AI Waifus and You 101

[Azehara]

My main issue with all of the examples posted above is that they all have that plastic shine to it that makes it easy to identify that its Ai. I feel like in that aspect Novel can better replicate a more natural feel to the images (except random hair and hand fuckery). I do like how the desktop ui seems to be way more free when it comes to nailing down poses and/or even basic photo composition. Hell, its nearly impossible to have 2 clearly defined subjects doing a very specific action unless you in paint and manipulate an image manually on photoshop. Recreating some meme poses can be extremely hard on Novel.

I think its main strength lies from being able to create/design character concepts.

 

[Clem the Gem]

Does this count as AI art? Probably old news to everyone, but I recently have been having fun with something called Suno where you can generate AI music complete with vocals. Either feed it your own lyrics directly and tell it what kind of music you want; or even more fun is to give it a short prompt to work with. It knows stuff too. When I mentioned "the streamer Gawr Gura" in my prompt for example, it was giving lyrics about the ocean, shark girl, playing games and so on.

Custom lyrics (2 variations):

The Pepeloni by @clemthegem | Suno

70s rock song. Listen and make your own with Suno.

suno.com

The Pepeloni by @clemthegem | Suno

70s rock song. Listen and make your own with Suno.

suno.com

AI generated:

Nolan's Obsession by @clemthegem | Suno

folk rock song. Listen and make your own with Suno.

suno.com

Virtual Desperation by @clemthegem | Suno

80s heavy rock song. Listen and make your own with Suno.

suno.com

Spoiler: The prompt:

"the tale of Nolan Crush, a man obessed the streamer Gawr Gura. He got banned from the forum The Virtual Asylum for being too autistic."

(I really wanted to do a Mexican mariachi or some shit to better appeal to the forum demographic, but couldn't get anything I was happy with)

With a free signup you get 50 credits that reset daily. Each generation costs 10 credits and gives you 2 variations each time.
Pro Tip: You can use custom mode to enter a prompt and generate lyrics at no cost until you get something you like before generating the full song.


In other news, I released a PonyXL version of my super popular world famous Onolumi LORA:

Onolumi - 3 Outfits - Pony | Stable Diffusion LoRA | Civitai

NEW PonyXL Version: Better looking and much more consistent when prompting for different outfit variations. Does struggle somewhat with the face wh...

civitai.com

 

[Watamate]

Does this count as AI art? Probably old news to everyone, but I recently have been having fun with something called Suno where you can generate AI music complete with vocals. Either feed it your own lyrics directly and tell it what kind of music you want; or even more fun is to give it a short prompt to work with. It knows stuff too. When I mentioned "the streamer Gawr Gura" in my prompt for example, it was giving lyrics about the ocean, shark girl, playing games and so on.

Custom lyrics (2 variations):

The Pepeloni by @clemthegem | Suno

70s rock song. Listen and make your own with Suno.

suno.com

The Pepeloni by @clemthegem | Suno

70s rock song. Listen and make your own with Suno.

suno.com

AI generated:

Nolan's Obsession by @clemthegem | Suno

folk rock song. Listen and make your own with Suno.

suno.com

Virtual Desperation by @clemthegem | Suno

80s heavy rock song. Listen and make your own with Suno.

suno.com

Spoiler: The prompt:

"the tale of Nolan Crush, a man obessed the streamer Gawr Gura. He got banned from the forum The Virtual Asylum for being too autistic."

(I really wanted to do a Mexican mariachi or some shit to better appeal to the forum demographic, but couldn't get anything I was happy with)

With a free signup you get 50 credits that reset daily. Each generation costs 10 credits and gives you 2 variations each time.
Pro Tip: You can use custom mode to enter a prompt and generate lyrics at no cost until you get something you like before generating the full song.


In other news, I released a PonyXL version of my super popular world famous Onolumi LORA:

Onolumi - 3 Outfits - Pony | Stable Diffusion LoRA | Civitai

NEW PonyXL Version: Better looking and much more consistent when prompting for different outfit variations. Does struggle somewhat with the face wh...

civitai.com

Damn that site is amazing...

