|
KakerMix posted:However Google has been and still continues to be pretty decent with music to my not-music-making ears, recently updated after Suno was the darling: yea music is the area where i am a "real professional" and honestly it's incredibly depressing that anyone finds this stuff listenable. not surprising at all, but still, depressing. what i really don't like about it is that at least with visual ai art you can make some weird and unexpected stuff. the ai music tools i've played around with churn out the most generic sounding goo no matter what you type in there. also it gets really basic poo poo like time signatures wrong. that said, we've had automation and sampling in music for a long time and both of those have resulted in all kinds of cool stuff, and there are some useful applications of ai in music already (noise reduction, stem isolation etc) but it seems pretty far behind the visuals and language stuff Earwicker fucked around with this message at 19:01 on Feb 3, 2024 |
# ? Feb 3, 2024 18:53 |
|
|
# ? May 30, 2024 10:34 |
|
I've been enjoying the gently caress out of the new sref feature
|
# ? Feb 3, 2024 19:23 |
Tarkus posted:I've been enjoying the gently caress out of the new sref feature Neat!
|
|
# ? Feb 3, 2024 19:36 |
|
AARD VARKMAN posted:Google's new AI art generator ImageFX is out and it can do dimetrodon KakerMix posted:I'm not really surprised, for some reason Google seems extremely far behind anything to do with "AI". I'm slowly going insane trying to make their LLMs do anything half as useful as OpenAI or Meta's offerings, and as we've seen, their image gen options are complete rear end compared to Midjourney, DALL-E, Adobe Firefly, or StableDiffusion. IT security has been dragging their feet for months about opening up access to use tools from other companies, I'm pretty sure our executive staff signed some exclusivity deal with Google against our AI team's advising. Earwicker posted:the ai music tools i've played around with churn out the most generic sounding goo no matter what you type in there. also it gets really basic poo poo like time signatures wrong. these are cool as hell. sref is my favorite new Midjourney option, second only to the --weird setting Soulhunter fucked around with this message at 19:52 on Feb 3, 2024 |
# ? Feb 3, 2024 19:38 |
|
oh my god Google these are the worst hands I have ever seen. did they only focus on training it to do high res output or something? the res is impressive but nothing else like turn down the res was this even trained for this? Oh neat POTUS Donny Two Scoops in handcuffs why does it actually do that well? e: Okay google understands some things other haven't so far... like walls covered in fingerpainted ketchup in a courtroom sketch pixaal fucked around with this message at 20:45 on Feb 3, 2024 |
# ? Feb 3, 2024 20:38 |
|
lol
|
# ? Feb 3, 2024 20:39 |
|
impossible, absolute shitshow, happy bidet
|
# ? Feb 4, 2024 01:14 |
|
I just got some beautiful art I wanna share. Snowy forest nights by the beach. ---and this as well. Vlaphor fucked around with this message at 02:12 on Feb 4, 2024 |
# ? Feb 4, 2024 02:01 |
|
Managed to get ImageFX to make two things: A couple bing: Some stupid sexy sasquatches from Midjourney Soulhunter fucked around with this message at 02:07 on Feb 4, 2024 |
# ? Feb 4, 2024 02:03 |
|
Swagman posted:impossible, absolute shitshow, happy bidet These are amazing on every possible level! Thank you oh so very much for such a wonderful birthday present! Get ahold of me for a Plat Certificate, you've earned it! KwegiboHB#5538 or KwegiboHB at gmail dot com
|
# ? Feb 4, 2024 02:04 |
|
Earwicker posted:yea music is the area where i am a "real professional" and honestly it's incredibly depressing that anyone finds this stuff listenable. not surprising at all, but still, depressing. The example given of Google's thing is just meandering nothingness but suno.ai regularly creates stuff that sounds like real music. The biggest problem I have with it is the fidelity rather than the structure of the songs or the melodies it creates.
|
# ? Feb 4, 2024 02:55 |
|
I've had a bit of a Depths kick lately, so here's some of the best ones I got Bing to spit out: And here's a few others I liked:
|
# ? Feb 4, 2024 02:58 |
|
Those are kicking rad, I am def going to try that theme. moist banana bread fucked around with this message at 08:01 on Feb 4, 2024 |
# ? Feb 4, 2024 06:17 |
|
Monster Truck Sasquatch sounds like a new film from The Asylum.
|
# ? Feb 4, 2024 10:15 |
|
Kinda surprised at how "Spiderman accidentally shooting himself in the face with his web shooter" turned out.
|
# ? Feb 4, 2024 17:02 |
|
I need to be able to have my characters do more than just stand and stare at the camera. Implied movement, implied tension between two characters, facial expressions, and so on. Is there some sort of canonical text or resource showing how great comic book artists manage this?
