|
           
|
#
?
Apr 12, 2024 22:41
|
|
|
|
| # ? May 28, 2024 22:56 |
|
|
Swagman posted:this lovely lady: Serious "creepy doctor who" vibes, here
|
|
#
?
Apr 12, 2024 22:47
|
|
|
Swagman posted:sweet pics Model?
|
|
#
?
Apr 12, 2024 23:05
|
|
|
KakerMix posted:https://voca.ro/1kZBR1OJU2ws 6.5/10 would find this acceptable as an opener act before the headline. e: Hadlock posted:Serious "creepy doctor who" vibes, here Very candyman (which apparently gave children nightmares and was banned from reairing) 
pixaal fucked around with this message at 23:41 on Apr 12, 2024 |
|
#
?
Apr 12, 2024 23:39
|
|
|
Hadlock posted:Serious "creepy doctor who" vibes, here It wouldn't feel out of place in Vancouver's Stanley Park, reminds me of https://en.m.wikipedia.org/wiki/A-maze-ing_Laughter e: also of this fellow I found on a back trail at a park: 
snorch fucked around with this message at 02:05 on Apr 13, 2024 |
|
#
?
Apr 13, 2024 01:25
|
|
|
Cars from several Star Trek alien races:    
|
|
#
?
Apr 13, 2024 01:38
|
|
|
https://www.youtube.com/watch?v=b18oEY8hg9g Well I made a AI mixtape with most the stuff from last night. Its 2 fully basic prompt songs, 2 H.P. Lovecraft poems turned into songs, and a cover of a song my buddies did during covid. I pulled it into protools and did a minor eq adjust with a little compression. I got a moderation warning while trying use James Joyces love letters as lyrics. They are just too spicy of content.
|
|
#
?
Apr 13, 2024 06:22
|
|
the Mods, they knew!
|
   
|
|
#
?
Apr 13, 2024 16:36
|
|
|
Serotoning posted:Model? https://civitai.com/models/161068/newrealityxl-all-in-one-photographic im using 2.1, not the experimental version 
|
|
#
?
Apr 13, 2024 23:12
|
|
|
More. MORE!!!!!!
|
|
#
?
Apr 14, 2024 18:41
|
|
|
Since installing Stability Matrix, I have been having the time of my life browsing civitai for interesting checkpoints and loras. One of the greatest things about it is that if you find a checkpoint or lora that you think you might like but aren't totally sure what it can do, just scroll down the page a little bit to see what other users of the site have created with that model. Because I have never seen more insane, off-the-wall poo poo in my entire life. Every new checkpoint is a new comedic high point in my life.  Do not click this link and scroll down. You might need to be registered and have "show NSFW" settings ticked to see what's going on, and yeah, I wish I could unsee it too.
|
|
#
?
Apr 15, 2024 01:52
|
|
|
XYZAB posted:
that lora is... weird, and I don't mean the subject matter it's attempting to replicate an already existing lora by training on images generated by this original: https://civitai.com/models/269148/vessels-style-sdx from looking at the examples I think the original is better
|
|
#
?
Apr 15, 2024 02:10
|
|
|
TIP posted:that lora is... weird, and I don't mean the subject matter They're all weird. There isn't one non-weird lora. Example: bread. Not weird, right? Wrong. 
|
|
#
?
Apr 15, 2024 02:51
|
|
|
Udio is pretty cool, especially when using manual mode. If not for some of the abrupt transitions, this sounds a lot like the beatles, a mix of all the early pink floyd eras etc. Pretty wild for 30 minutes of dicking around (most of that waiting for generation): https://www.udio.com/songs/c1SbWg2JopUevDjUrheUKE Noone tell Raja Ram https://www.udio.com/songs/2eDXFvy7ZS8B7oJixKQuzF Also, anyone have any idea what happened here? It seems to have just freaked out over the prompt: https://www.udio.com/songs/9KZKeogGCqXgFxQmj11tu2 no pubes yet sorry fucked around with this message at 03:07 on Apr 15, 2024 |
|
#
?
