Stay in touch
Leave us your email address and we will contact you back
SoundID AI is an Automatic Audio Content Recognition (ACR) Platform, specially designed for broadcast monitoring, exploiting the latest AI capabilities to handle tons of sound samples per second and removing all technical constraints faced earlier (performance, accuracy in a noisy environment…) providing broadcasters, advertisers, artists, monitoring and regulatory agencies with the appropriate technology to automatically analyze broadcast streams continuously and seamlessly, even when the searched content is played in the background!
Its valuable contribution to the broadcast professional sector can be summed up by saying that SoundID AI continuously provides a complete and time-stamped report for every broadcast of each of the sounds you want to track (music, ads, podcasts, jingles…), on any live stream we capture and analyze (TV, FM, AM, DAB+, HD Radio, web,…) or any offline content provided (podcasts, films, shows, videos…) supporting now scale-out technology to handle tens of thousands of sources in parallel.
By developing our own audio fingerprinting algorithm for SoundID, we had already met the great challenge of achieving an accuracy rate of 99+% where our competitors only reach <93.2% in ideal (noise-free) conditions. We were already able to analyze hundreds of reasonably good quality audio streams in real-time, based on a reference library that could reach hundreds of thousands of sounds.
Lately, facing the huge daily growing number of sounds to be recognized, the need to analyze an increasing number of broadcast feeds on top of all the existing ones, but also to completely remove the signal quality constraints, we have decided to completely redesign our solution to onboard the latest technologies to develop and deliver our own new world-class AI (Artificial Intelligence) audio fingerprint algorithm. We needed an algorithm that does not only crushes the numbers but was also highly tolerant to noise by understanding what has been analyzed.
To this extent, we took to our advantage a popular field in artificial intelligence which is called “deep representation learning“, involving high-dimensional vectorial spaces in which we managed to set a representation of our audio sounds.
We collaborated with NVIDIA to ensure the best use of the latest generation of GPU available today.
SoundID AI was trained to map an audio sound even when a noise is applied in front of the original sample. Therefore, it can make the distinction between disturbances, and the actual audio we need to recognize. After millions of training rounds, its neural network is capable of recognizing audio samples even when the noise is tens of orders of magnitudes louder than the original sound, in other words, we will successfully identify a sound even when the noise is so strong that humans are no able to recognize the original audio, no doubt we created a new situation where the machine is doing better than the humans.
As such, this completely innovative approach aims to provide broadcast professionals with technology capable of recognizing all sounds, even when they are played in the background or mixed with another in a multimedia stream (music in podcasts, songs in TV shows or interviews, podcast broadcasted during a live radio show…).
Furthermore, the new AI version of SoundID is now able to analyze tens of thousands of live media channels based on a constantly growing reference library. Together with this second generation of SoundID, we are happy to give access to our customers, on top of their own library, to probably the largest audio fingerprint database in the world, with references to more than 90 million audio tracks, and thousands of new titles every week.
To accomplish their mission of monitoring the media or all modern types of broadcasts, the regulatory companies or the associations for the protection of the author, composer and publisher rights must benefit from a technology ensuring their clients (artists, advertisers, …) the broadest coverage and the highest reliability.
On the other hand, broadcasters, podcasters, and all platforms making content available on the air or online may be required to declare each use of a sound covered by copyright, precisely what they are now able to do thanks to SoundID AI.
Moreover, artists, musicians, and all kinds of designers of audio creative content benefit from comprehensive reports with increased accuracy and coverage to ensure they receive their fair share of royalties.
SoundID AI, available on-premises or as SaaS subscription is then available and affordable for all business use cases including the support of young artists able to interact directly with our online platform to track the broadcastings of their first songs.
Lastly, the advertising sector is not to be outdone, since it can equally benefit from yet unequaled precision which makes it possible to confirm any broadcast of the thousands of fresh contents typically produced and broadcast daily on a virtually unlimited number of channels.