Stable Diffusion, NovelAI, Machine Learning Art - AI art generation discussion and image dump

[Catgirl Tyranid]

I've been playing a lot of GURPS games with friends on Tabletop Simulator, and AI slop is perfect for making character images to use as tokens. I made my account here using one from a game we recently concluded, a campaign set during the first crusade. I wanted to go for a Tyrion Lannister archetype, but horribly inbred. It only took a few photoshop alterations to get what I had in mind. It must really suck for people who would normally take commissions for this kind of thing. What would this have even cost, and how long would it have taken to get back?


Ver archivo adjunto 7161204Ver archivo adjunto 7161205

You wouldn't have paid money for this in the first place so it doesn't actually affect the art community.

 

[Bâtard Consanguine]

You wouldn't have paid money for this in the first place so it doesn't actually affect the art community.

Lol, true. But others do, and I was just curious what the going rate and lead time for commissions are.

 

[NSJAP]

0:21

 

[Rembrandt]

Lol, true. But others do, and I was just curious what the going rate and lead time for commissions are.

Most likely severely overpriced as it'd be the only source of income for the artist.

 

[The Mass Shooter Ron Soye]

Ars Technica: Midjourney introduces first new image generation model in over a year (archive)

AI image generator Midjourney released its first new model in quite some time today; dubbed V7, it's a ground-up rework that is available in alpha to users now.

V7 works with most parameters from previous versions (--ar, --seed, etc), including users' existing --sref codes from 6.1, as well as the recently introduced personalization feature. In fact, V7 is the first Midjourney model that has personalization enabled by default, meaning users will have to train it by picking at least 200 images to build their aesthetic profile.

 

[PandaChai]

News Hams breaking into the real world, sort of

Nooz 'ams, init bruv


News'ms, pip pip

 

[The Mass Shooter Ron Soye]

Ars Technica: Google adds Veo 2 video generation to Gemini app (archive)

The company's Veo 2 video generator is rolling out to the Gemini app and website, giving paying customers a chance to create short video clips with Google's allegedly state-of-the-art video model.

0:08

 

[Delany's Donkey]

Richard III on his way to murder his nephews and take the English throne

 

[Rembrandt]

Ars Technica: Google adds Veo 2 video generation to Gemini app (archive)


Gemini-2.mp4

State of the art but still can't stop cartoon characters from talking unprompted

 

[Baraadmirer]

Well, I've pretty much moved to Illustrious now. The way it handles prompt language is much better than 1.5, though looking for a photorealistic checkpoint is still hard.

 

[Soggy Floppa]

Well, I've pretty much moved to Illustrious now. The way it handles prompt language is much better than 1.5, though looking for a photorealistic checkpoint is still hard.

Have you seriously still been using SD1.5 up until now? Just use SDXL or Flux if you want photoreal.

 

[Baraadmirer]

Have you seriously still been using SD1.5 up until now? Just use SDXL or Flux if you want photoreal.

I was concerned about VRAM issues, but with Forge I feel more comfortable with using such models even if it causes my computer to lock up for a bit. Might even think about ReForge at some point.

 

[Slav Power]

Might even think about ReForge at some point.

Speaking of, it's been discontinued as the dev's personal life has interfered with the project. If you're VRAM restricted, consider learning ComfyUI since it handles memory constrains well. I was able to run SDXL models on a 1060 with 16GB of RAM. It would offload most of it to the system RAM, run like shit and make the system unusable, but it would work instead of shitting itself.

 

[The Mass Shooter Ron Soye]

AI-generated videos now possible with gaming GPUs with just 6GB of VRAM

Lvmin Zhang at GitHub, in collaboration with Maneesh Agrawala at Stanford University, has introduced FramePack this week. FramePack offers a practical implementation of video diffusion using fixed-length temporal context for more efficient processing, enabling longer and higher-quality videos. A 13-billion parameter model built using the FramePack architecture can generate a 60-second clip with just 6GB of video memory.

 

[Rembrandt]

AI-generated videos now possible with gaming GPUs with just 6GB of VRAM

I remember saying a few pages ago that it used to take >40GB to do video gen. It's exciting to see this tech progress and improve in real time.

 

[Overly Serious]

Ars Technica: Midjourney introduces first new image generation model in over a year (archive)


Ver archivo adjunto 7177919Ver archivo adjunto 7178021

Is it easy to have multiple profiles with Midjourney? Because the AI learning to do one particular style it thinks I prefer would be one of the last things I would want. I want to be able to change the style I want from image to image.

 

[Potatoherder]

Civitai is starting to do a massive crackdown on content and the owner confirms it's payment processors pushing it.

https://civitai.com/articles/13632

edit:

https://ghostarchive.org/archive/Zb0ez

 

[AnOminous]

Civitai is starting to do a massive crackdown on content and the owner confirms it's payment processors pushing it.

Payment processors get the rope.

 

[Delany's Donkey]

Civitai is starting to do a massive crackdown on content and the owner confirms it's payment processors pushing it.

https://civitai.com/articles/13632

Ver archivo adjunto 7267868
edit:

https://ghostarchive.org/archive/Zb0ez

This is 9-11 for menstrual blood diaper piss fetishists.

 

[Rembrandt]

AI-generated videos now possible with gaming GPUs with just 6GB of VRAM

I highly recommend using this addon on top of Framepack, adds some much needed boosts to generation speed and some QoL additions to the UI. Despite what is says on the tin the software is weird when it comes to utilizing RAM and GPU VRAM, this helps alleviate the issue. Went from 15-30 seconds per iteration down to 2 as it should have been.