New-Update-on-SONOS-App-on-New-AI-Powered-Speech-Enhancement-for-Arc
A few weeks ago, SONOS PR team contacted me and sent this new update on the AISE (AI-Powered Speech Enhancement) for the SONOS Arc Ultra for better, clearer dialogues for movies and shows. This latest development intrigued me a lot as I am aware that, being a reviewer, the simplest way to increase the vocals would be to increase the mid-level frequencies on the equaliser. So, what is this New AISE feature from SONOS?
So before writing this article, I decided to dig deeper to find out more and wanted to speak to someone from SONOS. Thanks to the PR team, they arranged an online meeting with the Director of Advanced Technology, Matt Benatan.
SONOS uses AI technology to increase specific audio frequencies, highlighting the vocal stream in the audio. Generally, in the equalizer setup of the Television, if you increase the Mid-levels bar, the audio will increase, but all the background noise will also increase at that frequency. So, AI identifies the character’s voice in the movie or show, which can be a heavy bass or a low, timid voice. It highlights those vocal channel streams and gives a clearer output that can be heard without increasing the TV’s volume or the system.
That’s why we’re introducing our updated Speech Enhancement feature. It provides four levels of control that let you adjust dialogue clarity to match your needs, including one specifically made for those with hearing loss. It will be first available on our Arc Ultra Soundbar via a free software update on May 13, 2025.
A Smarter Way to Hear Every Word
At Sonos, we have always aimed to help people enjoy great sound. We knew there was more we could do for those with hearing loss, so we entered a first-of-its-kind collaboration with RNID (Royal National Institute for Deaf People) to design a Speech Enhancement solution that genuinely addresses the unique challenges this community faces while watching TV.
It was a hard truth, but one that deeply motivated our team. While TV soundbars have offered basic speech enhancement for years, they often lacked the effectiveness and sound quality needed to solve the problem truly. We embarked on a long journey to build a meaningful solution, and AI provided a breakthrough.
“By implementing machine learning into our speech extraction technology, we figured out how to separate dialogue from other sounds in the center channel and clarify speech in real time,” said Harry Jones, Sound Experience Engineer at Sonos. “This lets us draw out just the dialogue at the most needed times, without overly impacting volume or taking away from the holistic cinematic experience.”
The result is a dynamic Speech Enhancement tool with four different levels to choose from – the highest of which is expressly designed for those with hearing loss – via the Sonos app home screen:
- Low – A subtle, artistic nudge that emphasizes dialogue while maintaining the original experience and creator intent.
- Medium – A medium enhancement that provides better dialogue clarity and a tasteful balance of the surrounding mix elements.
- High – A higher setting that makes dialogue prominent while reducing other mix elements.
- Max—The most pronounced setting, where dialogue clarity takes full priority, is designed for those with hearing loss. Unlike the more balanced approach of Low, Medium, and High levels, the Max level controls the dynamic range of non-speech elements, placing dialogue firmly at the forefront of the experience.
Designed with Real People, for Real Life
Working with RNID, we collaborated with 37 participants of various ages and hearing abilities to gather their detailed everyday listening experiences and test the feature across a range of content types for nearly a year.
“We wanted to ensure that speech enhancement would work for AI, even those who might not realise they have hearing loss,” said Lauren Ward, lead RNID researcher. “One in three adults in the UK experiences hearing loss, and it is reported that just under one in four adults in the USA do too. This tool has the potential to impact a large number of people.”
We also worked with award-winning film sound mixer Chris Jenkins to bring speech extraction techniques used in the studio right into people’s homes while keeping other mix elements like sound effects and music artistically intact.
“Sonos’ new Speech Enhancement feature is a huge step forward in addressing dialogue challenges that come with the breadth of content available to people today,” said Jenkins. “It’s also a testament to the importance of retaining a human touch when building with AI – there were countless hours of listening sessions where we worked through the details together, adjusting each setting to ensure it delicately enhances dialogue while remaining true to the creator’s intent.” “When creating Speech Enhancement, we knew we wanted to put the perspective of people with hearing loss front and center from the earliest development stages,” said Benatan. “What we learned from RNID researchers and participants perfectly complemented input from Chris Jenkins, allowing us to consider a broader range of listener perspectives. It has been an incredible collaboration, and we’re grateful for their expertise and time in developing this experience together.”
About the Author
