djpretzel   Administrators 🎮

You're right - OP fails to clarify at all, and mentions "a prompt" - singular, not "prompts" - in its opening, bold line. This is exactly what I'm saying is unclear; not specific enough. Right, that's ALSO unclear, because the waters have now been muddied with the ethicality/shade angle... which I don't think makes sense to even discuss until OCR has its own policy on whether OC ReMixes can be part of training sets, themselves... thoughts on that? I don't get the feeling that either of us is entirely up to speed on the current tools, and hell, they're adding features on a daily basis. It's one thing to ban something and then say "Until such time as..." and explain how the tools would need to evolve... but that wasn't done, right? Even with the current, limited tools, I can see someone spending a few hours with multiple prompts & revisions, tweaks at specific time markers, and some form of basic melodic input as part of the prompts, and then generating something that met most of OCR's OTHER criteria and ALSO involved more human input - more actual decision-making & human touch - than at least SOME mixes already accepted by the panel. The OP also says "wholly or in-part," so hypothetically one could generate a simple chord progression/structure using genAI and then add gorgeous, original, human vocals & instrumental performances on top, Made By Real Homo Sapiens!™, and it still wouldn't be kosher... Is that how MOST folks are using the current tech? No... they're doing simple prompts, just as you say. Is it a good idea to ban a tool based on how most people might be "abusing" it? Then go ahead and remove the torrents, right? Otherwise, get to the real crux of what you're actually trying to avoid... the underlying principle. Well I wasn't there for the conversation, and for all I know Shariq decided this himself... who can say? Did the community decide already, or is it still deciding? It's a great, relevant, & essential convo to have; it's a little disappointing when a community's very first official statement about a whole new field/technology is to ban it with language that is unspecific and already - relevant to the state-of-the-art - outmoded. What would that "larger percentage" of human touch be? At any rate, regarding percentages, I'm 100% on board with the notion that the amount of human input & decision-making for submissions to the site remain a focal point, in perpetuity...

The point you're making about degrees of control, and the points I was making about the balance between human input and machine output, are essentially the same. I don't think the OP is focusing on the right idea - starts talking about ethics, which is a different concern - and one that makes a lot more sense to tackle once OCR itself has its own policy as to whether genAI can be trained using material on THIS site, right? If it's such a "completely different" technology than what the OP is referring to, take a look at: https://www.suno.wiki/faq/metatags/voice-tags/ - among other pages. These services are developing their own prompt tagging & nomenclature and some of them ALREADY let you provide an existing melody or sound clip that they will ingest & incorporate. The tech is ALREADY substantially beyond "a single basic prompt," and thus the OP is ALREADY insufficiently precise in what it is trying to communicate... yes, that's three capitalized alreadies. @Rozovian That was one of my favorite posts of yours; I think you're right in that some of what's being discussed would have ramifications for the actual standards. "No edits," and I hope your ideas are considered.

I'm with you on the tools not defining originality - at least most of the time - but isn't "inspiration" a little more personal? I do think a policy of requesting artists using generative AI to briefly describe their process makes sense, because the example that keeps getting brought up - a single, basic prompt - is indeed far too little human input to characterize the resulting output as being particularly theirs, or particularly inspired. This is the grotesquely simplified vision/version of generative AI that I think most artists have in their heads when they get mad, or go full Luddite and attribute the technology to Satan, or conflate it with Elon Musk because they hate him, etc. But, as I've tried to point out... what if the prompt isn't "simple"? What if there are multiple prompts and refinements? For a point of reference: Imagine interacting with & refining the initial musical prompt response in a 3D DAW grid with 2D HUD elements, using a series of voice commands, almost the way Tony Stark does when he's figuring out time travel in Endgame? I don't see folks really engaging with these ideas... because it's more fun to ban things? Couldn't tell you, really... but some of the tech from that scene is getting close (!!!) - sans the time travel :) Well, I don't think most folks are doing the training part themselves... as for the "simple prompt," I suppose... read above? Also, there are already plenty of black boxes involved in modern music production... this one might be Bigger & Blacker (Chris Rock!) than others, but the idea of complete control down to the nuts and bolts is already a little suspect, even without the genAI... Well, I've heard tell that when the White Man takes your photograph with his Devil Box, your soul is lost forever, so you might be on to something... I'd just avoid the rhetorical and/or ideological metaphor of cancer altogether; ever since Snap's "Rhythm is a Dancer," this has been generally advisable... If we're going to be throwing "lazy" and "stupid" around, I'd suggest you be specific as to whether by "AI" you are referring to: Generative AI specifically of a prompt-based, minimal-interaction nature, or of... ALL generative AI (regardless of prompt complexity), or quite literally of... ALL artificial intelligence (!!) Failure to meaningfully differentiate on your part means I can only guess as to whether I: Kinda-basically agree with you on some level... Think you are being technologically myopic, on a long enough timeline, or... Think you are actively opposed to technology that is already saving human lives & has been with us longer than generative AI, a position which I would essentially characterize as Luddism... Same goes for anyone else, incidentally... gotta think it through & be specific, right?

