I can feel the hidden potential within the generations- it can be so beautiful, i am speechless

Perhaps Stability AI is secretly encouraging us to go talk to real women and get laid so we can know their body structure better. Truly a genius move by them

Is this going to work on my RTX 4070Ti 12GB?

just inform everyone when this thing can be used on automatic1111

Today I thought about this option: Maybe this terrible SD3 was released specifically to show everyone what happens when developers are forced to introduce censorship?))

I think I should probably cut myself short any time the words "killed by licensing" come to mind, but I think this is what we're going to see happen to this model, great as it maybe could be if it were actually allowed to grow in an open environment. So I won't. Really I think Stability is walking about half a step to the side of the right path here, but this reeeeeally isn't a great way to really pull in the money.

SDXL isn't going to easily pull in money with a license that open, SD3 seems objectively worse in concept diversity and just plain compromised when it comes understanding of how many limbs a human has and just plain isn't worth agreeing to any sort of royalty structure to host as a result. This might've been a great move about...well I'm not counting every component of those MoE models, so let's say about six models ago.

I have absolutely no qualms with a royalty structure for profit-motivated pursuits, but good luck when the terms are still clear as mud and the model isn't even as good as base SDXL in a lot of very noticeable ways. If it was a baseline good model instead of clearly underbaked and massively flawed model, people might see value in paying, but I really don't think anyone's going to bite, or at least not many. I think the license is probably stifling the continuation of this model more than the quality of the outputs.

If we don't look at the terrible anatomy, horrific bodies, mangled human composition and interactions and all the artist styles that are not there anymore the model is pretty good.

It seems not only the model is censored, but the posts on here are censored too, anything rated X or R gets completely removed from the preview page, haven't seen this in any other model before, why!?

SD3 turned out rotten to the core, it's anti-user and anti-AI, saying that its just the insane amount of censorship that kills SD3 is an understatement.

The user agreement is clearly a trap to monopolize the AI market, forcing non-commercial clause for any model using SD3, it renders SD3 a toxic poisoned swamp.

No way I am using this model for anything, compared to SD1.5 and SDXL or Pony, SD3 is an horrible downgrade

In fact, is it roughly correct to interpret SD3 Medium as "SD2 + high resolution + enhanced text expression"?

I've messed around with it for the better part of a day and I'm not impressed at all. It feels like the future of SD is with anyone BUT StabilityAI.

They want a bigger text encoder with more parameters? That's cool. They want higher dimensions on their training images? That's cool too.

But start over from scratch. 3 clip models in one model is bad, and the model can barely do 1024 decently when it's being trained on 1536 images.

I've pulled more impressive images from SD 1.x based models than anything I've seen out of SD 3.

With SD3, people who did not know the companies became familiar with them. The issue is only mandatory for your pocket. In short, do yourself a favor and forget SD3."

Why is there no relevant information about FP16? What's the difference?

уберите цензуру!!!!!

hear yi, hear yi. Boobs will come. Give it time. BOOBIES!

NO THANK YOU!

AttributeError: module 'torch' has no attribute 'float8_e4m3fn' use T5XXLFP8 , why?

Painfully disappointing for all the wait. I would have at least expected them to resolve the face, hand and feet issues with character generation but it's still an issue.

stable confusion 3.........