Banned and Branded by @tatsunoko | Suno

electric melodic midwest emo song. Listen and make your own with Suno.

suno.com

 

[PassiveUnaggressive]

Damn that site is amazing...

Banned and Branded by @tatsunoko | Suno

electric melodic midwest emo song. Listen and make your own with Suno.

suno.com

Amazing AI cover. It makes me want to sing it's so good. I hope they throw it in the OP if they get around to a KF page about him

 

[God's Strongest Dragoon]

It was announced yesterday that Stable Diffusion 3 will be released on the 12th.


Civitai already has a filter option for "SD 3", who knows how quickly people will leap at it to start pumping out tons of custom LORAs and shit. It took Pony v6 for people to finally make the leap from 1.5 to XL. Seeing as how XL's base model was about 3.5B parameters and 3.0 is suppose to be around 8B parameters, hopefully it will be powerful enough to convince people. Or it's just going to take a group of autists/coomers making a specialized checkpoint model for people to fuck around with.

 

[Clem the Gem]

It was announced yesterday that Stable Diffusion 3 will be released on the 12th.


Civitai already has a filter option for "SD 3", who knows how quickly people will leap at it to start pumping out tons of custom LORAs and shit. It took Pony v6 for people to finally make the leap from 1.5 to XL. Seeing as how XL's base model was about 3.5B parameters and 3.0 is suppose to be around 8B parameters, hopefully it will be powerful enough to convince people. Or it's just going to take a group of autists/coomers making a specialized checkpoint model for people to fuck around with.

Huh, I didn't think AMD had anything to do with AI. Thought it was all Nvidia. Infact I recall when first getting into SD, the setup guides were pretty much "If you've got an AMD card, it basically doesn't work, but here's some tweaks and hacks you can try".

Anyway, looking forward to having to buy a bigger graphics card and even more storage to run SD3.
Or I guess maybe getting better performance than SDXL, that would be cool too.

 

[Alchemyth]

Huh, I didn't think AMD had anything to do with AI. Thought it was all Nvidia. Infact I recall when first getting into SD, the setup guides were pretty much "If you've got an AMD card, it basically doesn't work, but here's some tweaks and hacks you can try".

Anyway, looking forward to having to buy a bigger graphics card and even more storage to run SD3.
Or I guess maybe getting better performance than SDXL, that would be cool too.

They don't. AMD is a joke. Their new w7900 dual slot 48Gb for Ai workspaces is literally, LITERALLY dogshit. Less memory bandwith then a 3090, no flash attention support and good luck dealing with ROCM. For the price they're asking, you could get 6 3090's for a whopping 144gb of vram.

 

[httn]

Huh, I didn't think AMD had anything to do with AI. Thought it was all Nvidia. Infact I recall when first getting into SD, the setup guides were pretty much "If you've got an AMD card, it basically doesn't work, but here's some tweaks and hacks you can try".

Anyway, looking forward to having to buy a bigger graphics card and even more storage to run SD3.
Or I guess maybe getting better performance than SDXL, that would be cool too.

They are still massively behind Nvidia on AI hardware. However they are trying hard to pushback and are somewhat bending to AI companies wishes. They are in the process of open sourcing more of their driver and ROCM software stack in an effort to try and woo devs & companies. So currently shit, but worth keeping an eye on in the next year or two.

 

[Clem the Gem]

"So, what kind of music are you in to?"

"I listen exclusively to AI-generated heavy metal songs about pepeloni pizza. You probably haven't heard of them though."

---

A playlist for your pleasure:

A Pleasing Playlist of Pepeloni Performances by @clemthegem | Suno

Aah, the Peppeloni. You know the pepeloni? The no one

suno.com

Suno had an update a few days ago. Aparently the AI music/lyrics have been improved, and also the standard length increased to 4 minutes. I've found in my examples that the lyrics I entered might only last for a minute, then the AI will add another minute or so with filler. It could be a guitar solo, a slow interlude, repeating the same lyrics again, or mixing parts of lyrics together.

Also, can I just point out the auto-generated artwork for this one? It's a pizza, except some of the pepelonis look sort of like skulls. That's pretty metal if you ask me.

Anyway, I want to hear what others have made. I guess V-tuber related to stay on topic.

