Liontamer ⚖️ Posted Sunday at 07:36 PM Posted Sunday at 07:36 PM (edited) Artist Name: Craig Jackson Underneath the cozy melodies lies the quiet grind of village life. Gentle acoustic guitars and soft vocals reveal the emotional labor of building a home and community, turning the lighthearted theme into a heartfelt reflection. Games & Sources Animal Crossing - Title Edited 2 hours ago by pixelseph
Liontamer ⚖️ Posted Monday at 03:31 AM Author Posted Monday at 03:31 AM The track was 3:49-long, so I needed to hear the VGM used for at least 114.5 seconds to consider the source material dominant. :00-:33, :42-:48, :53-1:44, 2:03-2:37, 2:46-3:48 = 186 seconds or 64.35% source usage A good clip of the source usage was from the bass as a supporting part, but this was the only track among the 12 that invoked the source tune throughout most of the track. If you further explore (presumably human-made) music-making further, this was a more connected end result, relative to the original VGM, and thus the most cohesive piece of music among the group. Dynamically, the sections still felt fairly repetitive aside from different lyrics (though the GenAI vocals do a decent job of creating different variations and inflections, all things considered), all of the instrumentation and vocals had a warbly/buzzy quality throughout so the mixing quality's not strong, and the lyrics come off like trope-y GenAI again (including the spoken-word aside as a conclusion, enough already). NO ----------------------------------- Checklist: - Warbly vocals - Lyrics also feel AI-prompt generated, too on the nose with the rhyming, always extended character meta-narratives - Dynamics possibly undercut by limiter on volume Sorry if my perception of this undersells how much actual human-generated content is there, Craig. If I had to bet on it though, this comes across like prompt-generated lyrics and the song structure isn't dynamic enough to feel like it's mostly Craig's direct input on this. I definitely don't want a trend of people sending AI-generated content here, no matter how good it ends up being. The aim is to highlight skill, intention, and creativity with human-created, human-written, human-produced music. When you take human decision-making out of the equation, even if the end result sounds good, it wouldn't be people genuinely creating the music. ----------------------------- Alright, so I've listened to all 12 tracks and wanted to give some other thoughts that go broader than the submission. 1. I can't take Craig's claim at face value that these tracks weren't substantially, majority, or totally GenAI. That means compositionally, not just the lyrics or vocals. 2. I recognize that people will want to use it as a creative outlet, it makes music "creation" more accessible, and it can be made with good intentions, and we'll potentially get more AI-related submissions in the future. 3. We have already approved a track using AI elements. Synthesizer V was used for the vocals in https://ocremix.org/remix/OCR04711. That was a case where a musician licensed their voice for the product (rather than AI being trained on stolen content). The arrangement had been previously made/submitted a decade earlier (the first version's linked in the writeup), and this was an upgrade in the sampled vocal quality, not generative writing. There can be ethical instances of AI usage within a product. That said, context is everything. 4. GenAI content is against the spirit of music I personally want to hear, even though it'll continue to improve and evolve. I've gone down the rabbit hole some to find VGM arrangements involving GenAI. Some sounded promising from a music quality standpoint and were, structurally, more in line with VGM arrangements we'd accept in terms of source theme usage. Even this Animal Crossing submission sounds like a viable enough concept. 5. GenAI content could eventually become sophisticated and/or ubiquitous enough that it cannot be effectively screened or identified. (What about cases where someone claims to have referenced GenAI music for creative ideas or as a mockup that's then performed by real musicians? I suppose we cross that bridge when we get there.) 6. Even if there came to be ethical GenAI music (i.e. that only legally trained on approved/licensed/permitted content and was transparent in sourcing/crediting), that wouldn't be a human-created work. 7. IMO, we should explicitly add a clarifying bullet to part 2.1 of the Standards to say that we don't want tracks involving generated part-writing, composition, or arrangement, that we only want human-created music. 8. I'm disappointed at a future where we'll need to be more forensic, more skeptical, and less trusting about the steps used to create music.
paradiddlesjosh ⚖️ Posted 5 hours ago Posted 5 hours ago Larry already covered the major issues and while there is a preponderance of source material usage this time, the production quality is below the bar. Given this is in direct violation of our Submission Standards, namely 2.3, 7.1, and 7.2: And considering you weren't truthful or forthcoming about the usage of prompt-based generative AI tools in the creation of your submissions, I am no longer interested in being polite about them. Unless you plan on honing your craft as a composer, arranger, and mix engineer, shove this slop where the sun don't shine and never darken our digital doorstep again. NO
Chimpazilla ⚖️ Posted 3 hours ago Posted 3 hours ago Cosigning with my fellow Js. Cool concept, and the composition and mix sound good enough to do very well on YouTube, Spotify and all the other socials, but for OCR we look for very specific things, including enough source use with thoughtful and creative interpretations, and real musicianship. While we allow limited use of AI tools to enhance a track, this submission (and the other tracks you submitted on the same day) is far too much AI for our standards. If we have misjudged this situation, I'd welcome some real, solid proof that these tracks were made with a lot more human involvement than we are hearing, but my feeling is that such proof cannot be provided. Best of luck to you with these, you've hopped on a shiny new lucrative bandwagon with these AI tracks, we just don't want to host music like this on OCR. NO
pixelseph ⚖️ Posted 3 hours ago Posted 3 hours ago As much as I'd like to see the human involvement in this piece or the other submissions by Craig, it is overshadowed by the audacity of submitting an entire album's worth of generative AI slop and lying about how large a role generative AI played in it. I have no pithy comments to make - my fellow judges summed up my feelings pretty well, and there's no need to belabor the point. NO. DO NOT resubmit anything using Suno or other such tools again and lie about it.
Recommended Posts