro, fragments of speech occasionally appear in the noise track. Transient sounds such as doors may spread unpredictably across all three tracks. Clicks can end up on either voice or noise tracks. And under difficult conditions, aggressive processing can produce the aforementioned reptile voice artifacts. That reality inevitably raises a question: Is it worth building a workflow around technology that has not yet fully matured? Looking Ahead What interested me most about this experiment was not the current state of separation technology. It was the creative possibilities that emerge once separation stops being a temporary editorial process and becomes a permanent part of the production track. The tools available today are almost certainly just the first generation. Future systems will separate voice, reverb, noise, and perhaps additional production elements with increasing accuracy. Artifacts will diminish. Processing speeds will improve. What currently feels experimental may eventually become routine. When that happens, the more important question will no longer be how we achieve separation. It will be what we choose to do with it. If machine learning is going to become part of our workflow, that is the future that interests me most. Not fewer people making fewer decisions, but the same creative teams working with more freedom, more flexibility, and ultimately more time to focus on the art of mixing.