STABILITY AI NON-COMMERCIAL RESEARCH COMMUNITY LICENSE AGREEMENTDated: December 06, 2023 By using or distributing any portion or element of the Models, Software, Software Products or Derivative Works, you agree to be bound by this Agreement. "Agreement" means this Stable Non-Commercial Research Community License Agreement. “AUP” means the Stability AI Acceptable Use Policy available at https://stability.ai/use-policy, as may be updated from time to time."Derivative Work(s)” means (a) any derivative work of the Software Products as recognized by U.S. copyright laws and (b) any modifications to a Model, and any other model created which is based on or derived from the Model or the Model’s output. For clarity, Derivative Works do not include the output of any Model.“Documentation” means any specifications, manuals, documentation, and other written information provided by Stability AI related to the Software."Licensee" or "you" means you, or your employer or any other person or entity (if you are entering into this Agreement on such person or entity's behalf), of the age required under applicable laws, rules or regulations to provide legal consent and that has legal authority to bind your employer or such other person or entity if you are entering in this Agreement on their behalf.“Model(s)" means, collectively, Stability AI’s proprietary models and algorithms, including machine-learning models, trained model weights and other elements of the foregoing, made available under this Agreement.“Non-Commercial Uses” means exercising any of the rights granted herein for the purpose of research or non-commercial purposes. Non-Commercial Uses does not include any production use of the Software Products or any Derivative Works."Stability AI" or "we" means Stability AI Ltd. and its affiliates."Software" means Stability AI’s proprietary software made available under this Agreement.“Software Products” means the Models, Software and Documentation, individually or in any combination.1. License Rights and Redistribution.a. Subject to your compliance with this Agreement, the AUP (which is hereby incorporated herein by reference), and the Documentation, Stability AI grants you a non-exclusive, worldwide, non-transferable, non-sublicensable, revocable, royalty free and limited license under Stability AI’s intellectual property or other rights owned or controlled by Stability AI embodied in the Software Products to use, reproduce, distribute, and create Derivative Works of, the Software Products, in each case for Non-Commercial Uses only.b. You may not use the Software Products or Derivative Works to enable third parties to use the Software Products or Derivative Works as part of your hosted service or via your APIs, whether you are adding substantial additional functionality thereto or not. Merely distributing the Software Products or Derivative Works for download online without offering any related service (ex. by distributing the Models on HuggingFace) is not a violation of this subsection. If you wish to use the Software Products or any Derivative Works for commercial or production use or you wish to make the Software Products or any Derivative Works available to third parties via your hosted service or your APIs, contact Stability AI at https://stability.ai/contact.c. If you distribute or make the Software Products, or any Derivative Works thereof, available to a third party, the Software Products, Derivative Works, or any portion thereof, respectively, will remain subject to this Agreement and you must (i) provide a copy of this Agreement to such third party, and (ii) retain the following attribution notice within a "Notice" text file distributed as a part of such copies: "This Stability AI Model is licensed under the Stability AI Non-Commercial Research Community License, Copyright (c) Stability AI Ltd. All Rights Reserved.” If you create a Derivative Work of a Software Product, you may add your own attribution notices to the Notice file included with the Software Product, provided that you clearly indicate which attributions apply to the Software Product and you must state in the NOTICE file that you changed the Software Product and how it was modified.2. Disclaimer of Warranty. UNLESS REQUIRED BY APPLICABLE LAW, THE SOFTWARE PRODUCTS AND ANY OUTPUT AND RESULTS THEREFROM ARE PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING, WITHOUT LIMITATION, ANY WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. YOU ARE SOLELY RESPONSIBLE FOR DETERMINING THE APPROPRIATENESS OF USING OR REDISTRIBUTING THE SOFTWARE PRODUCTS, DERIVATIVE WORKS OR ANY OUTPUT OR RESULTS AND ASSUME ANY RISKS ASSOCIATED WITH YOUR USE OF THE SOFTWARE PRODUCTS, DERIVATIVE WORKS AND ANY OUTPUT AND RESULTS.3. Limitation of Liability. IN NO EVENT WILL STABILITY AI OR ITS AFFILIATES BE LIABLE UNDER ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, TORT, NEGLIGENCE, PRODUCTS LIABILITY, OR OTHERWISE, ARISING OUT OF THIS AGREEMENT, FOR ANY LOST PROFITS OR ANY DIRECT, INDIRECT, SPECIAL, CONSEQUENTIAL, INCIDENTAL, EXEMPLARY OR PUNITIVE DAMAGES, EVEN IF STABILITY AI OR ITS AFFILIATES HAVE BEEN ADVISED OF THE POSSIBILITY OF ANY OF THE FOREGOING.4. Intellectual Property.a. No trademark licenses are granted under this Agreement, and in connection with the Software Products or Derivative Works, neither Stability AI nor Licensee may use any name or mark owned by or associated with the other or any of its affiliates, except as required for reasonable and customary use in describing and redistributing the Software Products or Derivative Works.b. Subject to Stability AI’s ownership of the Software Products and Derivative Works made by or for Stability AI, with respect to any Derivative Works that are made by you, as between you and Stability AI, you are and will be the owner of such Derivative Worksc. If you institute litigation or other proceedings against Stability AI (including a cross-claim or counterclaim in a lawsuit) alleging that the Software Products, Derivative Works or associated outputs or results, or any portion of any of the foregoing, constitutes infringement of intellectual property or other rights owned or licensable by you, then any licenses granted to you under this Agreement shall terminate as of the date such litigation or claim is filed or instituted. You will indemnify and hold harmless Stability AI from and against any claim by any third party arising out of or related to your use or distribution of the Software Products or Derivative Works in violation of this Agreement.5. Term and Termination. The term of this Agreement will commence upon your acceptance of this Agreement or access to the Software Products and will continue in full force and effect until terminated in accordance with the terms and conditions herein. Stability AI may terminate this Agreement if you are in breach of any term or condition of this Agreement. Upon termination of this Agreement, you shall delete and cease use of any Software Products or Derivative Works. Sections 2-4 shall survive the termination of this Agreement.6. Governing Law. This Agreement will be governed by and construed in accordance with the laws of the United States and the State of California without regard to choice of lawprinciples.

can it used in webUI?