Never go FULL hippie. I hope instead that, as with so many previous technologies, we explore the boundaries of what is possible, democratize the creative process and open it up to a much more diverse audience, and STILL have the wisdom and depth (spiritual or otherwise) to differentiate between art across the full spectrum of "human-only" to "machine-only", including the vast majority of output which will probably sit somewhere in the middle. Of all the reasons I see articulated for hating Rogan, this at least feels like it's based on some sort of principle... but as with @Rozovian's post, it seems like the principle hovers somewhere in the vicinity of Luddism. CAN it be avoided? The ships are already sailing; furthermore, consider that whatever might be done to "avoid" what you're concerned about could end up being worse than the symptoms... Think of how many might write an entire symphony because the task is no longer overbearing; also think of all the kids who stopped learning math when calculators became cheap, etc. Not every technological development is an either-or proposition that kills human utility entirely, and usually more than a few new doors are opened. Deep Blue didn't kill chess, and the popularity of The Queen's Gambit actually prompted a spike of sorts, etc., etc. Well, make up your mind - potentially, or not? It's hard to ring the alarm bells and tell folk there MIGHT be an issue... Here's how I see it... there are many types of music. There is "free expression" music written just for the sheer fun of it, there are soundtracks written for myriad forms of entertainment/education/media, there's commercial music written in many different genres, folk music performed for rituals, etc. In the short term, from a market angle, it seems like AI is going to have the most impact on musicians doing work-for-hire stuff for smaller productions, stock music, and more "utilitarian" composition where the artist name isn't front-and-center and there tends to be more of a churn/grind dynamic. In other arenas, I think it's more likely that humans will remain relevant & involved for a good long while, but they'll be leveraging SOME of the AI tooling as part of their process. I am relatively confident kids will continue learning instruments; Esther just performed a trombone solo for her 5th grade band concert, and she killed it. Very proud. The utility of learning an instrument and performing as part of a group extends far beyond whatever market value the resultant audio recording might have - which is very little, to anyone other than the parents :) Some things don't change. Yeah I believe thinking is going to get you further than feeling, on this one. If you want to feel bad (and who doesn't?!) for a more defined, concrete reason, I think it would be something like this: Eventually AI will come for almost any profession and task you can name. On a long enough timeline, hardly any human talent or utility escapes. But on that same timeline, due to that widespread market disruption, hopefully we end up advancing technology to the point where we are all far better off, where new concerns present themselves, and where we begin the long process of transforming our systems into something Roddenberry-esque. What sucks is that art & music & writing seem like they might get hit first, a little earlier than other disciplines, and thus not benefit from the market protections I believe we will eventually see arise. So it's a timing thing, mostly. But there is plenty of tragedy to be had in matters of timing, of course. It's going to be too blurry for that; you're criticizing poor old Joe for being "cheap and binary" but I think you & many others keep on conceptualizing AI as a binary all-or-nothing proposition when I think it's far more likely that we'll be looking at a MIX of human/machine input & "collaboration" into the future. That's why I chimed in originally - a single prompt feels quite wrong, yes, but if you start adding more & more prompts, refinements, or even allow for direct input of audio (I think one of the services just added this actually), you move away from the "a machine made this" vibe and towards more human input, along a spectrum. Because literally no technology in the history of technology has waited for "the whole world" to "come together and do public philosophy"? Nuclear might come closest... The line of thinking reminds me of this: https://squareallworthy.tumblr.com/post/163790039847/everyone-will-not-just

