"While we were preparing, Sena taught me what a 'glowie' was; The definition of that was so normal to me, I thought 'glowie' was something to do like when you read sexy manga, they put like a fat fucking white wad instead of the dick? that's what I thought what it was, like the censor"S.S. Isa

AI Waifus and You 101

Watamate

Previously known as Tatsunoko
Early Adopter
Joined:  Oct 8, 2022
Here's the Sniper Wolf Botan card that I mentioned I was working on in the main thread. I'm not sure if TVA attachments keep the json info so I'll add a few mirrors. I spent a lot more time on the actual image than the character definition, so feel free to make any suggestions on things I should change there.


Shishiro Botan Wolf.card.png
 

Clem the Gem

Unknown member
Early Adopter
Joined:  Sep 10, 2022
Interesting, first I'm hearing of this site and "cards", at least related to AI art. What do you do with it?

Also, Chub heheh

Oh yeah, the picture's nice too.
 

Watamate

Previously known as Tatsunoko
Early Adopter
Joined:  Oct 8, 2022
Interesting, first I'm hearing of this site and "cards", at least related to AI art. What do you do with it?

Also, Chub heheh

Oh yeah, the picture's nice too.
I've explained most of it in this post:

But a short recap and some amendments since then:
If you're familiar with Illusion (RIP) games à la Koikatsu, Honey Select. It's very similar to that. Where it's a picture that also has code attached to it. In the case of AI cards, it's mostly descriptions of the character that can then be imported into SillyTavern/TavernAI.


SillyTavern is only a frontend GUI, but there are free Horde AI models that you can use without having to jump through any hoops. But keep in mind these are fairly low quality and public, so responses might take a few seconds.

Sadly since my previous post on the subject things have gotten a bit more scarce/difficult if you want to use OpenAI GPT 3.5/4 or Claude. If all you're looking for is a chat your best bet is Clewd. Which is basically using the free Slack Claud AI model. The only real downside is it's fairly filtered and requires finagling with the Jailbreak.

Links are from 4chan, so do note that they include NSFW imagery purely to scare away Reddit and "normies":

Those links should be enough to get you up and running though if you're genuinely curious about trying it out. And for more info/keep up to date, the best resource is 4chan's /g/ board.
 

Clem the Gem

Unknown member
Early Adopter
Joined:  Sep 10, 2022
I've explained most of it in this post:

But a short recap and some amendments since then:
If you're familiar with Illusion (RIP) games à la Koikatsu, Honey Select. It's very similar to that. Where it's a picture that also has code attached to it. In the case of AI cards, it's mostly descriptions of the character that can then be imported into SillyTavern/TavernAI.


SillyTavern is only a frontend GUI, but there are free Horde AI models that you can use without having to jump through any hoops. But keep in mind these are fairly low quality and public, so responses might take a few seconds.

Sadly since my previous post on the subject things have gotten a bit more scarce/difficult if you want to use OpenAI GPT 3.5/4 or Claude. If all you're looking for is a chat your best bet is Clewd. Which is basically using the free Slack Claud AI model. The only real downside is it's fairly filtered and requires finagling with the Jailbreak.

Links are from 4chan, so do note that they include NSFW imagery purely to scare away Reddit and "normies":

Those links should be enough to get you up and running though if you're genuinely curious about trying it out. And for more info/keep up to date, the best resource is 4chan's /g/ board.
Ah OK, so it's all chat AI and nothing to do with AI art (there's just a picture to go with the character). I guess I wasn't paying attention when you said soemthing about spending time on the picture. I was at work, OK!
Yeah, I did read most of this stuff before, and yes I'm.. quite familiar with Illusion cards. I also heard they're not actually dead and just rebranded as something else. Though I haven't kept up with them in quite a while.
 

Watamate

Previously known as Tatsunoko
Early Adopter
Joined:  Oct 8, 2022
Now even though I don't think there are many people here who use SillyTavern besides myself, I figured I'd post this.

I made a World/Lorebook for Hololive members. When you use a card of a Hololive member, depending on which jailbreak you use and if it fits in the story, it can sometimes have another member show up, or chime in. And while Claude 2.0, and probably other AI models do a decent enough job at their personalities and kayfabe, it does get their appearances wrong a lot. I'm sure this is because a lot of the wikis and articles don't go into written depth about how they look besides "Fox Girl" or "Lion Girl", and the Chat AI obviously can't see pictures.

Ergo, I made the World/Lorebook. It's basically my prompts for Stable Diffusion lightly edited to be more AI-readable.