|
# ? Feb 4, 2024 17:19 |
|
Stupid_Sexy_Flander posted:Kinda surprised at how "Spiderman accidentally shooting himself in the face with his web shooter" turned out. was this Dalle? Because the "I can't take it anymore" feels very much like something it would pull in.
|
# ? Feb 4, 2024 17:50 |
|
I love this, it looks plausible but still vaguely anthropomorphic
|
# ? Feb 4, 2024 18:37 |
|
|
# ? Feb 4, 2024 21:02 |
|
pixaal posted:was this Dalle? Because the "I can't take it anymore" feels very much like something it would pull in. Yup, used bing.
|
# ? Feb 4, 2024 22:23 |
|
Here's a few more: I think trying to focus on the buildings more made it not want to make them as dark and atmospheric. I also put a few constructs in a museum:
|
# ? Feb 5, 2024 00:50 |
|
The negative review of Google's thing (what is it even called) is so weird to me. Shouldn't this have been in their bag with the whole "crawling the internet for a search engine" thing? I guess they weren't evil enough?
|
# ? Feb 5, 2024 03:20 |
|
what was the last "good" thing google have produced
|
# ? Feb 5, 2024 03:38 |
|
Bing Image Generator, prompt "manic pixie dream trad wife"
|
# ? Feb 5, 2024 04:47 |
|
has a harry potter house, enjoys tim burton, mcr moist banana bread fucked around with this message at 07:29 on Feb 5, 2024 |
# ? Feb 5, 2024 06:45 |
|
Christ, I just discovered there's an entire industry built around ai models and nude* pics, and people paying a crazy amount of money to them via various online services. I have no idea why I am surprised, but I am. One of em makes like 10k a month. *I didn't check out the pics so I dunno if it's like lingerie catalog levels of spicy or just straight up porn, at least for the model I was looking at, plus I didn't want to sign up and spend 5 bucks to find out. She did have the normal number of fingers though so that puts her above anything Google can generate.
|
# ? Feb 5, 2024 13:20 |
|
I've been out of the loop for a little while now, but these images are making me want to dive back in and prompt some image gen: What's the latest hotness for self hosted image generation? Is it still A11111 or Foooocus or whatever the other ones are called? And what model(s) do people recommend? Also if this is covered by a good resource available online somewhere, I'd be happy to check it out, but googling anything these days and expecting good results is a pipe dream.
|
# ? Feb 5, 2024 13:52 |
|
Google's imageFX does extremely well if you can get it to invoke a likeness everything in the image suddenly is 100 times better it even does text! POTUS Donny Two scoops in orange jumpsuit , speech bubble " Best crimes " Google has the dumbest words blocked, outfit, costume, cosplay... but I can do TSLA CEO with flaming wreckage and a blue bird very late edit: apparently there's a limit to how many images you can generate (a day?). I got an error "Maximum images generated". My impressions are Google's image AI understands some things even Dalle does not, much like Dalle 2 showing some promise but ultimately just not having enough quality to push into usefulness. I didn't really make anything else worth sharing I keep getting duplicated subjects this feels like something trained on 1024x1024 and just set to 1500 x 1500 pixaal fucked around with this message at 22:20 on Feb 5, 2024 |
# ? Feb 5, 2024 16:06 |
|
|
# ? Feb 5, 2024 23:27 |
|
going for a late-victorian grand tour vibe
|
# ? Feb 6, 2024 00:25 |
|
This is the good poo poo. Prompt / engine please.
|
# ? Feb 6, 2024 00:46 |
|
hell yeah, welcome to my skull cave lair
|
# ? Feb 6, 2024 01:02 |
|
The Phantom has some awesome vacation homes.
|
# ? Feb 6, 2024 01:44 |
|
Mola Yam posted:going for a late-victorian grand tour vibe Prompt?
|
# ? Feb 6, 2024 01:50 |
|
I like this one. I bought a plat certificate for you but you have to get ahold of me so I can give it to you. I can't just apply it to your account.