 

[Alchemyth]

"So, what kind of music are you in to?"

"I listen exclusively to AI-generated heavy metal songs about pepeloni pizza. You probably haven't heard of them though."

---

A playlist for your pleasure:

A Pleasing Playlist of Pepeloni Performances by @clemthegem | Suno

Aah, the Peppeloni. You know the pepeloni? The no one

suno.com

Suno had an update a few days ago. Aparently the AI music/lyrics have been improved, and also the standard length increased to 4 minutes. I've found in my examples that the lyrics I entered might only last for a minute, then the AI will add another minute or so with filler. It could be a guitar solo, a slow interlude, repeating the same lyrics again, or mixing parts of lyrics together.

Also, can I just point out the auto-generated artwork for this one? It's a pizza, except some of the pepelonis look sort of like skulls. That's pretty metal if you ask me.

Anyway, I want to hear what others have made. I guess V-tuber related to stay on topic.

https://litter.catbox.moe/kpcown.wav
Impressive but after 2 minutes it went a little schizo with mixing and matching lyrics, so I edited it out. Still really cool.

Lyrics:
[Verse] In the depths of the web we roam / A place where we all call home / Autists and schizos unite / Talking Veetubers day and night
[Verse 2] Drama unfolds at the speed of light / Who's right and who's caught uptight / Screens flicker and usernames glow / In the asylum all secrets flow
[Chorus] Welcome to the talking asylum / Where the chatters never tire / Veetubers and drama never end / In the asylum we're all friends
[Verse 3] Rants and memes are our daily bread / Wild thoughts and rats racing in each thread / No shame no silence no pause / In the asylum shitposting's our cause
[Bridge] Laugh and cry at virtual faces / In our digital hiding places / Hearts connected through the screen / In the asylum idol dreams are seen
[Verse 4] Tales of streams and fanarts bright / Shared across this eternal night / A mad world spinning out of sight / In the asylum we ignite

 

[Watamate]

"So, what kind of music are you in to?"

"I listen exclusively to AI-generated heavy metal songs about pepeloni pizza. You probably haven't heard of them though."

---

A playlist for your pleasure:

A Pleasing Playlist of Pepeloni Performances by @clemthegem | Suno

Aah, the Peppeloni. You know the pepeloni? The no one

suno.com

Suno had an update a few days ago. Aparently the AI music/lyrics have been improved, and also the standard length increased to 4 minutes. I've found in my examples that the lyrics I entered might only last for a minute, then the AI will add another minute or so with filler. It could be a guitar solo, a slow interlude, repeating the same lyrics again, or mixing parts of lyrics together.

Also, can I just point out the auto-generated artwork for this one? It's a pizza, except some of the pepelonis look sort of like skulls. That's pretty metal if you ask me.

Anyway, I want to hear what others have made. I guess V-tuber related to stay on topic.

I wanted to make one for every Myth member, giving each one their own distinct style. If people are interested I might do more for other gens.

Gura:

The Abyssal Queen Returns by @tatsunoko | Suno

swedish power metal aggressive symphonic song. Listen and make your own with Suno.

suno.com

Ame:

Tick Tock The Clock by @tatsunoko | Suno

playful American ska punk, lots of trumpets song. Listen and make your own with Suno.

suno.com

Kiara:

Phoenix Reborn by @tatsunoko | Suno

English K-pop song, female vocals, electro pop, fast paced, Drum and bass, bass drops song. Listen and make your own with Suno.

suno.com

Calli:
V1

Reaper's Ride by @tatsunoko | Suno

Spaghetti Western, Electro Swing, Rap, Drum & Bass, female vocals, Piano, song. Listen and make your own with Suno.

suno.com

V2

Reaper's Ride by @tatsunoko | Suno

Spaghetti Western, Electro Swing, Rap, Drum & Bass, female vocals, Piano, whistling song. Listen and make your own with Suno.

suno.com

Ina:
V1

Ina's Shadows by @tatsunoko | Suno

Post-hardcore alternative metal in the style of Unravel by TK, Female Vocals, Gothic, Reverb song. Listen and make your own with Suno.

suno.com

V2

Ina's Shadows by @tatsunoko | Suno

Post-hardcore alternative metal in the style of Unravel by TK, Female Vocals, Gothic, Violin, Reverb song. Listen and make your own with Suno.

suno.com

 

[Clem the Gem]

I wanted to make one for every Myth member, giving each one their own distinct style. If people are interested I might do more for other gens.

