AI Waifus and You 101

[GOD'S STRONGEST BUILDERBEAR]

There should totally be Sleepy lora

that said heres few attempts at Sleepy cause she birthday

Spoiler

 

[Azehara]

I should do Sleepy, haven't done her in a while. I have been playing around just prompting shit on NAI without any inpainting or images as base.

Spoiler: Dizzy Koinushi and Serafi

 

[Clem the Gem]

Guess that's one advantage of simpler character designs. Easier for everyone - including AI - to draw.

I haven't forgotten about that Mozumi LORA I was going to upload "tonight" a week and half ago. Infact I've been wasting days of my life trying to get the different outfits working like I mentioned before, but think I'm about ready to just pick the least worst of the bunch and come back to it one day if I ever discover the secret formula.

 

[Watamate]

Guess that's one advantage of simpler character designs. Easier for everyone - including AI - to draw.

I haven't forgotten about that Mozumi LORA I was going to upload "tonight" a week and half ago. Infact I've been wasting days of my life trying to get the different outfits working like I mentioned before, but think I'm about ready to just pick the least worst of the bunch and come back to it one day if I ever discover the secret formula.

You're a stronger man than I. I've been trying to train a Pochi Iida LORA, and I've given up at this point. There's a hint of her art style, but not enough for me to actually be happy with it. Disregard the LORA names, they're still different iterations each picture.



And just some nice gens from my current obsession with a specific LORA, not using my Pochi LORA.

 

[Clem the Gem]

You're a stronger man than I. I've been trying to train a Pochi Iida LORA, and I've given up at this point. There's a hint of her art style, but not enough for me to actually be happy with it. Disregard the LORA names, they're still different iterations each picture.



And just some nice gens from my current obsession with a specific LORA, not using my Pochi LORA.

Maybe I'm just terribly uncultured, but I flicked through her Pixiv page very quickly and had a hard time actually noticing a theme or style shared by all the pictures, so I could see why the AI would be having trouble.
I have not tried making a style LORA yet (except by mistake with my character ones) so I can't comment much on that.
I've been referencing these sites which may hlep..

Making a Lora is like baking a cake. | Civitai

First of all this was originally a guide I created as requested by a discord Channel(it has changed quite a bit), it is also posted in rentry (That...

civitai.com

Essential to Advanced Guide to training a LoRA | Civitai

1. Introduction "Fear the curses that hide in your training" - Disclaimer: I won't teach you to make images like this one, don't worry. This aims t...

civitai.com

Stable Diffusion LoRA Models: A Complete Guide (Best ones, installation, training) - AiTuts

Definitive guide to using and training LoRAs for Stable Diffusion. Learn how to use LoRAs, which LoRAs are popular, and how to create them.

aituts.com

Hmm, that specific LORA you mentioned might work nicely with the animation stuff I showed before..

 

[PleaseCheckYourReceipts]

You're a stronger man than I. I've been trying to train a Pochi Iida LORA, and I've given up at this point. There's a hint of her art style, but not enough for me to actually be happy with it. Disregard the LORA names, they're still different iterations each picture.



And just some nice gens from my current obsession with a specific LORA, not using my Pochi LORA.

You probably need to feed the LORA more of the manga. Reine's design is in her broader style but tuned for production corporate design and color palettes. So it doesn't surprise me you'd have issues. Thus, you probably need to either pick copying Reine (which its pretty good at) or her manga/line work style. Given the latter part of this post, I've got a guess what you're after.

Funny bit about design. The Nene/Mio ones work, but something about Fubuki with J-cups just seems wrong. Not as bad if you put them on Kanata, but her whole aesthetic and our experience with it just makes it seem really off.

 

[Alchemyth]

Don't mind me, just posting this little gem from /lmg/.

 

[Watamate]

Would there be an interest in a LORA compendium for Hololive members? And potentially other corpos/indies depending on how much I want to work or update it.

 

[Clem the Gem]

Would there be an interest in a LORA compendium for Hololive members? And potentially other corpos/indies depending on how much I want to work or update it.

Well, I think just about every Holo member already has at least one high quality LORA up on Civitai which are pretty easy to find. But I guess an organised list could be helpful since Civitai's search is a bit shitty sometimes.

I did plan to try making a bunch of the lesser known vtubers (and some well known who don't have any yet) before I ran into problems.

 

[Watamate]

Well, I think just about every Holo member already has at least one high quality LORA up on Civitai which are pretty easy to find. But I guess an organised list could be helpful since Civitai's search is a bit shitty sometimes.