With SD v3 "a woman?" almost sounds like 'Eurika!!!'

Hello, potential finetuners! ^^/

As you might already be aware, the SD3 license is quite the hassle, and to make matters worse, there are also significant problems with censorship and anatomy. It's quite disheartening. But don’t lose heart—there are better alternatives out there! :)

Have you looked into the PixArt-Sigma and HunYuan DiT models? They're quite recent and, in my opinion, on par with SD3, without any of those troublesome issues! Plus, the PixArt-Sigma models can already be finetuned using OneTrainer from Nerogar. How cool is that? :D

i highly recommend people to avoid SD3 at all cost!. It's terrible, just stay use SDXL or SD 1.5

i am getting this message "server execution error : ' model.diffusion_model.input_blocks.0.0.weight' "

i downloaded SD3medium fp16 model

how do i resolve this issue

may i....can i say... i really don't understand, very very puzzled here...

Why is this topic so heavily censored? I mean, first two days there was bunch of great memes, jokes, etc, now all of that images are dissapeared from "official" model page. Or am i missing something?

Further, appears that SD3 is perfectly capable to create NSFW content, not worse than any other model, but, not a single image on official model page? I tried, but, it newer appeared... again, what am i missing? Is it some site rule breaking if posting such image created with SD3 or what?

To be hones, i have some computer issues these days, so maybe all of this is just problem on my end of the table, so apologies if this goes into "hallucination" drawer ;)

Censorbility.AI 🤐

Friends, if you have an RTX 2060, remember that it takes about 40 seconds to generate an image on SD 3 using the basic ComfyUI workflow settings, good luck.

My PC :

- GPU: NVIDIA GeForce RTX 2060 6GB

- Memory: 32GB RAM

- CPU: Intel(R) Core(TM) i7-10700F

Generation settings :

- Steps: 28

- Resolution: 1024x1024

Should I able to post my genereted images to instagram account or I any need licience and permision ?

I've heard sd3 is just as disappointing as sd2, really?

Automatic1111 isn't loading this model for me.

That's the great thing about Stable Diffusion, that you can let your thoughts run free and create anything you can imagine, without a prudish idea like (Midjorney and DALL-E) producing nothing sensible... I hope that in the next few weeks SD3 versions will be released that are again uncensored! In this state I have to say, UNUSABLE!

I cannot load this this model on Auto1111 why is this how to fix it? Waited whole night to download this massive 14gb with my slow net but now it doesn't work 😭

any tips on realism? parameters? sampler?

SAY NO to the censorship rip-off!

Even if SD3 was good in itself, it would suck! Because of the license and the censorship!

NO to the rip-off

NO to the ban on creativity

♥

I am unable to use it with tiledDiffussion
Any advice?

If anyone is wondering where these random good gens are coming from, go to Reddit and look up negative prompts that are nsfw filled. By doing so you restore some function of the poisioned model. You can see the difference with and without the negative prompt many paragraphs long.

BANNED FROM CIVITAI

Hopefully Stability will learn a very hard lesson from this.

SD3 censorship means that one might as well just go use the Bing Image Creator (which I do from time to time.) It's safe, fast, and you don't have to have a huge expensive graphics card to run it at full-tilt. There is no other reason to use SD than the platform just staying out of the way.

Wow, Stable Diffusion 3 medium is even worse than SD2. This has to be a new low.

Looks like we'll keep using SDXL checkpoints until something better comes along.

The hate in the comments is absolutely crazy. You people don't care about censorship point blank, you just want to use a model to make your porn and hentai. You get one model where you can't make your porn to goon to and you go apeshit.

Damn didn't know training data for these people called "women" were so hard to find. Must be an extinct species.

Another disappointment. For me, the main issue with SD models isn't even the censorship, but they still can't properly work with clothing, and especially with shoes.

