Yes. Quit supplying it with the electricity it needs to run.
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
AI and Future of Music Production
- Thread starter amirm
- Start date
It's music Jim But not as we know it.
Chr1
Major Contributor
- Joined
- Jul 21, 2018
- Messages
- 2,930
- Likes
- 3,092
Quite astounding how many warnings that are being ignored really.
Here's another!
www.theguardian.com
Here's another!
AI models capable of devastating attacks on governments and business months away, rare Five Eyes statement warns
Signal agencies in Australia, the US, the UK, New Zealand and Canada sound alarm after Trump blocks foreign nationals from Anthropic’s Fable AI model
www.theguardian.com
Thanks for pointing out . I wish we had been given a choice
Audio Science Pal
Senior Member
- Joined
- Feb 22, 2022
- Messages
- 343
- Likes
- 489
Back on the subject of AI and the future of music production, I find it interesting that @Brian Hall, who has limited access to musicians and singers due to his remote geographic location, is explaining the AI tools and the processes he uses to enable his creativity.
- Joined
- Mar 1, 2021
- Messages
- 3,270
- Likes
- 6,913
Meanwhile a.i companies shit scrape artists songs and uses them to train its model.
Last edited:
OP
- Thread Starter
- #487
Was it posted that Spotify cut a deal with Universal Music (UMG) so that people can create covers of their releases and upload them to Spotify? I assume there is some kind of revenue sharing. Nice to see some solutions are being put forth instead of just a conflict.
sonitus mirus
Addicted to Fun and Learning
- Joined
- Apr 19, 2021
- Messages
- 613
- Likes
- 884
I would love to create covers or major releases for my own enjoyment. I kind of can do that now as I have all the gear and subscriptions to do it, but it is not as simple as I want it to be. I want to hear what Yes or Rush songs would sound like with a different singer for some of their songs, as an example. Or a slower version of something or an acoustical adaptation. It could open the doors to a new market where old music can be resold once again. I mean, what if Enya was singing all of the classic Black Sabbath songs from the early 70's. 
MattHooper
Grand Contributor
- Joined
- Jan 27, 2019
- Messages
- 11,617
- Likes
- 20,853
This thread is creeping me the fuck out.
He gets to the point … greedy, dissembling f*ckwits basically "oh no we don't do that thing that we do".
I would love to create covers or major releases for my own enjoyment. I kind of can do that now as I have all the gear and subscriptions to do it, but it is not as simple as I want it to be. I want to hear what Yes or Rush songs would sound like with a different singer for some of their songs, as an example. Or a slower version of something or an acoustical adaptation. It could open the doors to a new market where old music can be resold once again. I mean, what if Enya was singing all of the classic Black Sabbath songs from the early 70's.
Not to attack your personal taste, but writ large this is (refreshed) bread and (private) circuses to mask the accumulation/concentration of extreme wealth for its own sake. All because people are a bit bored and a bit stupid? I mean we could work to address obvious problems, instead of making them worse by opportunity cost and neglect.
Might already be a good Sabbath song.
Last edited:
The team behind ACE-Step 1.5 took a strict "alignment-first over quantity-first" approach to prevent the downstream copyright risks that plague other music models. According to their technical report and Hugging Face disclosures, the models were trained entirely on a legally compliant, commercially safe dataset composed of:
- Fully Licensed Music: Catalogues acquired directly from partnering music labels, distribution platforms, and artists under explicit commercial training agreements.
- Royalty-Free Libraries: Large-scale commercial production libraries, loops, stems, and atmospheric cues.
- High-Fidelity Audio Stems: Approximately 6 million of the tracks utilized in the mid-stage training were broken down into individual, high-quality vocal, drum, bass, and melodic stems to teach the model how arrangement layers interact.
Global & Language Distribution
The dataset wasn't just Western pop. Because they wanted robust multi-lingual performance, the dataset spanning those 27 million processed songs included music across 50 plus languages.To make sure the model didn't trip over non-Roman scripts (like Mandarin, Japanese, or Thai), they used a trick during training: they converted 50% of non-Roman lyrics into phonetic representations (Romanization). This allowed the model to share phonological sound charts across completely different languages, which is why it maintains clear vocal pronunciation even on rarer linguistic tokens.
What Didn't Make the Cut
The filtering process was incredibly aggressive. Even though they started with 27 million tracks, millions of songs were thrown out if they had poor acoustic quality, messy mix-downs, or mismatched text descriptors. The final "polishing" phase (Phase 3 SFT) was squeezed down to just 2 million pristine, ultra-clean songs, ensuring the AI learned from high-fidelity audio structures rather than low-bitrate or poorly mastered tracks.Sounds a like a really great basis on which to create music - more worried about copyright infringement than any other metric. By implication does this mean that :
1. Their training data set is hugely incomplete and rather limited
2. their training data is made up largely of uninteresting s**t that was not good enough for somebody to release and be paid for, but which they can safely copy without fear of being sued.
I have posted several songs I generated with it in this thread. Judge for yourself.
Hi Brian
You write great lyrics / poems and I totally get and respect that these tools allow you to bring those works to life musically.
It would be fascinating as a comparison if we could put your excellent lyrics into a setting with experienced musicians and a blank sheet of paper to see what they came up with bouncing ideas around with each other and you .
I did that in my 20s and early 30s. Most of the "musicians" didn't care about music. They just wanted drugs and women.
Oh ! Well 1-0 to AI on that score
(for now !)
Not as likely to explode as their human counterparts.
Chr1
Major Contributor
- Joined
- Jul 21, 2018
- Messages
- 2,930
- Likes
- 3,092
Indeed. My biggest fear with where all of this is going, is if a malevolent Super AI ultimately takes over. Learns (my) musical tastes, then enslaves humans.
("Zoo" style or otherwise.)
"Last Christmas", by Wham would be on repeat. Horribly utilised to tyrannize us.
Most likely miscellaneous increasingly bizarre and disturbing AI created cover versions introduced over time later too.
In the "good version" we will always be granted suicide as an option. (By method/drug of choice.)
Oof. Soz.
("Zoo" style or otherwise.)
"Last Christmas", by Wham would be on repeat. Horribly utilised to tyrannize us.
Most likely miscellaneous increasingly bizarre and disturbing AI created cover versions introduced over time later too.
In the "good version" we will always be granted suicide as an option. (By method/drug of choice.)
Oof. Soz.
Last edited:
