FEED Summer 2024 Web


What is the company’s origin story? “It’ll be impossible to automate a football mix – it’s too complicated.” That’s what Rob Oldfield heard from a senior sound supervisor when he first visited an OB truck for the Burnley FC versus Blackburn Rovers match on a rainy Sunday afternoon as part of a university project. As an audio researcher, that sentence was like a red rag to a bull. Having seen the extremely labour-intensive process of manually mixing 12 microphones around the field to mix the pitch effects microphones on that day, the founders of Salsa Sound set out on a mission. They aimed to develop some algorithms which could recognise audio events and choose the best microphone to mix to automatically. This was achieved with a deep neural network, trained to recognise the sound of a ball being kicked and/or a referee’s whistle while mixing to the best microphone accordingly. Fast-forward a few years, and Salsa Sound does far more than just virtual crowds and automated pitch mixes. Its Mixair software fully automates mixes for live sport, including crowd and commentary, on-premises or in the cloud, Salsa Sound is working on many exciting projects at the moment. As well as adding in a tonne of new features to Mixair, it’s building software tools for automated audio quality control. Go in to any MCR and there are loads of TV screens for QC on the video, but the environment is eerily quiet as very few audio feeds are being listened to. Salsa Sound’s QC tool (coming Q1 2025) will address this using AI to recognise, flag up and fix common audio issues like speech intelligibility problems, audio glitches or encoder artefacts etc. The company is also passionate about immersive audio, so much of its development is focused on this. In particular, it’s currently adding an S-ADM authoring function to its NGA tools, which will be hugely beneficial to the broadcast community in the coming years. What is the next step? As Oldfield says: “For years, microphones have been seen purely as sound- capturing devices, but they are more than this. They are (cheap and readily available) data gatherers.” Seeing audio as data rather than simply sound opens up a whole new world of data analytics, where audio can add value far beyond just an audio mix. What one thing does the company need most? Salsa Sound’s most pressing need currently is financial and human resource, so it can fulfil its potential and bring all ideas to fruition. and can support essentially any audio protocol. What is it working on right now?




Powered by