I recently saw a meme that made me chuckle, it went something like this; “Let’s calm down about 3D printers until we can get 2D printers to work.” They say the best humour is grounded in reality, but it does make one wonder, instead of using AI to replace audio engineers and producers, how about making stuff that can help?
It’s Not All Bad
There’s been some great strides made using AI (Artificial Intelligence) and ML (Machine Learning) for audio tasks in the last few years. One clear example is in the area of noise reduction with great tools from brands like Accentize, Hush, and Waves transforming the noise reduction world.
These tools often take previously unusable or compromised audio and help to clean it up in ways that conventional noise reduction couldn’t. In some cases, for example, with Accentize dxRevive and Hush, these tools can even restore missing sound using reconstructive audio magic.
There’s also some other clever tools such as Accentize Chameleon that can sample a piece of audio and extract and recreate the reverb.
So, there are clear examples of how AI and ML powered tools can help sound engineers. After all, who wants to spend all day cleaning up bad audio when some modern tools can do it in seconds?
Assistant Not Star
Apple has recently announced a slew of new AI powered tools in Logic 11, many of which use generative AI to replace creatives, rather than help them. For those with no band to use in the songwriting process, it’s a useful idea and in some ways no different from us using drum machines in the 1980s. Some people simply don’t have access to the talent to get the job done.
However, before we throw the baby out with the bathwater, I had been thinking about other ways that AI and ML tools could be used to assist engineers and producers and was then reminded when I saw an online comment by friend of the blog, engineer and producer, Phil Dubnick, who felt the same.
There’s plenty of ways to use AI that don’t replace us but help us, here’s a few.
Naming Stuff
Perhaps the most boring job in the world, and the likely reason some of us end up with sessions filled with audio files called ‘Audio 1’, ‘Audio 2’, ad nauseum. Sometimes we forget, sometimes people are just too lazy, but NOT naming your files with meaningful names is asking for a bag of hurt.
AI and ML models already exist that can identify different instruments from the audio, so how about some smart DAW developer has our DAWs name tracks for us. Yes, of course you could rename them, but a simple default to ensure you never end up with a piece of audio called ‘Audio X’ again, would be greeted with delight by many.
Which leads me to a similar matter, how about we use the technology to name the channels and tracks too?
The Assistant Mix
Here’s one thing that I’ve discovered over the last few years while exploring technologies like ChatGPT. It's very good at doing the work you would have historically had a junior person do, but bad at detail and accuracy. It’s great at taking a lot of disparate data and assembling it into a format you can use to finesse something good, but you still need someone who knows what they are doing before it leaves the building.
Another way AI could help is to give to modern engineers something that has existed for years, the assistant rough mix. In the days when studios had numerous staff, it was commonplace to have an assistant engineer set up a rough mix before the senior engineer came in and did their magic.
It’s not beyond the realms of AI/ML to do the same thing. One could even give it a reference mix first to let it know the ballpark mix required. It could then get a rough mix up, complete with your favourite plug-ins so you can then get to work and do the final mix. AI mastering already uses some of this technology to approximate the sound of the final master.
Cleaning Up
We’ve already discussed how AI/ML powered software is helping to clean up audio, again this is an area where some engineers will have assistants to help. They go through the audio and check for clicks from bad edits or cuts and then apply fades and crossfades. They also go through and find other problems to fix, such as bleed, breaths, plosives, etc.
Again, no one got into audio to have to do this work, but someone has to do it, why not get AI/ML powered software to do it?
Can We Get Some Help Over Here?
AI and ML powered technology is here to stay, but I’d rather the smart people coding the software concentrated on helping professional audio engineers and producers rather than trying to replace them.
I’ve spoken to many professionals over the years and asked them what one piece of advice they would give to someone starting out. So many times the answer has been the same; “make yourself useful.”
The AI and ML powered technologies most loved by professionals in this industry right now are the ones that do the boring jobs. The ones that mean engineers and producers can spend their time being creative. That’s how you win over hearts and minds.
What about you, what other ways do you think AI and ML can help you?