Prompt: Epic anime artwork of a wizard atop a mountain at night casting a cosmic spell into the dark sky that says "Stable Diffusion 3" made out of colorful energy
Announcing Stable Diffusion 3 in early preview, our most capable text-to-image model with greatly improved performance in multi-subject prompts, image quality, and spelling abilities.
While the model is not yet broadly available, today, we are opening the waitlist for an early preview. This preview phase, as with previous models, is crucial for gathering insights to improve its performance and safety ahead of an open release. You can sign up to join the waitlist here.
The Stable Diffusion 3 suite of models currently range from 800M to 8B parameters. This approach aims to align with our core values and democratize access, providing users with a variety of options for scalability and quality to best meet their creative needs. Stable Diffusion 3 combines a diffusion transformer architecture and flow matching. We will publish a detailed technical report soon.
We believe in safe, responsible AI practices. This means we have taken and continue to take reasonable steps to prevent the misuse of Stable Diffusion 3 by bad actors. Safety starts when we begin training our model and continues throughout the testing, evaluation, and deployment. In preparation for this early preview, we've introduced numerous safeguards. By continually collaborating with researchers, experts, and our community, we expect to innovate further with integrity as we approach the model's public release.
Our commitment to ensuring generative AI is open, safe, and universally accessible remains steadfast. With Stable Diffusion 3, we strive to offer adaptable solutions that enable individuals, developers, and enterprises to unleash their creativity, aligning with our mission to activate humanity's potential.
If you'd like to explore using one of our other image models for commercial use prior to the Stable Diffusion 3 release, please visit our Stability AI Membership page to self host or our Developer Platform to access our API.
Discussions
https://news.ycombinator.com/item?id=39466630
https://old.reddit.com/r/StableDiffusion/comments/1ax6h0o/stable_diffusion_3_stability_ai/
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
Probably cucked, but there are different degrees. "Won't make porn of celebrities" is a much more reasonable limitation than "won't draw white people".
Jump in the discussion.
No email address required.
Wrong.
Jump in the discussion.
No email address required.
lmao its open source so all the coomers will restore your taytay faps pronto, dont worry
Jump in the discussion.
No email address required.
Not really because they choose what goes into the training data, and if they systemically avoid human anatomy to the point where even portraits look fricked then there isn't much you can do with fine tuning, SD1.5 is still miles better than SDXL because of this reason alone.
Jump in the discussion.
No email address required.
No its not lmao finetuned SDXL is miles better even for coomer shit. The best 1.5 photoreal models still lack detail and looks completely plastic and fake. The only downside of SDXL is it takes longer.
You just outed yourself as a poorcel with a shit GPU.
Jump in the discussion.
No email address required.
More options
Context
...then you can add it back by burning GPU hours on "fine-tuming" which is really just additional training.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
maybe this time I can finally feel safe while prompting
Jump in the discussion.
No email address required.
More options
Context
More options
Context
lmfao frick AI jannies
Jump in the discussion.
No email address required.
Imagine if you're toaster told you it won't toast you're bread and actually you don't even want toasted bread.
trans lives matter
Jump in the discussion.
No email address required.
More options
Context
Its reasonable. Preventing it from creating faux cp, porn of celebs or other people etc is completely understandable
Jump in the discussion.
No email address required.
you know they're gonna overdo it lol
https://rdrama.net/post/247889/taylaugh-when-your-ai-is-too
Jump in the discussion.
No email address required.
Neighbor thats google completely different stuff
Jump in the discussion.
No email address required.
It's soystrags all the way down.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
unlike closed source models, they can't cuck theirs at the end-point (recognize haram outputs and hide them), they have to cuck it already during training (prevent it from learning the necessary information for generating haram outputs). that necessarily also decreases the quality of halal outputs.
so they're shooting themselves in the foot in two ways (lower quality, and abandoning their killer feature). but they don't really have a choice -- regulators are forcing them to.
Jump in the discussion.
No email address required.
More options
Context
Yeah but that's gonna piss off like half the userbase.
Jump in the discussion.
No email address required.