Great technical insight into the nuances of the SD3 model👇
https://civitai.com/articles/5794

Bad model, bad license, bad moderation on StabilityAI's part. I got banned on their discord for being negative about their model and the license, they called negative feedback "trolling", serves them right to be boycotted!

Important thing to say 3 times:

Remove all "Negative prompts"!

Remove all "Negative prompts"!

Remove all "Negative prompts"!

Watch this video to generate better images with SD3 2B (it was done by Matt3o, who wrote the IPAdapter ComfyUI node, so he is not just some random Youtuber): https://www.youtube.com/watch?v=OrST6Nq1NUg (the good stuff starts at around 8:50)

TLDR: Various ways to "hack" SD3 when you see bad generation:

1. Use non-standard resolution such as 1042x1042 (not multiple of 64).

2. Avoid words such as "lying". Or use the 2 CLIP + T5 ClipTextEncoderSD3 node, but only use "lying" in clip_g.

3. Use "random noise" such as "aaaaa aaaaaaa aaaaa" in your negative prompt.

I don't know why people complain about SD3, but it works nicely for me, even better than SDXL or any other I have used. So far, the best tip I have found is to use CFG 2 or even lower, especially for realistic stuff. Also, Don't be afraid to experiment with width and height. Also, avoid negatives if you can. Steps 28 or 32

Is not perfect but so is any base model.

PS!! Now I use 2cfg only it work on everything its crazy! Just steps either 28 or 32. If you put cfg higher you get shit no kidding

If anyone trains a Base Model, using this one as a base, specialised in furry/anime/realistic/cartoon/comic images, and that model returns good images with very few attempts, it will be have a lot of merit, extra points if that hypothetical model can natively generate images with coherent texts.

Not usable because of copyright issues....

StabilityAI has updata their license. It's not good for anyone to continue ban SD3, even though it has many drawbacks.

Open SD3

Open SD3!!!!

Bro open sd3

Why is this still not available?

Why?

FINALLY, WE MADE IT GUYS

wb

The most you can do is train models and lora for SDXL, otherwise the model is completely useless, you can produce 20X more and better with SDXL. The license of this is stupid.

What is the 14.68GB "stableDiffusion3SD3_sd3MediumInclClip.safetensors"?

Is it the top-end model that includes everything in the released SD3 Medium?

There doesn't seem to be an explanation in the Quickstart Guide.

The bugs or oddities I see in this official checkpoint are surprising, as compared with the offiical SDXL checkpoint model:

- If I use 512x512, the images seem to get cropped vertically. I guess SD3 expects a taller canvas. Using a minimum of 768 height mostly solves it.

- This checkpoint can't draw Britney Spears, Madonna, or Summer Glau! WTF! I'm not talking about drawing them naked, just drawing their faces. It doesn't know them! The SDXL official checkpoint can draw them. It does know Taylor Swift, so I guess I'll use her in my experimental prompts.

- If you use anything but Euler, you get weird hellish images. (I'm using automatic1111 where that is not the default.) Why would an official checkpoint release happen without support for more options?

When will Stability.ai release the new finalized SD3 model? This model should be removed from everywhere at all - since nobody understood why such a terrible thing was put out...

This works best for termux android in clip included version without t5xxl its only 5 gb and runs flawlessly on Samsung s20+ it as close to flux ima get:) but I can't seem to get good female images,I e gotten a few but only portraits no posing or full body angles:(

https://github.com/KintCark/COMFYUI-ANDROID-TERMUX

Here is my github page with comfyui on termux android. Ull need at least 10gb of ram 12 but termux only reads 10gb if u got 12gb and if u got 8gb it reads 7 gb 🤔

Maybe I don't understand something, but I can't get SD3 to run on Automatic Forge. Is there any way to run it? I get an error - TypeError: 'NoneType' object is not iterable

Because I'm new to this, it would be very helpful to know which folder each of these models should be put. I have figured out that the ???clip_g and ???clip_l go in the ComfyUI/models/clip/. I've also figured out that sd3.5_large_fp8_scaled.safetensors and sd3.5_medium_incl_clips_t5xxlfp8scaled.safetensors belong in the ComfyUI/models/checkpoints folder (https://blog.comfy.org/p/sd3-5-comfyui). I'm not sure where the other sd3.5 files belong. An explanation or instructions on where I can find this out will be most appreciated.

I am using SD3 Medium Q4_K_S GGUF model, have added the L&G CLIPs + t5. It is still asking for a VAE. Where can I get that ? I used the VAE from flux but the O/P image is just a mess.