We’ll only use AI if it betters the sound experience: Sonos’ Harry Jones

Sonos believes they simply wouldn't be able to get to the point they are at to reproduce the original art of music, without the help from artists and creators.

Very few genuinely understand audio and its layered complexities. Often, the ones who absolutely know what they’re talking about tend to be engineers, because for them, substance matters more than anything else. As I came away from a conversation with Harry Jones, Sonos Sound Experience Engineer, he noted a very important thing — that their products get better the longer you own them. Case in point, he said, are the Sonos Ace headphones released a year ago, which recently received a vital software update adding Adaptive ANC (or automatic noise cancellation), TrueCinema, and Multiuser TV Audio Swap features.

Jones said Sonos’ intention with audio processing and personalisation was to “continue to explore and push the boundaries”

“It's not something that you buy once and that is the experience that you get, because our products do get better over time through software updates,” he noted.

Jones said Sonos’ intention with audio processing and personalisation was to “continue to explore and push the boundaries”. One thing is clear, as the American audio giants build the product portfolio for the years to come and alongside improve the current ones that users own, there is little intention to fall back on the buzzword called artificial intelligence (AI) to create something that isn’t worthwhile to the experience or doesn’t improve the sound experience. Jones spoke about the pursuit of the perfect sound, how listener feedback is balanced with benchmark results, and the specifics of getting sound right in your home. Edited excerpts.

At Sonos, how do you balance what sounds good to people versus what measures well in benchmarks and on technical equipment? If listeners say they prefer something else, how do you decide?

It's a really good question because this delves into what differentiates us as a brand. When we start development on a new product, we have a group of extremely talented and experienced audio engineers, and they'll be iterating on different tunings. The tuning itself is something that we spend a lot of time on, and you're right that a lot happens with measurement mics, anechoic chambers and lab environments, which get us 70% to 80% of the way there. A lot of people at that point would stop and release. But we know that throughout this entire process, there are a lot of trade-offs including for example, spatial audio. Those presentations are mastered in rooms with 20 loudspeakers, and it's the perfect kind of environment for that sort of thing. But we have to do that with just sort of one single box. If I had it my way, for that matter, if anyone could have it their way, and I'm sure you'd be the same, we’d have 20 loudspeakers in the living room. But that's just not a realistic expectation.

So where we come in is that we are able to represent the original intention of the music or the film or the TV piece, in a way that satisfies the original artist. It’s absolutely pivotal to us as well, from a sound philosophy point of view that we realise over the years that the only way that we can really get to that point, is with the direct input of those who created the content themselves. People who have mixed your favourite musical or being a sound designer on your favourite TV series, chances are they have worked on our speakers. It's an important thing for us and something we are very proud of, just to talk about our creative community. We have our soundboard group, which is a core group of mixers, engineers, and producers in music, film and TV. They are Oscar-winning and Grammy-winning people, and very well know their industry. We also have a wider network of more than 450 creators in this industry, who are really at the top of their game.

We simply wouldn't be able to get to the point that we're able to represent this original art without them. Obviously, if listeners do prefer something else, we offer things like, TruePlay or EQ changes in the app. But our job really is to present the best possible tune that we can with input from industry, the trusted voices. Our collaboration with creators in this sense is absolutely vital.

Sonos systems are listened to in different kinds of indoor listening environments — the noisier apartments in India, versus more peaceful and often soft-furnished homes in the West. Should sound systems try to 'fix' a room's acoustics or work with them, and how might AI tuning evolve in the next few years, beyond what TruePlay does today?

This interrogates the kind of real-world scenarios rather than us just saying our speakers sound amazing, which they do (laughs), but it really depends on the room that you put them in a lot more. Over the years, it's become pretty clear how critical TruePlay has become to start to really correct some of those, or at least help push it in the right direction, in case of poor room acoustics. Obviously, we can't fight physics, but of course, TruePlay will start to help the acoustics of the speaker by trying to account for the space. And really help the brilliance that we're really known for to really help it stand out. The fundamental belief with TruePlay is that getting great sound at home shouldn't be especially difficult. It should be you asking why doesn’t my speaker sound very good? You should be able to put it down in most places, run TruePlay, and get that extra quality.

If there are any trends over the years that I've seen which are kind of like low-hanging fruit for consumers to help the sound experience, it is definitely a placement. We’ve seen trends nowadays with media walls where people will put sound bars in sort of little spaces or shelves. There are up-firing drivers, there's out-firing drivers, and people don't quite understand that. And so you get a big buildup of energy within this very small acoustical space and people say they can't hear the dialogue and whatever. A bit more knowledge or understanding of that would really help to put it on a TV stand rather than in a cavity. Even simply, the sort of the simplicity of having the sound bar in front of you and your surrounds behind you, and then TruePlay can really help you get to that next level once you understand the sort of the correct placement of it all.