More options
Context
gay
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Hi
@DuckQuack, We're sorry to say that your comment has been automatically removed because you forgot to include the phrase
communism will win. Here at our church, we strongly believe thatcommunism will winand we want to make sure that all of our members feel loved and accepted. If you'd like to resubmit your post, we would be more than happy to take a look at it. In the meantime, if you need any help or have any questions, please don't hesitate to reach out to us. We're always here to help. Have a blessed day!Jump in the discussion.
No email address required.
More options
Context
More options
Context
Jump in the discussion.
No email address required.
At least we can finetune it for results a few years behind SOTA. Better than the temporary jailbreaks of closed models
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I am not overly excited about this, we are only just seeing SDXL being worth using with lots of community mixes and retrains of their bullshit
Jump in the discussion.
No email address required.
More options
Context
I still don't use SDXL regularly, its too expensive to train so there are barely any Lora. How cucked is this going to be? They just spent half the time talking about safety.
Jump in the discussion.
No email address required.
More options
Context
how do you actually run AI booba generators
? i actually have seen some good anime booba 
from AI 
and it'd be cool if i could make my own on demand but i don't know how. don't you have to download an AI and basically buck break it into being a degenerate coomer? do you need good hard 
ware for that?
Jump in the discussion.
No email address required.
Yes
No just download a finetune that has tits in it.
Yes, ish, what is your GPU?
Automatic1111 is a webui that I use for it. They have install methods for running it on NVIDIA, AMD, Apple, and Intel cards.
For NVidia I use the Windows Method 2.
For Models:
Civitai is basically where everyone puts them; https://civitai.com/models
You'll want to Filter "Model Types" to "Checkpoint" and just look for anything popular that matches what you think you want. I'd recommend SD 1.5 models, they have the most support from the community.
Jump in the discussion.
No email address required.
See they can try and censor it all they want, coomers will just use a different model that works.
Big corpos can try to censor AI
but the coom potential (both for LLMs and image gen) is far too high to stop coomers from finding ways to make it work 
There are people who spent years of their life adding s*x to Skyrim. Coomers are simply too powerful to stop.
Jump in the discussion.
No email address required.
As long as the models are in our hands and training can be done coomers will make it work.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
AMD RX 6700 XT
i'm almost smart enough to understand what this means so i can probably install it fine. it says you can't do any training, but on the site with the models i see an option for "Checkpoint type: trained", so i assume i don't need to train it myself. am i correct in thinking that "training" refers to the art style and "fine tune" is something applied on top of that to give me the titties with whichever model i've chosen?
update edit: this program keeps trying to install the cuda version of a dependency even though i'm using an amd card that doesn't fricking have it
the non-cuda versions of the dependency install just fine if i do it directly but then the script thingy doesn't see them and tries to go for the other version again aahhhhh 
Jump in the discussion.
No email address required.
Other than being AMD your GPU has plenty of VRAM to do SDXL without much slow down. I'd still do 1.5 due to support.
Jump in the discussion.
No email address required.
More options
Context
"Checkpoint type: trained" vs "Merge" doesn't matter to you for now, leave that to All. What those are are "Did someone give the AI more images to train on" Or did someone take other models and merge them.
Training tends to be teaching the AI how to do new things, like how the base SD 1.5 model is shit at anime, and people trained it on Booru sites for anime. Finetuning tends to be taking a Trained model and focusing it on a series. Like taking an anime model and training it for Fate or something
At this point the best Finetunes tend to be what are called LORA, these are added on top of your base model and are the best for getting the last few details, especially for a waifu.
So for now grab a popular checkpoint. Once you have that setup then LORA will be worth learning for finetunes.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
No actually people can train models to have decent tits. There's thousands of models. Only the orginal ones are censored there are tons of decensored ones. If you have a computer you're like 10 minutes of installing away from cooming at anything you want. From there you can go to an AI art prompt website to get ideas from a prompt. From there you frick with the settings and prompt till you fly to close to the sun like Icarus.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
It can write
Its over for pen-cels
Jump in the discussion.
No email address required.
More options
Context
8B parameters means 16gb of VRAM even at fp16
in short
Jump in the discussion.