Apr 15, 2024 03:01
|
|
|
https://twitter.com/AngstHarrier/status/1779667525066015180 I'm not very keen on Udio doing 33 second segments at a time, but quality is considerably better than Suno so... *hand weighing motion*
|
|
#
?
Apr 15, 2024 04:56
|
|
|
How do you create multiple udio soundbytes that sound alike to create a song, can you feed it a snippet from the end of the first 30 seconds or what
|
|
#
?
Apr 15, 2024 05:17
|
|
|
Hadlock posted:How do you create multiple udio soundbytes that sound alike to create a song, can you feed it a snippet from the end of the first 30 seconds or what you just select extend and then pick whether you want to extends forwards or backwards and it will deliver you a new clip that includes the original and the extended bit
|
|
#
?
Apr 15, 2024 05:20
|
|
|
I've been using the stable-diffusion-webui-forge package since Friday with no issues, and out of the blue I can't get it to launch anymore. code:
|
|
#
?
Apr 15, 2024 07:32
|
|
|
did you try providing what the last line is asking for?
|
|
#
?
Apr 15, 2024 07:43
|
|
|
Ugh I can't stop playing with this stupid poo poo, it's keeping me up late https://twitter.com/AngstHarrier/status/1779794455765426665
|
|
#
?
Apr 15, 2024 10:00
|
|
|
King of the Hill thread brought up the idea of Ultra Instinct Dale after a combo King of the Hill / scooby do cross over and I had to make a set. It really seems to want to add One Piece into this Bing variations of: ultra instinct Dale gribble in Shaggy Rogers costume, orange baseball hat reading "MACK", green shirt with holes, thin wispy goatee, ultra instinct aura, glowing eyes, powerful radiating aura, fighting pose       
pixaal fucked around with this message at 14:12 on Apr 15, 2024 |
|
#
?
Apr 15, 2024 13:26
|
|
|
Commie Lasorda posted:Ugh I can't stop playing with this stupid poo poo, it's keeping me up late lol, that's really good
|
|
#
?
Apr 15, 2024 13:27
|
|
|
Bunch of misc. generations I've made lately.  Some alien things using old Alien concept art as part of the cref blend:      Some character consistency experimenting, generated the first then used it as a cref for the next four:      Another character consistency attempt while shifting styles for a T-rex on rollerblades:     As for Udio, I found it's pretty viable for podcast / radio ads due to the short but quality nature of what it generates. Enjoy this completely normal ad for McDonald's. https://www.udio.com/songs/eNkqEZyrvABxpoo8GqHsx7 Soulhunter fucked around with this message at 15:55 on Apr 15, 2024 |
|
#
?
Apr 15, 2024 15:53
|
|
|
             
|
|
#
?
Apr 15, 2024 16:22
|
|
|
I can dig it
|
|
#
?
Apr 15, 2024 16:28
|
|
|
  
|
|
#
?
Apr 15, 2024 18:12
|
|
|
maxwellhill posted:did you try providing what the last line is asking for? I don't know how. In Stability Matrix you basically just go to the package screen and press the launch button, and it does all the rest for you. Or at least it used to, in that it seemed to just breeze through the event log no problem. Edit: I don't understand why it's asking for a model and sampler straight out of the gate when usually that's all taken care of in the A1111 dialogue boxes once the WebUI pops up.  Edit 2: I'll try taking the commands out of the launcher and doing it all manually in cmd or powershell or something after work tonight. Unfortunately the launcher doesn't let me interact directly with it in a cmd.exe style way that lets me invoke the man page or get any additional clarification, as it's basically just spitting out a read-only log at me, and using cmd for anything is usually the last choice I'll willfully make so I would rather *not* have to start loving around in there if there's any way to avoid it. XYZAB fucked around with this message at 19:32 on Apr 15, 2024 |
|
#
?
Apr 15, 2024 19:18
|
|
|
Commie Lasorda posted:https://twitter.com/AngstHarrier/status/1779667525066015180 Commie Lasorda posted:Ugh I can't stop playing with this stupid poo poo, it's keeping me up late
|
|
#
?