If you're unfamiliar with what World/Lordbooks are, in this case, whenever someone is named the AI will temporarily add the information that is connected to that keyword. I haven't tested it in depth yet, since I just made it. But hopefully, it will reduce stuff like randomly referring to Fubuki as a blonde, etc.

This is a base example though of how Claude 2.0 with the right Jailbreak can introduce characters in a solely Towa-focused card (entirely unprompted):

1693491789118.png
1693491856138.png


I should add, that I wasn't using a wholly accurate Towa card, it's one that made her more bully/punkish.
 
Last edited:

Faceless Waifu

prompt: sister cleaire, event horizon, no eyes
Early Adopter
Joined:  Sep 9, 2022
Gonna awaken the thread with a rather interesting bit of info for those who are into the art AI generation side of thing, around a couple weeks ago (or last month, not sure), the usual site for anything LORA/Model/Checkpoint/Lycoris/you name it for image generation stuff, Civitai, recently revealed their built-in AI art generator that you can use right there and then.

You need to have an account to gain access to it, but unlike the PixAI art generator that I shared before, it is free without any sort of credits system, although you still need to wait for queue everytime you make one with no way of getting high priority slot.

Example on how the menu look like
1695660558962.png
1695660576623.png
The menu itself looks more like the one you see in the desktop variant of stable diffusion, although I can't find advanced stuff like ControlNet or other things like HiRes option. The queue time is a bit long (I fell asleep waiting so I cant estimate how long) but the result is more or less workable.

Result while testing out Ririka (Hololive Regloss) LORA someone made in civitai
1695660797949.png
1695661070262.png 1695661080332.png

Personally I find it a bit easier to use than PixAi if only because:
- You have Civitai's array of stuff to access without issue/trouble, while PixAI relied on people to repost it to their own market or you uploading the stuff you want yourself as public or private thing in their site to use it
- Civitai's image generation can let you use more than 3 LORAs, while PixAI only lets you use 3 (unless you have membership)
- Civitai can let you make up to 10 image per generation task, PixAi is maxed at 4 image per generation task
- Civitai has no credits system so it's truly free, PixAi needs credit to reliably generate stuff (and let you use it more for higher priority generation)

But since I only used it once to generate this example, I still need to get used to Civitai one to get a hang of it. That said, it is one good alternative if you want to try generating image with AI without wondering on whether your PC/Laptop can do it or not.

Speaking about PixAI, they also has set up membership plans if one want to seriously use the place to generate AI art.

1695661524650.png
Seems rather expensive for just a month of making AI stuff.
 

Clem the Gem

Unknown member
Early Adopter
Joined:  Sep 10, 2022
I did see that, but haven't tried it. I also noticed when browsing (some?) LORAs and models, you have the option to preview them right there and then, using this same thing I assume.

Just the fact that it's free with no credits makes it better than anything else out there if you ask me. If you're OK with whatever you make being seen by anyone else of course...

I would still urge anyone interested to try a site like this as a taste of AI generation by all means, but really to run SD locally if you think you have the hardware.

I haven't really seen any other SD news lately other than SD XL apparently being the next big thing. Has anyone else tried it? So far I've been struggling to see what the fuss is about - generating images really no different to what I've been getting with SD 1.5, but taking longer to do so since you're supposed to use 1024 x 1024 as a base.
 

Faceless Waifu

prompt: sister cleaire, event horizon, no eyes
Early Adopter
Joined:  Sep 9, 2022
Just the fact that it's free with no credits makes it better than anything else out there if you ask me. If you're OK with whatever you make being seen by anyone else of course...
For the Civitai one, your results is still gonna be in the feed that only you can see unless you want it published, so no one would notice your result right there and then unless you explicitly decided to post the result right there.

And yes, it does allow you to use the Civitai's image generator to immediately test out the LORA and models with it, which is a bit annoying because that's where the previous "Download" button was beforehand.
 

Clem the Gem

Unknown member
Early Adopter
Joined:  Sep 10, 2022
For the Civitai one, your results is still gonna be in the feed that only you can see unless you want it published, so no one would notice your result right there and then unless you explicitly decided to post the result right there.

And yes, it does allow you to use the Civitai's image generator to immediately test out the LORA and models with it, which is a bit annoying because that's where the previous "Download" button was beforehand.

Oh I wasn't sure if you had the option of a private feed, but I was thinking more about how I'm sure someone somewhere behind the scenes gets to see everything that is uploaded whether it's private or not. If I'm going to try out the "AssGape Extreme Gaping" LORA, or perhaps "Futa Cock Close-Up + cum", I'd rather no one get to see it thank you very much.