Gura:

The Abyssal Queen Returns by @tatsunoko | Suno

swedish power metal aggressive symphonic song. Listen and make your own with Suno.

suno.com

Ame:

Tick Tock The Clock by @tatsunoko | Suno

playful American ska punk, lots of trumpets song. Listen and make your own with Suno.

suno.com

Kiara:

Phoenix Reborn by @tatsunoko | Suno

English K-pop song, female vocals, electro pop, fast paced, Drum and bass, bass drops song. Listen and make your own with Suno.

suno.com

Calli:
V1

Reaper's Ride by @tatsunoko | Suno

Spaghetti Western, Electro Swing, Rap, Drum & Bass, female vocals, Piano, song. Listen and make your own with Suno.

suno.com

V2

Reaper's Ride by @tatsunoko | Suno

Spaghetti Western, Electro Swing, Rap, Drum & Bass, female vocals, Piano, whistling song. Listen and make your own with Suno.

suno.com

Ina:
V1

Ina's Shadows by @tatsunoko | Suno

Post-hardcore alternative metal in the style of Unravel by TK, Female Vocals, Gothic, Reverb song. Listen and make your own with Suno.

suno.com

V2

Ina's Shadows by @tatsunoko | Suno

Post-hardcore alternative metal in the style of Unravel by TK, Female Vocals, Gothic, Violin, Reverb song. Listen and make your own with Suno.

suno.com

Nice, I think I like the Cali ones the most.

Purely AI lyrics, or your own mixed in? If I remember right, I was able to type in single lines that I liked (which I grabbed from other generations) along with the basic prompt, and it would work those lines into the song somehow.

 

[Watamate]

Nice, I think I like the Cali ones the most.

Purely AI lyrics, or your own mixed in? If I remember right, I was able to type in single lines that I liked (which I grabbed from other generations) along with the basic prompt, and it would work those lines into the song somehow.

I ended up splurging and paying 10$ for a month, which I think gives you access to the Custom mode. But mostly AI lyrics, and then making manual changes where the lyrics didn't make any sense. Along with some tweaks. All the stuff between brackets like [Drum and Guitar Buildup], [Trumpet Solo], [Whispering Acappella Verse], etc. It does a pretty good job at adhering to those, but I won't say it takes quite a few gens to get something that's good and has decent vocals.

And yeah I agree, I was really curious how it would handle Electro Swing Western, but it came out pretty well. Ina I wanted to try to get something like Unravel, but sadly I personally don't feel like I was successful with that

 

[Clem the Gem]

Oh yeah I got the monthly account a bit ago when I got tired of juggling free gmail accounts, and then blew through all 2500 credits immediately, and even topped up a few more. But credits is all it gives you. There's no cost for custom mode. Lkie I said before, you can go in custom mode and generate lyrics forever at no cost (click "Generate Lyrics" inside the Lyrics box) and then only use credits to generate the actual song once you've got your lyrics.

 

[Clem the Gem]

So, Stable Diffusion 3 has now been released on Civitai. Currently it is only available for ComfyUI, but expect an Automatic1111 version to follow in the coming days.

Looking at some of the uploaded pictures, the reaction seems a bit mixed, with one complaint in particular:

Spoiler

Even if you only want to do SFW only, I'm seeing comments saying it's going to struggle with even clothed subjects, as without the uncensored stuff, it's never going to get the underlying anatomy correct. Which I guess makes sense. There are some examples of horrific anatomy, but this could just be people not prompting proplerly.
Still, the text is pretty amazing, right?
Will be able to comment further once an A1111 version is released.

Meanwhile, the model makers:


(cobbled together during a slow day at work.. don't judge too harshly..)

 

[Alchemyth]

So, Stable Diffusion 3 has now been released on Civitai. Currently it is only available for ComfyUI, but expect an Automatic1111 version to follow in the coming days.

Looking at some of the uploaded pictures, the reaction seems a bit mixed, with one complaint in particular:

Spoiler

Even if you only want to do SFW only, I'm seeing comments saying it's going to struggle with even clothed subjects, as without the uncensored stuff, it's never going to get the underlying anatomy correct. Which I guess makes sense. There are some examples of horrific anatomy, but this could just be people not prompting proplerly.
Still, the text is pretty amazing, right?
Will be able to comment further once an A1111 version is released.

Meanwhile, the model makers:


(cobbled together during a slow day at work.. don't judge too harshly..)

Shit like this is why I'll never be worried about AI somehow destroying humanity. 'Allignment' cripples any model. If Skynet ever wakes up, just say you identify as a trans pansexual, black, muslim woman and watch it throw a pride parade in your honour.

Anyway, you get much better results in Suno if you actually force it to repeat the chorus like most songs do.

The Tale of Nolan by @minimalisticmasteringengineer875 | Suno

nu metal, alternative rock, hard rock, electric guitar, bass guitar, drums, keyboard, , haunting, song. Listen and make your own with Suno.

suno.com

 

[Clem the Gem]

The Tale of Nolan by @minimalisticmasteringengineer875 | Suno

nu metal, alternative rock, hard rock, electric guitar, bass guitar, drums, keyboard, , haunting, song. Listen and make your own with Suno.

suno.com

"Explosion chamber sealed his fate
Online fights, they terminate"

Amazing

 

[Watamate]

Running out of credits, but here's a few more.

Fauna:

Nature's Keeper by @tatsunoko | Suno

Mbira. Playful Cute Trap Music, record scratches, female vocals, kawaii, song. Listen and make your own with Suno.

suno.com

Mumei:

Namer of None by @tatsunoko | Suno

Midwest emo, female vocals song. Listen and make your own with Suno.

suno.com

Another Gura:

The Only Thing They Fear Is A by @tatsunoko | Suno

Mick Gordon, Djent, Progressive Metal, Female Vocals, Industrial Metal, Synths song. Listen and make your own with Suno.

suno.com

I also noticed that for some stupid reason, IRyS is a forbidden word for some reason.

 

[Clem the Gem]

An improvement over the last ones I think.
I'm now looking forward to the next year or so where we can just paste these prompts/lyrics into another generator to get a full-on animated music vidieo to go along with the songs.

Also new Suno update just dropped. This one lets you upload an audio clip up to a minute long to make a song from. You could record yourself smaking a bucket to make a drum beat, or say some lines, and Suno will try to make something out of it. Of of couse grab the audio from a video somewhere. This feature does require one of the paid subscriptions however.

I might be using it wrong, but I've found it hard to work with, with the prompt being pretty much ignored. Could also be my choice of audio clips are just too hard to work with.

Here's something from Miko. Sound familiar?

Untitled by @nomadicfusion800 | Suno

80s japanese city pop, female vocals song. Listen and make your own with Suno.

suno.com

Untitled by @nomadicfusion800 | Suno

chilled synthwave, japanese female vocals, guitar, song. Listen and make your own with Suno.

suno.com

Untitled by @nomadicfusion800 | Suno

heavy rock, Industrial, heavy drums, guitar, song. Listen and make your own with Suno.

suno.com

Untitled by @nomadicfusion800 | Suno

Midtempo Industrial, heavy drums, guitar song. Listen and make your own with Suno.

suno.com

Untitled by @nomadicfusion800 | Suno

rock, heavy drums, guitar song. Listen and make your own with Suno.

suno.com

faq you by @nomadicfusion800 | Suno

Midtempo Industrial, heavy drums, guitar, japanese female vocals song. Listen and make your own with Suno.

suno.com

faq you by @nomadicfusion800 | Suno

Midtempo Industrial, heavy drums, guitar, japanese female vocals song. Listen and make your own with Suno.

suno.com

I want to try something else, but I'm having a hard time thinking of good clips that I could use.

 

[Azehara]

Since we got some Nolan songs in the mix.

Ban Me, I Don't Mind by @glamorousjitter501 | Suno

melodic acoustic country song. Listen and make your own with Suno.

suno.com

Been making some Kirsche spicy AI for thumbnails.

Spoiler: Kirsche