Automixing at ISE 2015

We had some fun at ISE 2015 but for a serious reason. Automixing has been around for a while and most speech is mixed that way. For music it is a different story. Some people say that music can never be mixed by a machine. However we know that a number of manufacturers are working on this technology. The problem is that they are all focussing on the considerable engineering challenges and not on the economic and social consequences of introducing such a product.
We live in a world where computers drive trains and land planes, robots perform surgery, automatic translation is getting better and Bill Gates himself warned us recently of the dangers of AI.

If music is mixed by machines who will use it? Will sound engineers lose their jobs? How much should it cost? We wanted to find out.

We decided to carry out an experiment. Rather than just talk about Automixing music, we decided to present the technology as fully functioning, competent at the job and available now. We felt that presenting an audience with the ‘reality’ of Automixing would elicit a more honest response.

Amid a great deal of secrecy we created a fictitious product called ‘Intramix’ that was supposedly developed by a Korean research institute. We presented on their behalf purely for language reasons.
We gave a presentation that explained a number of the technologies the system was using. We tried to make these technologies sound plausible:
  • Proximity sensors on microphones
  • A register of pre-recorded sounds that it knew. We showed it recognising a voice and if you changed microphone it followed and adjusting the settings accordingly
  • Rock solid feedback suppression
  • The ability to identify repeating patterns such as a chorus
  • Spectral analysis (used already by another Automixing system)
  • ‘Heisenberg pre-processing’, where the signal is split, analysed and then processed before being output
  • Connect style motion tracking
  • Simple user feedback such as ‘like’ and ‘dislike’ - and even the idea that this feedback could be crowdsourced

We then played a 24 track piece of music three times - first of all completely unmixed. Next with it trying to mix the music and taking about 40 seconds to get things under control and finally the piece from the beginning sounding good because it had learned.
We wrote some custom software so that during the playback meters flickered, console windows filled with data and mysterious lights seemed to adjust in time with some musical events.

We then asked the audience questions and they answered with keypads (except for one question where the keypads failed and we asked for a show of hands).



Audience demographic
We first wanted to know who the audience were and most were users of mixers. We had no venue owners or managers.

“Chart


How did it sound?
Next we wanted to ask what people thought of the mix. Even though the response was 1 to 5 (bad to excellent), one person insisted on asking to give it a zero. They admitted later that they were trying to play devil’s advocate. In actual fact the mix was done by a competent professional engineer. No one gave it a 5, was this prejudice because they ‘knew’ it was a machine?

“Chart


Who will use it?
This time we wanted to know what sort of applications would be the most likely customer for this sort of technology. The audience were allowed multiple votes and we suggested that they vote for their top three most likely uses. The most popular choice was for it to compete with non-consumption where people don’t normally have an engineer and about a third thought it would do ok in several applications.

“Chart

What will happen to Engineers?
If Automixing really does work what will happen to engineers? More than half think that skilled engineers’ jobs will be at risk.

“Chart


Large shows
On the assumption that large shows are unlikely to replace an engineer entirely, we asked how Automixing might assist, though we did offer engineer replacement as an option. Most people seem to think it will take away some of the ‘chores’. Though nearly half seem to indicate that the mixer will do most of the work unless the engineer chooses to get involved - ‘auto mix except where selected’ and ‘touch override’. Seems like people like to be lazy as long as they have the impression of still being in control!

“Chart


Product form factor
We asked if it should be a standalone black box, or just have the technology built into a mixer. We said that the former would be cheaper and would not take up expensive space in a venue, the latter would allow better or total user intervention. The vast majority prefer some form of human intervention.

“Chart


Purchase options
Here we discussed it being built into a desk as a software upgrade, a separate unit that worked alongside a desk or the technology paid for on a rental basis - perhaps offset by the savings brought about by not using an engineer. Most people seem to want it as part of a conventional mixer.

“Chart


When will you use it?
We wanted to know how quickly the technology would be adopted when it is invented. Over 50% will adopt it themselves but 22% would only do so if others requested it. Fewer than 20% would never use it.

“Chart


What will encourage you to use Automixing?
Obviously it’s a big leap to let a machine take over your show, so what will convince you that it’s ok?
It appears that no one would be impressed enough to use it themselves if a major artist such as Madonna used it and they won’t rely much on the press either. But if a major manufacturer were to adopt it, or customers were given a month to play with it, you might have them hooked.

“Chart


How much should it cost?
We asked people to be sensible and not just to think that everything should be free but it appears half of us are cheapskates and don’t want to pay more for anything. If they can get a quick pay back then nearly half would buy it and the answers here seemed to justify our question about it being paid for on a per-use basis.

“Chart


How much of mixing is art?
One audience member was quite angry with this question. If the stance is that music can’t be mixed by machine because music engineering is ‘art’, then the counter argument is ‘do you really think that every second of every song, on every channel and every setting is an artistic decision? Surely some of it must be function?’ But how much? Overall the audience was split 50:50 over whether there was more or less art than function. This question will run and run.

“Chart


How do you feel about music Automixing?
We wanted to see how people felt emotionally about this. The majority seem to think it will be a helping hand to an engineer and good enough in some scenarios. Few think that it will not succeed at all.

“Chart


What will Manufacturers offer?
We asked the audience to put themselves in the minds of a manufacturer and it seems almost all would offer it, mostly as an option and some as standard.

“Chart


What about Engineers?
How will engineers cope in the world of Automixing? How many will leave the industry; or worse take up lighting? It seems most see it as another tool for them to use.

“Chart


What about Venues?
Even though we had no venue management present, we asked how venues might adopt it. Remember that the decision to use this technology may not be made by the audio industry but by a savvy venue owner wanting to save money or not deal with an engineer who is late all the time.
However, most see it as a way of helping the engineer; though a sizeable majority think that financial considerations would sway the decision. Nearly 90% would use it in some way or other.

“Chart


Summary
After the questions we revealed that the whole thing was fake. We were glad that no one left halfway through and was telling people outside about this amazing technology. Almost all the audience had the good grace to admit that we’d convinced them and no-one cried foul during our presentation.

We think we put on a good show. It would have been good to have a really big audience or to have done it with a live band. We even considered having engineers outside protesting, telling people not to go in. You can only do so much when you have no budget for mucking about and we did what we set out to achieve.
We lied in front of an audience for 45 minutes, invented fictitious technologies and convinced a room full of people to answer sensible questions about it. I’m immensely proud of my team for pulling this off. I don’t think any other company in the audio industry would do something like this.

I really want to thank the audience for being good sports. I sincerely hope that they saw this as a useful way of trying to find out how a new technology might affect us. It was a very special time and we could only ever do it once.
We hope people find the information useful. All information has been published and there was no other agenda other than to investigate the impact of a new technology.