Ibrahim Maali, our CTO, recently spoke to Music Biz and provided some valuable insights into Song Sleuth’s alliance of AI/Machine Learning and manual human review. Here are some of the key takeaways from their conversation about user generated content, technology, the music industry - and, of course, how Song Sleuth fits into this intricate and ever growing ecosystem.
To read Ibrahim Maali’s Keeping Tempo interview in full on Music Biz, click here.
The cracks on the wall of Audio Fingerprinting
The monitoring systems in online platforms currently are, for the most part, reliant on audio fingerprinting. This works based on a unique fingerprint that’s generated for every single song in a database – in the same way that each person has their own individual fingerprint.
The problem with audio fingerprinting is that, even though the technology generates very accurate results, it has a hard time encountering the intended work when it deviates from the original composition used as a reference. In simple terms, this means that a lot of covers, live performances and remixes are left unnoticed by the existing monitoring systems, like ContentID on YouTube.
User Generated Content: a vastly unexplored yet profitable realm
UGC – or User Generated Content – refers to any form of content that has been created and posted by users on online platforms. This can range from images and videos to audio or text posts. In the music industry, however, UGC usually takes the form of audiovisual content – like concert videos, covers, remixes, edits and others.
Needless to say, this type of content is a huge part of what keeps several social media platforms, like TikTok and YouTube, alive and prospering. The UGC landscape is huge, with our team constantly encountering and learning about new categories of UGC and types of music usage.
In their latest music report, YouTube announced they paid out $2 billion in royalties due to UGC – which shows a 60% increase from the previous year. In his interview, Ibrahim stated that we estimate over $100 million a year are missing proper attribution on YouTube alone – and this number continues to grow. For active touring artists and for those whose songs get covered often, we’ve witnessed missed opportunities that would’ve amounted to hundreds of thousands of dollars in monthly recurring revenue.
Finding the Unfindable
By uniting human intelligence and AI/Machine Learning, Song Sleuth has built a process through which the most accurate and high quality results can be found. This combination is what enables our growing and validated dataset to constantly refine and enhance our models.
Any and all properly licensed content is disregarded by our system, which, along with our internal review process, guarantees that our clients can be confident in the total accuracy of what we deliver to them.
Helping fans give back to their favorite artists
Amongst the sea of content that’s uploaded to YouTube on a daily basis and the limitations of audio fingerprinting that we’ve previously established, a lot of User Generated Content is left undetected. And, as Ibrahim stated in his interview, at Song Sleuth “we aim to rectify that so the fans who come to support and film their favorite artists can actually bring benefit to those same artists for their efforts.”
As the demand for live music events keeps hitting record levels – Live Nation Entertainment’s latest report shows the highest quarterly attendance ever, with over 44 million fans across 11 thousand events – it becomes more important than ever to make sure that artists and rights holders are being paid all that they’re entitled to. Song Sleuth is the solution to that.