Ad yeah, I've clicked that test button a few times by mistake already..
 

Watamate

Previously known as Tatsunoko
Early Adopter
Joined:  Oct 8, 2022

Clem the Gem

Unknown member
Early Adopter
Joined:  Sep 10, 2022
I was debating posting these since they are kind of cringe and very haram, but I think @Tatsunoko and I share some tastes, so here you go.


Also, browsing Civitai I came across another Vtuber I recognise, so this one's for you @God's Strongest Mozumite



And what's this? Another model that might interest you..


I can't remember if you've tried making anything yourself yet, but why not use the website posted a few posts above and give this model a spin?

[Edit] And like an idiot I forgot to include the one image I spent the most time working on and was what started this whole theme. Added to Fubuki folder.
 
Last edited:

Watamate

Previously known as Tatsunoko
Early Adopter
Joined:  Oct 8, 2022
If some of the more AI-interested members on this forum want to be handheld and dip their toes into AI Chatbots. I've recently figured out how to create my own proxy. I'll be rather picky and spots will be very limited (3~) due to not wanting to get my own key banned.

My only request is for said people to use https://www.anthropic.com/earlyaccess, with a non-obvious throwaway email (so a Gmail or an actual business-looking mail) and fill in a serious-looking application. So that in case my key does get banned, I could replace it with a working one (if you'd be inclined to let me add said key to my proxy). Do note that it can take weeks/months for them to get back to you if they accept your request.

So DM me if you're interested in it.
 
Last edited:

Clem the Gem

Unknown member
Early Adopter
Joined:  Sep 10, 2022
spots will be very limited (3~)
So enough for everyone who ever looked at this thread.

Honestly though, I'll leave a spot for someone else. While I'm interested in where chat AI is going, I just haven't really ever felt like using it. I guess I just don't know what I'd do with it.
 

node

Santa's Little Diversity Hire
Joined:  Jul 25, 2023
Does anyone know if there are any Phase LoRAs or Hypernetworks? This is about the best Panko I could coax out of bing create/Dall-e 3 and I'm willing and able to use SD to get something better.
 

Faceless Waifu

prompt: sister cleaire, event horizon, no eyes
Early Adopter
Joined:  Sep 9, 2022
Does anyone know if there are any Phase LoRAs or Hypernetworks? This is about the best Panko I could coax out of bing create/Dall-e 3 and I'm willing and able to use SD to get something better.
I think this is the Panko you seek? It's the only one I can find on Civitai.
 

node

Santa's Little Diversity Hire
Joined:  Jul 25, 2023
I think this is the Panko you seek? It's the only one I can find on Civitai.
That'll be perfect for making my double spook Panko for Halloween. Thanks a bunch! I did all the SD/Hypernetwork stuff back when rentry was the main source of info and models, I have no idea what new sites have sprung up in the last year, so that's super helpful.
 

Faceless Waifu

prompt: sister cleaire, event horizon, no eyes
Early Adopter
Joined:  Sep 9, 2022
That'll be perfect for making my double spook Panko for Halloween. Thanks a bunch! I did all the SD/Hypernetwork stuff back when rentry was the main source of info and models, I have no idea what new sites have sprung up in the last year, so that's super helpful.
No prob. Civitai last I checked has some LORAs for them Phase Connect girls, though I haven't check pipkin's name yet. Give it a try to check, but there might be other LoRas that are outside of civitai too.
 

node

Santa's Little Diversity Hire
Joined:  Jul 25, 2023
No prob. Civitai last I checked has some LORAs for them Phase Connect girls, though I haven't check pipkin's name yet. Give it a try to check, but there might be other LoRas that are outside of civitai too.
Thanks, maybe I'll use those to troll the Pipposting thread. But don't worry, now that I know about Civitai, I'll look myself.
 

Clem the Gem

Unknown member
Early Adopter
Joined:  Sep 10, 2022
That'll be perfect for making my double spook Panko for Halloween. Thanks a bunch! I did all the SD/Hypernetwork stuff back when rentry was the main source of info and models, I have no idea what new sites have sprung up in the last year, so that's super helpful.
Yep, there are a few but still pretty limited . I check every now and then to see if there's anything new.
I think the biggest change to SD since those days is the addition of the ControlNet extension, so you should check that out if it's new to you.
I wrote a short guide on the previous page if you didn't see it that should get you started.
 

node

Santa's Little Diversity Hire
Joined:  Jul 25, 2023
Thank you all very much, tons has changed, and all for the better! Early SD gens were absolute nightmares, so I'm glad that it's now comparable to professional products.
 
Top Bottom