Apr 15, 2024 20:33
|
|
|
Jumping on the Udio train, I have created several tracks about mountain dew Hardcore Thunder's End Thrash Monarch of the Dew Emolectronic pop Kings of Cheap Am I making music? No. Am I having fun expressing my stupid tastes musically? Hell yes.
|
|
#
?
Apr 15, 2024 20:47
|
|
|
DJ Fuckboy Supreme posted:Emolectronic pop Countdown until Kevin Barnes sues the absolute poo poo out of Udio. At least with Suno there was some sort of ambiguity where most of the training data were coming from. This is so blatantly of Montreal-coded that I can almost pinpoint the songs they used in their model. Dismemberment Plan is in there, too. Which brings up a strange consideration that I don't fully know how to articulate. It's similar to the concerns raised about visual artists having their art stolen and used in training data without their consent and generally includes ALL of those exact same arguments, I just have a sense that it's much more apparent in music: If you're training your audio models on massively famous, incredibly unique, talented individuals, how do you not make it blatantly obvious whose work you used to train those models? You trained it on those artists because they sound they way they do, in order to recreate the sound that they do. It's more ambiguous within the visual medium, and possibly more easy to ignore if you'd rather not think about it, but with music, when you recognize someone's voice, you recognize their voice. Now, do we witness artists embrace this technology as we did when Aerosmith gave Run DMC the thumbs up with their sampling of Walk This Way in the 80s that signaled a sea change in the public discourse? Is Kevin Barnes going to walk out in the middle of the music video for Udio's next big groundbreaking, culture-unifying hit song about Mountain Dew of all things? The future is loving wild for this. Edit: Reconsidering my assumption here. I failed to consider that I may be biased more toward a familiarity with music than visual art (music degree, former college radio music director), and that this might skew my ability to accurately call out sources in one medium versus the other, in a way that someone more familiar with visual art and artists might recognize recompiled visual elements in the way we've been witness to two years of development in the realm of visual art generation. It might be equally as offensively apparent to them whose works they are seeing remixed in front of them as it is to me upon hearing the voice of someone being used in a song I just generated. XYZAB fucked around with this message at 04:10 on Apr 16, 2024 |
|
#
?
Apr 16, 2024 00:48
|
|
|
XYZAB posted:If you're training your audio models on massively famous, incredibly unique, talented individuals, how do you not make it blatantly obvious whose work you used to train those models? there is already a very common practice in the marketing world of hiring composers to make music that sounds almost like a specific famous artist or song, usually just close enough to the line to not get sued, though every once in a while the line is crossed and someone gets caught. but those times are rare, for the most part there are already massive libraries of music that sound extremely similar to but are not various famous composers and musicians. so while it's one thing to listen to this stuff and identify as a listener that they are training it on x, y or z, it's another thing to prove in court that they are definitively training it on that material and not on one of many human-made soundalikes that also already exist. Earwicker fucked around with this message at 03:00 on Apr 16, 2024 |
|
#
?