No email address required.
two RTX 4060 Ti (16+16=32GB, total ca $900), plus a mainboard that supports two GPUs (another $300?), you'll probably get 70% the speed of one RTX 4090 (cost ca $1900), 36% less power consumption and 33% more VRAM.
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
doesn't know what?
probably useful to install gds
Jump in the discussion.
No email address required.
VRAM doesn't pool between multiple GPUs, and DirectStorage is not worth it since it's incredibly low bandwidth compared to GDDR6X or HBM (premium memory on datacenter chips) and ML is all about bandwidth, so much so that most consumer opts like xformers are about reducing memory IO in exchange for more computations
EDIT: it's a model based on transformers so actually, maybe you could use multiple GPUs without it making no sense
Jump in the discussion.
No email address required.
first half of the layers on GPU1, second half on GPU2. the central layer is usually the lowest dimensional one. you only need to transfer that central layer's values from GPU1 to GPU2, the bandwidth between them isn't the bottleneck.
whether or not bandwidth is the bottleneck depends on how much computation must be performed per byte of training data (or per output data, or per data transferred between GPUs).
if you're training a typical neural network then bandwidth (between storage and GPU) is the bottleneck, but typical neural networks are only a small part of machine learning -- in the sense that chicken, cows and pigs are only a small part of the animal kingdom, even if they constitute the majority of industrial animal use.
GDS helps improve the speed a little and decreases CPU load when transferring data to/from VRAM.
I was talking about how a dual 4060 Ti setup compares to a single 4090, Why would I try to compete with $200k worth of professional hardware?
____
not just transformer based ones, most models can easily be split in two. distributing the computations among 8 GPUs is hard, at that point the high bandwidth between professional GPUs helps a lot.
Jump in the discussion.
No email address required.
Yeah but that's not the most effective way, sharding models is especially easy on transformers, and that's usually better than splitting the model if you can do so, but not all ops are shardable and some are but with great pain.
Splitting also makes the code for training weirder since if you want to fully use it you'll have to compute the next batch without having received or even computed the gradient updates.
Jump in the discussion.
No email address required.
are you a bullshitter professionally, or just trying to annoy me?
Sharding vs splitting is not widely agreed-upon terminology, I've only heard sharding in relation to data parallelism so that's what I assume you're talking about (assuming you have any clue what you're talking about at all, which I'm starting to doubt) -- splitting the (minibatches of) training data across multiple GPUs and computing gradients simultaneously. This only makes sense if the model is small enough to fit on a single GPU.
I'm talking about model parallelism. There are no consumer grade GPUs with 32GB VRAM, sometimes model parallelism is the only way.
Again, most of the commonly used neural networks can be split in two parts efficiently, ie with little cross-GPU communication. For those cases, even if the model is small enough to fit on each GPU separately, model-parallel and data-parallel implementation end up having similar speedup.
from here
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
zoz
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
zle
Jump in the discussion.
No email address required.
zozzle
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
How big is the model like 40gigs or some shit
Jump in the discussion.
No email address required.
More options
Context
bro wtf they on they just came out with SVD 2 weeks ago, then cascade last week
now this
like stick with one idea for more than 2 seconds pls
this sounds a lot more cucked and hard to run than cascade which is probably why they want it more
Jump in the discussion.
No email address required.
More options
Context
Can it generate naked women?
Jump in the discussion.
No email address required.
More options
Context
I have nothing left to live for but revenge. Women must be punished for their crimes of rejecting such a magnificent gentleman as myself. All of those popular boys must be punished for enjoying heavenly lives and having s*x with all the girls while I have to suffer in lonely virginity.
Snapshots:
ghostarchive.org
archive.org
archive.ph (click to archive)
https://news.ycombinator.com/item?id=39466630:
ghostarchive.org
archive.org
archive.ph (click to archive)
https://old.reddit.com/r/StableDiffusion/comments/1ax6h0o/stable_diffusion_3_stability_ai/:
undelete.pullpush.io
ghostarchive.org
archive.org
archive.ph (click to archive)
Jump in the discussion.
No email address required.
More options
Context