This is in response to a possibly (?) AI generated cover of "Like a Prayer" topping Australian radio play charts in July and ARIA deciding it was illegitimate.
>But the track, released by Queensland-based DJ and producer Josh Fawaz, has angered some in the music industry who believe it was produced with artificial intelligence-powered audio engineering software.
> ABC News has contacted Fawaz multiple times this week to seek his response to the criticism, but he has not responded.[1]
I think Ben Lee's point in the NYTimes article is good:
> The Australian musician Ben Lee, who is known for hits like “Cigarettes Will Kill You” and “Catch My Disease,” said non-generative A.I. was used by artists in the process of music production, including for pitch correction and noise reduction. But generative A.I. and its impacts on the industry needed to be looked at more closely, he said.
> “I’m not the kind of person who is at all fundamentally opposed to technological changes in my industry,” he said. “It feels like healthy pushback is a really vital part of us looking responsibly at the way to integrate these technologies into our industries.”
I remember when autotune came out and people wanted to ban that too.
The rules seem pretty loose:
> For example, if a song’s lead vocal or key instrumental performance is generated by A.I., the track would be ineligible, ARIA said. But a human lead vocal with A.I. backing vocals would be allowed.
Can you really not think of a generative AI service or a well known model from the top of your head? Ask someone in your circle (who's into this scene) what they consider to be generative AI.
Isn’t autotune something that you give an input and params and get the same output every time? I don’t think it’s generative in the sense that people mean when they say generative AI.
If I’m right, is that different from a paint splatter brush in Photoshop? It “literally generates” a splatter without you having to place every drop.
Generative AI like LLMs and image diffusion models are perfectly capable of being deterministic. I have seen it repeated a lot that they aren't and that simply isn't true.
edit: to answer the second part, photoshop brushes are absolutely generative but I'm not sure they are AI.
It’s totally possible, but would seriously need to not use much parallelism, and definitely not going to use the GPU much. So it would be really slow, but possible.
Virtually all modern music uses pitch correction in some way, often very subtly. The human voice is not perfect, and it's very common for even incredibly skilled vocalists to, say, drift slightly flat on a long held note. Or have one note a fraction of a tone off in a tight run.
Typically it's done just to save time. The singer could do takes over and over until it's perfect, but usually if a take is 99% there, you just slightly correct it.
You can tell this is the case if you listen to live recordings, although live pitch correction also exists. Vocalists aren't perfect, even Broadway vocalists and professional opera vocalists will occasionally have pitches slightly off. The human voice is an instrument, but it's not a piano.
>But the track, released by Queensland-based DJ and producer Josh Fawaz, has angered some in the music industry who believe it was produced with artificial intelligence-powered audio engineering software.
> ABC News has contacted Fawaz multiple times this week to seek his response to the criticism, but he has not responded.[1]
I think Ben Lee's point in the NYTimes article is good:
> The Australian musician Ben Lee, who is known for hits like “Cigarettes Will Kill You” and “Catch My Disease,” said non-generative A.I. was used by artists in the process of music production, including for pitch correction and noise reduction. But generative A.I. and its impacts on the industry needed to be looked at more closely, he said.
> “I’m not the kind of person who is at all fundamentally opposed to technological changes in my industry,” he said. “It feels like healthy pushback is a really vital part of us looking responsibly at the way to integrate these technologies into our industries.”
I remember when autotune came out and people wanted to ban that too.
The rules seem pretty loose:
> For example, if a song’s lead vocal or key instrumental performance is generated by A.I., the track would be ineligible, ARIA said. But a human lead vocal with A.I. backing vocals would be allowed.
[1] https://www.abc.net.au/news/2026-07-10/most-played-song-in-a...