Jukedeck co-founder and COO Patrick Stobbs talks music and machine learning ahead of Production Social
Next week Production Social are gathering in London to explore how artificial intelligence is set to revolutionise production. Among the speakers will be Patrick Stobbs , co-founder of Jukedeck, a company that uses machine learning to generate royalty free music for commercials, film and TV. LBB’s Laura Swinton caught up with him to find out more.
LBB> Where does your background lie, music or AI?
PS> I grew up playing violin, piano and percussion, and singing in a lot of choirs. At the age of 8 I met my now-co-founder, Ed, as we happened to be singing the same treble line in a choir up in Cambridge. We spent 5 years singing together then, and then reunited at university where we again sang in the same choir and tried a few musical side-projects. The rest of our team are also keen musicians, though most also have backgrounds in machine learning or software engineering - they certainly bring the brainpower!
LBB> Can you tell me about the technology underpinning Jukedeck?
PS> We use machine learning to train a system that understands the rules of music theory and can compose original music, note by note, just like a composer would. The notes produced are then passed through a system we’ve built that automatically applies music production to a professional standard, resulting in unique tracks of music. Our system can also take existing artists’ tracks, and customize them to different situations.
LBB> What sort of responses have you had from companies working in TV, film and advertising?
PS> We’ve now had tens of thousands of videos use our music, including advertisements by brands such as Google, and TV content by the likes of UKTV. They tend to tell us a similar story - that in an age where video creation is exploding, sourcing and shaping soundtracks scalably is painful due to the time and cost involved. They need need a soundtrack solution for what they call their ‘mid-tier’ content - the social media videos, explainer videos and corporate comms. And that’s where we come in.
LBB> I can imagine that some composers might feel a bit like you’re trying to replace them – what would you say to that?
PS> Not at all – composers are the bedrock of the musical world (we’d know – we’re composers ourselves!). We just know that not everyone has access to a composer (it takes money, time, education)– and that’s where we come in. We want to allow non-musicians to create unique music at the touch of a button, democratising music creation. Moreover, owing to its speed, AI opens up the possibility of creating and adapting music in realtime - allowing composer’s music to be personalised to an individual listener. That hugely excites us and the composers we meet, and they tend to share the belief that this will be a part of the future of music.
LBB> Is there any particular genre that Jukedeck works best with?
PS> It probably depends on your taste (music tends to be pretty subjective!) but personally I’m a particular fan of our piano and folk genres - and they seem to attract the most attention on our site.
LBB> What sort of input can producers or creatives bring in order to generate a track?
PS> Creators can choose the style of music, mood, duration and speed, and a few seconds later you can download a completely unique soundtrack. They can even set the exact point of musical climax, so that it syncs with a particular moment in their video - and we’ve got loads more features coming soon that are going to make it easier than ever to get the perfect soundtrack.
LBB> I’ve heard it said of AI/neural networks deployed in creative fields is that it can generate stuff that feels very different and otherworldly and genuinely different as it doesn’t rely on the usual human thought processes and heuristics – I was wondering what your thoughts were on this?
PS> If you take a look at the art produced by projects such as Google’s Deep Dream
, or the recipes created by IBM’s Chef Watson
it seems clear that AI can indeed be highly innovative (in fact, one of my favourite examples of this is Google’s system that learnt how to master Atari Breakout
, by implementing a strategy that no human had seemingly thought to employ - engineering a tunnel so that the ball hits the top of the screen). And this is partly because AI systems have the potential to ingest more information than a human reasonably could, make sense of that, and lean on it when generating new works. This is particularly exciting when it comes to music ; we’ve already produced many tracks that could certainly be described as unusual, and see new styles emerging from the field of AI music that will hopefully further the musical realm.
LBB> Can you tell me about any high profile projects Jukedeck has composed for?
PS> As well as Google and UKTV, we’ve had various other larger companies such as Cisco and SAP use our music, but we’ve also had large music artists on YouTube use Jukedeck to write backing tracks for their vocals,. And we’ve partnerships in the offing which will be exciting to announce!
Production Social will take place on Tuesday October 3rd. For more info head here.