Well, you know I like principles. I'd just personally have a hard time writing that specific principle up, in a universal manner, without it essentially being distilled into Luddism. History is full of examples of technology making some jobs obsolete, or much easier, and I kinda think we're on the first of many waves of AI basically doing that; while I hope that we can get governance and social safety nets in place ahead of time, my observation is that most regulations are reactive, not proactive, in this regard. Perhaps you can rephrase what you think the generalized principle actually is, in a way that wouldn't essentially prevent any new technology from ever disrupting any job market? I believe it's difficult? Sure, I think requesting a brief statement on the use of AI, if it was employed - even if that statement is written by AI, so long as it is accurate - is a decent way of trying to capture the influence of the tech on any given track, in an honor-system mode. Ala Valve. I'm not convinced that usage in a large training dataset, where aesthetics/characteristics are gleaned and end up in a kind of weird soup that almost reminds me of human brains at times, is directly equivalent to usage as a "mere asset". People use OC ReMixes in far LESS transformative fashion, where I suppose you could call it a "mere asset", when they employ them in non-profit works, streams, let's play videos, etc., don't they? The goal of promoting VGM has some siblings - one related goal is music education, and I think OCR has always hoped that folks would learn from arrangements, study them, be inspired by them, etc. I guess the key word there is "folks" - it's definitely not the same with AI, because of the scale & speed. I don't want to sound defeatist, here, but I also want to repeat that there is enough music that is completely in the public domain that a pretty capable model could be built from JUST that corpus, so it becomes unclear what is being achieved, precisely... I think this is a great, essential conversation to be having, and this announcement does actually seem like the relevant place for it...

Nah mate, AI's been writing essays for folks longer than it's been writing music ;) Most ChatGPT essays are better than the average human can muster. The average person doesn't even tend to make music, so the bar there is less clear. Seriously though, having an AI textually auto-describe how much AI you used in a piece of your music actually sounds like a plausible and not-entirely-dystopian potentiality. I'd even say it's probable. A couple other policy points to consider: What about using ChatGPT, etc. for lyrics that you then use in a mix/submission? Again, I would hope one would tweak that output, but some of those tweaks can just be accomplished with additional prompts, refining things further. People are generally trash at writing lyrics to begin with (my hand is up, too!), and pop lyrics are often mindless and repetitive, e.g. baby, baby, baby, baby, baby, baby, baby. Baby. Baby. Yeah. Baby. So while a one-size-fits-all ban on AI might seem reasonable, I wonder about a lyrical/text exception, or at least additional clarification for this use case. It's a little incomplete to throw shade on the corpus being used to train generative AI (re: ethicality) without also having an explicit policy for training AI using OC ReMixes. Is it allowed? I vaguely recall bringing this up w/ staff but I believe the consensus was that a combination of unenforceability & general lack of clarity made such a policy unwise or unnecessary or both. That might still be the case, but then I guess I'm just not sure about throwing that specific shade without such a policy in place.

It's great to see new policy forged in my absence. I do expect this stance to change & evolve over time, as the tech also changes over time. As written, the first line mentions "a prompt" (singular) but then further down it mentions "tracks generated wholly or in part" and cites a couple of those shiny new services. I agree that music generated by a single short text prompt doesn't come remotely close to the expectations and concept of this site, or any similar community which emphasizes the creative process & human decision-making extensively. Right there with you on that one. My mind just tends to then jump to all the future hypotheticals that will one day crop up: What if it was 5 prompts instead of one? What about.... 50 prompts? 100? What if was just one prompt, but instead of text it was someone singing a complete arrangement and tapping their fingers on the table and the AI took that, respected all the beats/intervals, and built out the whole track around that concept? All will be possible, at some point... It really comes down to a ratio of input to output & the overall amount of human discretion and time involved. What I expect is that these lines will only get blurrier & blurrier, and more difficult to assess, as AI-based tooling becomes ubiquitous within DAWs and as part of creative pipelines, not just as a soup-to-nuts prompt-based magic track generator. I also do wonder about traceability/proof - beyond asking for project files as evidence of effort, seems like it would mostly be honor system. Even asking for project files will only work until AI is proficient enough to understand & navigate DAW interfaces and work within human-oriented tools, allowing for further human refinement. It's all a matter of time, and just more to consider when it arises. The ethicality argument, well... I don't think you actually need to go there, so I personally wouldn't. People learn from & mimic other people, just not at the same speed & scale as AI, and the body of work in the public domain alone is sufficient for a pretty badass model, even if that's not the approach Big Tech seems to have taken. The issue of creative input ratio is not only more critical, it also persists even if you (somehow) fix any ethical concerns.

