- This topic has 7 replies, 7 voices, and was last updated 1 month ago by .
-
Posts
-
This may sound like a stupid question (wouldn’t be the first time for me!) BUT, until the cost of training these models goes down, I fail to see how an AI music company focused on Sync will make money:
1) Look at this article, the TLDR is GPT cost $700,000 A DAY to keep running! Now, I’m sure Udio, and the others cost less to train since they are more focused, but then again, I’m not sure of the resources needed to preprocess the audio to allow proper training of the model.
2) New styles will be very hard for these models to emulate as there will be few examples to train on at EXACTLY the time when TV will want the “hottest” sound. How would Suno or Udio have faired when every trailer wanted DubStep and Skrillex was the only artist doing it?
3) Look at your BMI or ASCAP royalties… In the “new world” of streaming, it’s not uncommon for my music to be on 20 shows… major shows with big names…. and the total royalty to be $47.
So I just don’t know how, with the meager royalties, AI companies with a huge overhead are going to survive this environment.
Just FYI, my “day job” is in AI/Machine Learning and I get a front row seat to the budgets needed to stand-up and pay for training enviornments…. not pretty..
@abellboy great topic and great point. It seems to me that AI is always looking backwards. So, staying current with the latest musical trends would be difficult if not impossible.
If specifically mean generating full music with AI, then yeah, it’s simultaneously scary but also has all the issues you mentioned too.
I personally find it frightening in all areas of work, that being said it if makes everyone redundant and I can spend my days just making music in some sort of ai driven utopia (unlikely I know) then I’m game.
It’s good, it will only get better, but I think the bubble will burst with AI, at least in it current iteration. Think the .com situation in the 2000’s. The market is saturated and the public are already kicking back on AI driven content.
I’m aware there’s also issues with rights when it comes to AI music, so large studios won’t touch it. To me, the people doing simplistic jingles, soundbeds etc. will be hit hard. If you’re doing big scores for libraries I don’t think it will impact much, different demographics.
“To me, the people doing simplistic jingles, soundbeds etc. will be hit hard. If you’re doing big scores for libraries I don’t think it will impact much, different demographics.”
FYI: Jingles are big time, prime time. Any track that finds it’s way into a major ad campaign can be a substantial royalty generator that pay just as well as “large scores”. Big national Advertisers will not touch AI music due to the risks, but I am also afraid that we are already in a period of “who would really even know?” Music Beds created by way of AI music generators like SUNO and UDIO can easily be uploaded to various stock music sites and it is not at all out the realm that a TV commercial editor will search and find a 100% AI track and use it on a TV spot. We are kind of at a point where it’s really hard to tell what is AI music and what is human made music? And who is going to stop AI music creators from registering their titles at PRO’s as if they are the actual “composer” of the piece?
This area of copyright law needs immediate attention. The longer we kick the can down the road and pretend this issue is not relevant right now, the greater the risks. Generative AI music sites need immediate government regulation. I wonder what the PMA position is on this? They just held a conference in LA 9-24 to 9-26.
“FYI: Jingles are big time, prime time”
I know, I’m not casting aspersions on jingle writers, but pressing “generate” on a piece of music over and over that is brief is more likely to yield results than longer pieces that need to be listened to in their entirety. Its just playing the statistics.
Thing is, and this has been bugging me for a while, even if they changed the rules, copyright and policy what’s to stop people go to countries that don’t have those rules? I believe, from reading Music Business Worldwide that the majors are going after AI companies, so I think that’s going to have a big impact on the landscape.
I think there is a good chance that for things like “lower tier” cable tv reality shows, etc., once the legalities are resolved, sups will use AI music for background cues. I don’t know how long that will take but they certainly will go with whatever is least expensive. They already pay low (or zero) blanket fees to libraries and, for a long time, libraries have given in to TV production companies taking 50% of the publishing.
When I say “lower tier”, not a judgement, just a descriptor. I’m thinking of some shows on TLC, etc.
As to whether this will take months or years, no idea.
I don’t think *RIGHT NOW* AI is a threat to the higher end of the market, we’ll see.
I think generative AI music going to have a hard time supplanting us. Aside from the many hurdles it has to overcome like re-writing the copyright laws, being sued for stealing training data, and getting good at writing sync music which has a bit of a different remit to commercial music in general then I don’t see how the end user benefits.
The production companies already get music at a low cost, along with all the assets that enables them to better fit it to their project like alt mixes, cutdowns, and stems.
Faffing about with multiple prompts trying to get something that workd sounds like more long winded workflow to me. Maybe I am missing something but the convenience factor does not seem to be there.
- You must be logged in to reply to this topic.