Apr 16, 2024 02:57
|
|
|
Earwicker posted:there is already a very common practice in the marketing world of hiring composers to make music that sounds almost like a specific famous artist or song, usually just close enough to the line to not get sued, though every once in a while the line is crossed and someone gets caught. but those times are rare, for the most part there are already massive libraries of music that sound extremely similar to but are not various famous composers and musicians. so while it's one thing to listen to this stuff and identify as a listener that they are training it on x, y or z, it's another thing to prove in court that they are definitively training it on that material and not on one of many human-made soundalikes that also already exist. I have to admit I find it concerning that the discourse surrounding AI generated media is always couched in terms of its overall usefulness being nothing more than a shell game whose amalgam of parts that constitute the whole are meant to stay hidden from us forever and we're just supposed to be okay with that. And then I remember that not only do I enjoy the occasional hotdog cooked over a charcoal grill on a hot summer day, but major corporate entities like Facebook, Tik Tok, Google et al., aren't required by law to divulge their algorithms either, and it kind of pisses me the hell off. More in terms of that the algorithms that have been weaponized against society as a whole since ~2007 seem to have set a precedent that allows for anyone to get away with whatever the hell they want in the name of digital innovation, whenever the hell they want, in ways that completely ignore and make wholly absurd via plausible deniability any claims of possible copyright malfeasance (or other worse/illegal activity) from anyone whose work very realistically may have been folded into the training data. "There's no way to know," and "that could feasibly be a band that sounds exactly like of Montreal" are easy off-the-cuff rebuttals to make that don't fully satisfy the criticism though, because although I'm aware that the market for music-that-sounds-close-to-but-isn't-that-band is very real, I question the availability of that kind of use-specific, utilitarian, contract-based music that somebody spent a lot of real money on and this time has NO obligation to make publicly/freely available in the quantity required to train entire models, at a price-point that makes Udio able to let me and anyone else generate 200 songs per day for free with zero ads. Sure, it COULD be a band that sounds just like of Montreal, but saying that and proving that takes the argument full-circle in the other direction of "I still want to know how the soup is made, what are the ingredients?" And for the sake of this argument it might just as well be everything that Spotify had available for the low cost of $10 and took less than a month to compile as much as it might be Calvin Bjorns' for-hire studio group "of Shawinigan." Virtue signalling over. Look at me, arguing against the use of AI tools as if I didn't just buy a 40 series card to abuse the ever-loving poo poo out of this technology at a price-point that I can barely justify, instead of spending that money on real, tangible art from real artists who spend lifetimes honing their craft.  And as for my problem with the Forge package not loading; I went on a rampage installing a bunch of models, loras, extensions, textual inversions, and other crap that I didn't understand (in an attempt to understand it), and one of those extensions (ai-toolkit) basically broke the launcher. I've deleted the ai-toolkit folder and the A1111-Forge webui now loads as it did previously. Hooray. Time to make stupid pictures with the stolen labor of ten thousand individual lifetime achievements in art and photography! 
|
|
#
?
Apr 16, 2024 03:59
|
|
|
I spent a long time -- like hours flew by -- playing around with Udio. I know AI has problems, but this is just absolutely amazing, almost magical stuff. I'm not familiar with a metal album about Rime of the Ancient Mariner, I'd sure like one! I feel like a fuckin kid with infinite power. I ask the AI to make me a song, here are the lyrics, and the AI comes back and is like, Sir we got a symphony together and some of the performers from your favorite bands and we put this together for you, what do you think? And I'm like, No, not good enough. Do it again. Hmmm still not good enough, make it more piratey. Yes, that'll do. Get some opera singers. Anyway, I (ME I DID THIS ME) finished making the entire poem into a metal concept album in twelve tracks.  https://www.udio.com/playlists/vPwV8HGvCmn1NhEp3sm3ah I know I'm not an artist by any means for doing this, but it's not entirely without some authorship. I feel rather like a kid making a collage out of other people's paintings. There's direction here, is what I mean. And it sounds incredible. I finally got what I always wanted, a rad metal album about a poem about a jackass who shoots a bird and gets his comeuppance for it and then some. https://voca.ro/13bu9EjtAMPs Here's the first track. You know what I've noticed about Udio and metal, though, is that the voice gets noticeably more gravely as the song goes on. Every time, if the voice starts out smooth (the above example, he starts off gravely but sounds like hell by the end) by the end it sounds like poo poo. Not like AI bad or distorted, just like the way a guy's voice sounds after he's been singing all night long. But seriously you guys this album is so good. credburn fucked around with this message at 04:13 on Apr 16, 2024 |
|
#
?