|
# ? Feb 6, 2024 02:14 |
|
feedmyleg posted:Prompt? so it's mostly just adding "grand tour photograph antique 1887 --ar 3:4" to a MJ V6 prompt. but more specifically, i got chatgpt to give me some descriptions of parallel universe grand tour destinations, and fed those in. so this is "the Great Dam of Aquatia, an impressively constructed dam that not only controls flooding but also serves as a public space with gardens and walkways, showcasing engineering prowess and environmental consciousness" and this is "the Tower of Winds, a tall, sleek tower with an intricate design that channels the wind to create harmonious sounds, acting as a symbol of the city's dedication to harmony with nature" and then i've been experimenting with feeding them through depth anything https://huggingface.co/spaces/LiheYoung/Depth-Anything and then through leiapix, replacing the default depth map with the (much better) DA one https://convert.leiapix.com/animation to create some vintage stereographs of made up monuments (these are both cross-eyed ones)
|
# ? Feb 6, 2024 02:37 |
|
You are standing on a hillside overlooking the ocean. Up the hill is a dirt path leading to a dark, solitary cabin. Or is it a house? It's hard to focus, everything seems to be shifting back and forth with each step closer. One moment, it seems to have windows, the next it's a small chapel. For a moment, you swear you see someone standing outside, beckoning you forward. You take a breath as you reach the front door and the home settles into the shape of a log cabin. It was always a log cabin, of course. Why would you have thought otherwise? You must've hit the bottle pretty hard last night, you can barely remember how you wandered up here. Inside the front door, a window to the left casts a light across the room. A small table sits to the side of the room with a stone paperweight sitting on top. In the air are a series of dangling ropes, with a small dagger suspended in the air, turning slightly as it catches a glint of sunlight. You move toward the mirror to examine yourself, and... The door on the wall suddenly moves to the right, dragging itself across the wall with the sound of a heavy chair on an old wooden floor. The light comes from a different angle - that window wasn't there before, right? You look back up at the mirror to find yourself staring at a portrait of a old woman in a hooded robe glaring at you with intensity. There was never a mirror here. You open the door and peer down the steps, shouting, "Is anyone there?" An echo of a meek "hello?" bounces off the walls back up at you. You shout back, "Do you need help? Are you OK? I'm coming down!" and cautiously proceed into the dank basement, thinking carefully to cut the dagger free from the ropes upstairs and tucking it into your belt. In the basement, you find a young woman wearing a dress, shackled to the wall. She weakly whispers out: h e l p m e She chokes on the words, struggling to force them out. You take a few steps closer to free her from the shackles, and she starts to laugh, at first quietly, then in deeper, impossible tones. You're in arm's reach, but find you cannot move - your legs are stuck firmly in place. You feel compelled to reach for your dagger by whispers in your head. A small voice. You're suddenly aware the voice has been there since you arrived outside, but somehow you tuned it out. It is screaming at you: usetheknifeusetheknifeusetheknifeusetheknife You manage to barely move your arm an inch toward the dagger before the woman swipes it from your belt with her right hand in a flash, holding it under her left arm as her cackles become more unhinged. You realize that you misread the situation from the start - she wasn't shackled to the wall at all. Yet another trick of the eye. Slight of hand to slip her wrist free from the oversized cuffs. A trap. She moves in close, suddenly, jerking and twisting unnaturally in twitching movements, brandishing the dagger. Her face contorts. "Where's your guts now, big hero?" She sticks the knife in your abdomen, eviscerating you. She smiles wider than humanly possible, giggling quietly to herself, staring into your eyes closely to watch your soul slip away, savoring the moment. As you fade out, you hear another, deeper voice in your head say, Why did you let her take the dagger? You've made a terrible mess. Maybe next time, you'll do better... A feminine voice echoes through the void. ....see you again soon, hero.... You are standing on a hillside overlooking the ocean. Up the hill is a dirt path leading to an... empty plot of land? you should play Slay the Princess Soulhunter fucked around with this message at 03:04 on Feb 6, 2024 |
# ? Feb 6, 2024 02:54 |
|
null_pointer posted:This is the good poo poo. escape-velocity,hallucinogenic psychedelic,dissolving and coagulating,style of Gregory Crewdson using newrealityxl checkpoint dropped you a line, much thanks!
|
# ? Feb 6, 2024 03:10 |
|
LordAdakos posted:What's the latest hotness for self hosted image generation? Is it still A11111 or Foooocus or whatever the other ones are called? And what model(s) do people recommend? Also if this is covered by a good resource available online somewhere, I'd be happy to check it out, but googling anything these days and expecting good results is a pipe dream. Well I've only been into the hobby for a couple months but imo having both A1111 and ComfyUI is probably good. I just started combining comfyUI workflows into a single workflow file so I can just drag-select and ctrl-m "mute" all of them, then unmute one workflow at a time. Haven't looked into it but it should be possible to use symlinks to make Auto & Comfy share model directories since those are going to get large. I'm just using SDXL, SD v1-5, and I have a few LORA's, which if you haven't tried you simply must. I have no idea though and would be interested in anyone else's recommendations. Most of the images I've posted show my workflows in comfy since I was too lazy to modify them. I don't think it will be possible to make images that are as compositionally complex as many shown here are.
|
# ? Feb 6, 2024 03:33 |
|
|
# ? May 30, 2024 10:34 |
|
Swagman posted:dropped you a line, much thanks! This brings the GBS AI Art Thread, economic powerhouse that it is, up to a grand total of $12.60 made.
|
# ? Feb 6, 2024 03:50 |