top of page

Generative AI Update

HELLO! Here’s one I made earlier...



This is not me in the above video… it’s an avatar I created by uploading a 30 second clip of me speaking into my webcam and then adding a script. I used the app HeyGen to

create this video which is remarkably accurate and the fact that it’s not me is only given away by my accent, which according to my kids doesn’t sound that posh in real life! 😊 But how freaky is this…


So after that tantalising titbit, here’s some more specific information on the kind of thing that’s happening in the world of AI and how it crosses paths with real-life, human musicians. Read on for more detail…


NEW MUSIC-GENERATING TOOLS

First up, I’ll pick up where I left off on the last article about GenAI in music (written a lifetime ago in October 2023) and look at composers. There has been an explosion of interest in two recent tools on the marketplace, both used for music creation. They are Free Suno AI Music Generator and Free Udio AI Music Generator.


Both of these apps use text-based prompts i.e. written instructions to generate music. They are definitely aimed at the amateur music-maker. It seems increasingly unlikely that these tools present a genuine challenge to real (i.e. human) composers because, not surprisingly, the creative output is very derivative.


But the terms of service on each platform essentially leave it up to the user to decide the legal parameters for how to use the output. The assumption of most users would be that, having paid to use the service, the output is theirs to do what they want with. So for example, there is no explicit instruction to prevent the output being exploited commercially; for example as background music in an advertisement.


As such, the app is increasing the pool of royalty free music available. And that’s not a great thing for working composers and music creators.


There’s also a proliferation of tools that purport to help composers do their job more efficiently and effectively; take a look at DAACI for one such example.





LEGAL ISSUES

But what about the library of data that has been used to make these tools work in the first place? In GenAI speak this is the data used to “train the large language models (LLMs)”.


Here’s link to an article written by Ed Newton Rex, who is the CEO of the ethical generative AI non-profit organisation, Fairly Trained.


He’s a heavyweight in the world of gen-AI and wrote a fascinating article for Music Business Worldwide. It demonstrates with impressive detail, the massive coincidences between the music output produced by Suno and the famous tunes we all know, love and should pay to enjoy. You can read it here.


Meanwhile, legal challenges have been mounted by both the New York Times and the Authors Guild in the US against Open AI, which is the company that developed ChatGPT. The claim is one of copyright infringement and specifically that Open AI trained its model on copyrighted material and as such, has breached copyright law.


What this demonstrates is that some of these powerful organisations now feel that there is a genuine case to answer for. The outcome of these cases will be fundamental to the way in which artists and creators are recognised and remunerated for their original content when it’s ingested and used by Generative AI tools.


And circling back to Ed Newton-Rex, Fairly Trained is an organisation that wants to see legislated fair play, to protect creators in this wild west environment.


Guardrails are being introduced with the Generative AI Copyright Disclosure Act in the US and the AI Act in the EU.


There's another interesting article about the Authors Guild case you can read here or read more about the NY Times case here.


Meanwhile, we have to wait and see but I’d encourage you all to continue to lobby your unions for action to support musicians in seeking fair remuneration here. Take a look at the Human Artistry Campaign for more information.


VOICE USERS

My interest in this area was started when I was told by a friend about the increasing usage of voice-clones to narrate audio books and how this was robbing professional voice-users of their livelihood. Since then I’ve looked at lots of apps online which enable you to upload clips of your voice and can then generate a clone. Very much like my avatar but specifically for voices.


For the most part, there seems to be an element of fair-play inherent in the way these tools are taking in voice content insofar as the fact that many of them now specifically offer a licencing agreement with professional voice users. Take a look at ElevenLabs (one of many) for more on this.


If my voice on the avatar is anything to go by, then as with composers, the lack of nuanced human emotion is still the give-away but let’s continue to watch this space…





UPCOMING BLOG POST


We’re moving away from robots and taking a look at the pros and cons of introducing your musical instrument into your business as a capital asset.

This decision requires a viewpoint often overlooked by non-specialist accountants.


The assumption is that capital allowances are claimed on assets that lose value over time - but many musical instruments appreciate in value, particularly stringed instruments.


As such, you need to weigh things up carefully before jumping in to claim your capital allowances. So stay tuned for our next Blog Post!!



If you're not already signed up to our newsletter, you can sign up here!




Comments


bottom of page