Apr 16, 2024 04:07
|
|
|
XYZAB posted:And for the sake of this argument it might just as well be everything that Spotify had available for the low cost of $10 and took less than a month to compile as much as it might be Calvin Bjorns' for-hire studio group "of Shawinigan." it might be, but sadly even if so i suspect the end results will ultimately be some form of deal cut between the record labels and AI companies, rather than individual lawsuits by artists making much of a dent in this stuff. that's what happened during the previous big change brought by streaming, there were a few famous artist-led lawsuits/crusades early on such as the one led by metallica in the napster days, which was widely mocked and didn't stop anything, and then what happened in the end was the labels working it out with companies like spotify and youtube, so that the execs on both sides continued to get richer while most artists outside of the top 40 got hosed. i think it's pretty likely a similar situation will unfold with this stuff. i tried asking Udio to imitate my own work and one of the more well known podcasts i've worked on and one of the examples came close enough to sound like it could have been a modification of some of my source material, but it would be hard to prove.
|
|
#
?
Apr 16, 2024 04:10
|
|
|
Earwicker posted:it might be, but sadly even if so i suspect the end results will ultimately be some form of deal cut between the record labels and AI companies, rather than individual lawsuits by artists making much of a dent in this stuff. that's what happened during the previous big change brought by streaming, there were a few famous artist-led lawsuits/crusades early on such as the one led by metallica in the napster days, which was widely mocked and didn't stop anything, and then what happened in the end was the labels working it out with companies like spotify and youtube, so that the execs on both sides continued to get richer while most artists outside of the top 40 got hosed. i think it's pretty likely a similar situation will unfold with this stuff. Yeah, that's the end of the day on this. Proposed expansions of copyright or restrictions on AI, if successful, will definitely affect how much corporate rent-seeking is involved in using the technology and which particular corporations hold their hands out when you click the button. But it probably won't be in any way that gives non-famous artists much more money or job security. Still, that phenomenon is older than people losing their jobs to computers.
|
|
#
?
Apr 16, 2024 04:24
|
|
|
Earwicker posted:it might be, but sadly even if so i suspect the end results will ultimately be some form of deal cut between the record labels and AI companies, rather than individual lawsuits by artists making much of a dent in this stuff. that's what happened during the previous big change brought by streaming, there were a few famous artist-led lawsuits/crusades early on such as the one led by metallica in the napster days, which was widely mocked and didn't stop anything, and then what happened in the end was the labels working it out with companies like spotify and youtube, so that the execs on both sides continued to get richer while most artists outside of the top 40 got hosed. i think it's pretty likely a similar situation will unfold with this stuff. Unfortunately I also do not see an alternate path, so long as governments are not beholden to their constituents and instead can be bought and paid for by those corporate entities with vested commercial interests. Are you able/willing to share what podcast you work on and Udio's response for comparison? No worries if not. As for myself, I was kind of blown away with how genuinely poetic a spontaneous poem that I didn't ask for at the beginning of this attempt at generating a "person talking backwards" standup routine came off, that I had to google to make sure it wasn't wholly ripping someone off verbatim. I didn't get any results. That entire clip comes off like some well-known poet-laureate dropping some insane knowledge in front of a football stadium, before immediately detouring into gibberish that everybody seems to understand. It's bizarre, but like all of this stuff, the satisfaction I get comes from how little I have to put into its abuse in order to obtain something totally bizarre and off-the-wall yet otherwise seemless and perfect in its absurdity, which may not be the same satisfaction that others seek to obtain from this technology.
|
|
#
?
Apr 16, 2024 04:31
|
|
|
I'm having too much fun with Udio pranking your bullies is cool and good Lana Del Rey has conflicted feelings about hot dogs Snack Palace Highs Nancy Reagan Got Into A Fight With My Dog
|
|
#
?
Apr 16, 2024 04:38
|
|
|
Can you get Udio do to church sermons? Religion on demand?
|
|
#
?
Apr 16, 2024 04:39
|
|
|
Hadlock posted:Can you get Udio do to church sermons? Religion on demand? Well gently caress me sideways I know what prompt I'm using next
|
|
#
?
Apr 16, 2024 05:08
|
|
|
|
| # ? May 28, 2024 22:56 |
|
|
DJ Fuckboy Supreme posted:Well gently caress me sideways I know what prompt I'm using next  Edit: bonus points if the prompt includes "1890s Georgia Southern Baptist preacher" and "you reap what you sow"
|
|
#
?
Apr 16, 2024 05:13
|
|