I tried to keep this brief, but as you might know, that's not my forte. FIRST, the facts... On October 28th I informed staff that I was stepping down from my role as president/admin/owner/etc. of OverClocked ReMix, and on November 1st I also stepped down from the board of Game Music Initiative, the 501c3 non-profit organization that funds OCR. In short, I no longer feel I have the bandwidth to do these roles justice and to not only maintain, but advance, the missions of both projects. I will be working with Shariq Ansari (DarkeSword) to transition my responsibilities and ensure continuity of operations. The (excellent!) mix posted on Halloween was published without my direct involvement, subsequent posts up to the milestone #OCR04500 have been superbly executed, and I am confident that staff will continue the work necessary to operate - and evolve - OCR in my absence. I will be even less available than I have been, lately, so I apologize in advance for any lack of responsiveness. THEN, the feels... Where to even begin? It's hard to encapsulate over two decades of history; omissions are inevitable. What began as a neat side project I started in my parents' basement in 1999 snowballed into something far beyond my wildest expectations, due to the blood, sweat, tears, and unbridled, rampant creativity that thousands of you have contributed. Much of this happened before social media was even a thing and before the platforms/services we now heavily associate with the modern internet had come into being; it was a frontier, and we were on it, and we took it pretty seriously because we knew how amazing VGM is, how creative arrangements could effectively convey and explore that vast musical landscape, and how a small fandom communicating via email, IRC, & forums could collaborate to build mighty, new things. We took it seriously, often too seriously, but we ALSO played more than a few rounds of Shaq-Fu at conventions, made some truly ridiculous (but always musical!) joke mixes, and developed internal circles of lore with our own memes & jargon. NOT in strictly chronological order: there was some drama with now-legendary composer Jake Kaufman; VGMix entered the fray; we added a judges panel so it wasn't just me making stuff up; we released our first community album; the unmoderated forum birthed its own sort of... subculture; the site itself evolved to be database-driven and not just two giant dropdowns sorted by game/date; we posted mixes submitted by composers George "The Fat Man" Sanger and Jeremy Soule; we met/interviewed Hiroki Kikuta and Nobuo Uematsu; our album trailers by the incomparable José the Bronx Rican started blowing minds; we started appearing in person at Otakon, PAX, MAG, others - much love to all for having us; we bumped into Leeroy Jenkins at ROFLcon and gave him a hoodie; we started hosting from our own server and managing the technical side of things ourselves; thanks to Mr. Shael Riley (among others!), we got to remix the music for an actual Street Fighter game (!!); we released fifteen more albums... ...and then we turned ten, on December 11th of 2009. Quite a first decade, and I missed hundreds of things I shouldn't have. Hundreds of firsts, some tragic lasts, and millions of memories that can't quite be conjured by words. In 2011, we stood up for Fair Use at World’s Fair Use Day, an event organized by the non-profit Public Knowledge. In 2012, we launched our kickstarter for Final Fantasy VI: Balance and Ruin, it was taken down, we talked with Square lawyers directly for a couple hours and made the non-profit project structure clear & contractual, and we relaunched a successful kickstarter. That's not always how those things go! We launched Game Music Initiative in 2016, creating an official 501c3 charity to formalize the finances around OCR and potentially support other VGM-related projects, too. On a related note, I’ve absolutely loved seeing OC ReMixes featured by charity speedrunners Games Done Quick (GDQ) - it’s exactly the type of thing I always wanted to see, that synergy. Things do start getting a little quieter from then on out, and I think there are a ton of reasons for that, but it has been an incredible and improbable journey that I wouldn't have missed for the world. Thank you ALL for making it possible; OCR was always yours, I aspired only to stewardship of something I wanted to exist for everyone. FINALLY, the future… It's time - some would say past time - for OverClocked ReMix itself to be ReMixed. That's the point, right? Infinite permutation; endless possibility. You don't always know the day, month, or even year when your influence on something starts holding it back, or when the waning amount of time and energy you can dedicate becomes a liability. That type of certainty is often elusive; it can be a difficult diagnosis to even contemplate, and you need to look for & listen to signs. In addition to just being too much of a single point of failure for OCR (sorry, engineering mindset), the last year I've been asking myself whether it was time to let go, and I think the answer is sometimes in the asking. I have been stretched thin, like butter scraped over too much bread, and that's when you leave the Shire. Beyond representing what I genuinely believe is best for the future of OCR, I absolutely confess a personal wish to redirect reclaimed time & energy to my family and my own music. Being a husband to my wife Anna and being a father to our daughters Esther and Sarah is my meaning; I have always put them first, but now I can put them even MORE first. Esther just started learning trombone, so in a few years, expect a collab! Sarah is building her confidence learning piano & makes me proud every day. I want to write new music for them, and with them, and that requires more time than I've had. I believe the principles that have driven us - embracing all games & all styles of music, emphasizing interpretation & creativity, offering both curation and critique, and providing a non-commercial platform for those who seek it - are truly timeless, but there are many ways to honor them. I look to the new leadership/staff to galvanize, streamline, diversify, and re-imagine, within that immense space. I'll be leaving them with some ideas of my own; please let them know yours. I ask the community to support them, embrace change, provide guidance, and be patient; I believe it will be worth it! Thanks, - djpretzel