In terms of how we're working with AI to combat the environment, but also other issues, I suppose, on top of that, is that we've made massive leaps and bounds forward with our new speech enhancement feature. And we're actually now able to separate the speech from everything else and provide the user with these really tastefully “tuned 4 levels” and they increase in intensity as they go up. We understand that now dialogue is becoming more and more difficult to understand, and you know it's a myriad of issues. We don't pretend that this future can solve everything, but we can actually really hone in on, you know, a difficult mix, if you're hearing something where we're seeing rushed film schedules in which post-production has to happen in a very short amount of time — so the recording of everything is really rushed. It's like TV, a conveyor belt where it just has to go out, you go on Netflix and it's got like 18 seasons on it. It’s all getting very rushed.

Also, people have hearing impairments. We’ve worked with a charity in the UK called the Royal National Institute for Deaf People, or the RNID, and they really helped us define that sort of top-level, the maximum setting in the app. So we could really make film and TV watching more accessible. There are a number of things that we are trying to target in this space. And maybe on top of that as well, an investment in a Sonos system and the true sort of modularity and everything working together is a really important point. When it comes to the sort of distribution of audio throughout your home and helping in that sense, as well as flexibility, ease of use, and really having that sort of optimum sound experience sort of no matter where you are.

Do more speakers really mean better sound, because that tends to be the conventional wisdom, or is it more a factor of tuning and feeling the sound?

I do believe that more speakers does not mean better sound at all. We focus a lot of energy into how our speakers perform in a standalone configuration. For example, we would absolutely be tuning the speaker itself and that then becomes your benchmark for adding a stereo pair. I would say that because we are so modular as an audio company, things have to sound good by themselves too. We spend more than half the time on the single unit itself. And then when you have got one thing right, to then add something else on, the stereo presentation sort of looks after itself if it's correctly placed. Sure, having a secondary speaker is of course better than having one, simply because spatially it's a more impressive performance. But it does all come down as well to a user's budget, taste and space. Multi-room audio is a really important aspect where you can really decide how many speakers you want and add to your system over time.

It's this entire ecosystem that does just sort of get better the longer you have it. In India, we created an experience called Sonos Sound Suites where people could intentionally listen in these environments. We got to celebrate India's culture and revival of traditional music, which was amazing. These really do showcase how our products can work together as well as separately. And I think that's a really important division to make, because as I said, we'll spend a long, long time on the individual tuning of a speaker. The second one will maybe give you a little more output and a better spatial experience, but more does not always mean better at all, in my opinion.

With AI becoming an element in almost everything, at what point would you set a limit for AI in terms of audio processing? Should the industry move away from the search of a 'correct' sound mirage, and instead focus on personalisation of experiences instead?

It's interesting because I think, as a company, we've really strived over the years to honour that artistic vision for music and for films and for TV. As I said, get those creators involved early on so that they can have their opinion on what we're doing. I must say that we're only really going to use AI if it is better for the sound experience. We're only going to use it if it's honestly worth using, and never to bring it in as a gimmick or just to keep up with the industry. We're not about that at all. We will look at an AI implementation of something and we will try our best to make that the best experience possible rather than just using it for the sake of it, especially in marketing terms, where you can use buzzwords. AI as a sort of massive buzzword in order to sell more products, that's just not what we do.

The way that I see AI is, it's a great tool in our tool set. It's a very complex field with LLMs and different machine learning models. Many artists and audio engineers, who I get to speak with through what I do at Sonos, use it a lot in their workflows now. Anything from source separation, to create a sort of Dolby Atmos mixes of old records, to AI mastering as well. Throughout the production process, and then on the transmission side, there's a lot of AI involved. It's just, how do we use it in the right way?

As you say, the search for the correct sound, I don’t think we’ve ever had a problem with that. We’ve never had an issue with working with artists and getting it right in that sense because we're sitting right next to them and are able to ask them very direct questions. We’re never looking to like sugarcoat anything, we don't want nice answers. And I think a lot of companies do and they'll defend it, justify what they've done. But we want to hear the raw truth as to what they think of our speaker in that particular instance. In that sense, I think we're really tuned in. I think AI can really help there potentially, in terms of more of the personalisation stuff, especially if you think about AI speech announcement. But yes, it's in no way central to getting good sound. Good sound is good sound. People have different opinions on that, but for me, you can never replace the human ear with an AI.