I did plan to try making a bunch of the lesser known vtubers (and some well known who don't have any yet) before I ran into problems.

As I was trying to find a Lumi LORA, I found out that /VT/ already had me beat anyway.

/vtai/ LoRAs / Lycoris

FAQ | Cookbook | Proompts | Archive Corpo LoRAs l Indie LoRAs l SDXL LoRAs l LoRA Checklist l LoRA MEGA Archive TEMP LoRA Scrap List l TEMP /vtai/ thread requests Quick Update (03/07/24) Finally up to date.. >For Loras that have been made and are in good working order DO NOTE THAT THIS IS PURE...

rentry.org

/vtai/ LoRAs / Lycoris

FAQ | Cookbook | Proompts | Archive Corpo LoRAs l Indie LoRAs l SDXL LoRAs l LoRA Checklist l LoRA MEGA Archive TEMP LoRA Scrap List l TEMP /vtai/ thread requests General Information Due to hitting character limit, we have decided to move Indies onto their own rentry list. You can move to the Ind...

rentry.org

37.7 GB folder on MEGA

1225 files and 64 subfolders

mega.nz

 

[Watamate]

I've been doing random memes to see how they turn out. Some are kind of cursed.

Spoiler: Randomness

 

[The Proctor]

Curious; is there any decent A.I software that can generate semi-abstract backgrounds, logos or similar stuff?

 

[Clem the Gem]

Curious; is there any decent A.I software that can generate semi-abstract backgrounds, logos or similar stuff?

I'd imagine you could do it pretty easily with Stable Diffusion. Why not give it a try? https://civitai.com/generate

Here's "abstract background featuring birds"


Of course you would need additonal steps to make a larger image, but it's a starting point. Also there are so many models to choose from to get different styles.
Pretty sure I have seen models dedicated entirely to making logos.

 

[Clem the Gem]

Alright, I've wasted months of my life and I'm now am sick of looking at it so will just release this LORA as-is.

See it on Civitai

It feels like @God's Strongest Mozumite has been going kinda hard on the mozuposting recently, so I hope you can get some enjoyment out of this. I forget if you have been in this thread and mentioned running Stable Diffusion yourself, but if not, you can just generate images using this model right there on Civitai. Use responsibly.

Tails are still broken and the outfit tags don't do shit, so you'll need to describe the clothes in order to get the official outfits (See tags in description).

[EDIT] Just discovered anyone looking at the model while not logged in (so probably everyone) would have only seen a single picture, (the worst possible example too) since all the others have been automatically tagged with "sexy attire" and "female swimwear" and such scandalous smut is hidden by default. There doesn't appear to be a way to fix this.
You can now find all the images below:

Spoiler

 

[Watamate]

Not Vtuber related, but apparently, there's been a huge breakthrough in text to video generation. By Chrome's OpenAI team, no surprise there.

Sora: Creating video from text

openai.com

It's pretty baffling. There are still telltale signs that it's AI, but at the same time it manages to take into account crazy details like reflectivity during light scenes vs dark ones like this one in particular:



Like what...



All of the ones featured on their article are truly insane to me.

 

[Thomas Talus]

All of the ones featured on their article are truly insane to me.

I can tell that *something* feels off, but I can't articulate what.

 

[Brapfiend]

I see a lot of people nitpicking this, yet I just think of how far we've come since "The Rock Eating Rocks" AI vids. How long ago do people think that was? One year? Two?
Less than a year actually. Many people don't seem to grasp just how fast this tech is developing.

 

[Ｍｕｒｒａｙしないで]

I see a lot of people nitpicking this, yet I just think of how far we've come since "The Rock Eating Rocks" AI vids. How long ago do people think that was? One year? Two?
Less than a year actually. Many people don't seem to grasp just how fast this tech is developing.

I see people telling me to compare these videos to the eating videos, but it's hard to do that when these new videos don't feature anyone eating anything. Maybe the eating videos this thing makes are even more grotesque.

 

[Alchemyth]

I see people telling me to compare these videos to the eating videos, but it's hard to do that when these new videos don't feature anyone eating anything. Maybe the eating videos this thing makes are even more grotesque.

Does this help? In a few years we'll be prompting our own anime/movie/porn scenes. Or edit existing ones.

 

[PassiveUnaggressive]

I can tell that *something* feels off, but I can't articulate what.

That western town one doesn't have the stream at the bottom of the valley for one thing

The water overall is uncanny