Vic Viper = legend. Mix title makes me continue the Pet Shop Boys lyric: "...let's make lots of money" Great stuff! Super-catchy synthwave with some syncopation, then we get the halfway transition with the Krueger quote and go dark for a bit, but not too long. I was totally unfamiliar with the Dave Wise original, so this exposure was a bonus as well!

Very neat piece! And I smiled; "playful" seems like the right word - there's a familiarity to the march setting, relative to the artist's other mixes, but live performances add depth and the intentional use of dissonance and some more animated/"cartoony" compositional effects make this a different ballgame. I'll confess that the first time the Calvi influence pops up, since I wasn't expecting it, I thought something was off... but I wised up quick. Successful concept execution!

@Liontamer Asked me to chime in with a sanity check. I appreciate the time & detail the artist gave to their response. I had fun listening to the track; two things stood out to me, before reading judge comments... which mostly mentioned the same two things: Balance issues (keys quiet, brass loud, general sense of foregound/background not always clear) Repetition (repeated sections may vary a bit, but by the end of the arrangement, you're definitely getting some deja vu, and since it's such a high-energy affair, it starts getting a little exhausting) Samples really didn't bother me all that much and I think some of that is coming from #1, issues w/ balance. So to me, it's that certain instruments stand out as being quieter or louder than one might expect, and potentially what would work better for the track, and then there's just some structural re-treading which, even if bits are tweaked a little each time, still has a fatigue issue to my ears. But I think I need to make a distinction here: It's (generally) good when judges all agree about an issue, or issue(s), with a mix, but... Just because there's consensus doesn't make the issues themselves more egregious, just more discernible... So even if most of us agree what the issues are, I'm kinda with Larry in that they don't feel like dealbreakers... if artist is willing to rebalance a bit & trim a bit, I do think there's a better version of this mix waiting to be found, but if not, I personally think it should land above our bar, as-is... YES (borderline)

I really dig the vintage/throwback texture here - add some vinyl crackles to the intro and I swear it would fit right in on a Tarantino soundtrack

There are parts to this that feel more like jazz fusion, then parts that definitely hit like progressive rock - interesting blend, and Ron Burgundy would love the flute!!

INSANE guitar AND organ solos on this one; really delivers the goods!

Time flies. Good example of tweaking/manipulating a sequenced solo piano mix to sound pretty pleasingly natural, so good job, I think it worked!