Stable Diffusion XL (SDXL) Locally On Your PC - 8GB VRAM - Easy Tutorial With Automatic Installer

#SDXL is currently in beta and in this video I will show you how to use it install it on your PC. This tutorial should work on all devices including Windows, Unix, Mac even may work with AMD but I couldn't test it. I also have shown settings for 8GB VRAM so don't forget to watch that chapter.

Source GitHub Readme File ⤵️

Automatic Installer Script File ⤵️

Our Discord server ⤵️

If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on 🥰 ⤵️

Technology & Science: News, Tips, Tutorials, Tricks, Best Applications, Guides, Reviews ⤵️

Playlist of #StableDiffusion Tutorials, Automatic1111 and Google Colab Guides, DreamBooth, Textual Inversion / Embedding, LoRA, AI Upscaling, Pix2Pix, Img2Img ⤵️

0:00 How to use SDXL locally on your PC
1:01 How to install via Automatic installer script
1:35 Beginning manual installation
1:47 How to accept terms and conditions to access SDXL weights and model files (instantly approved)
2:08 How agreement page looks like and how to fill form for instant access
2:38 How to generate Hugging Face access token
2:53 Continuing the manual installation
3:36 Automatic installation is completed. How to start using SDXL
4:00 How to add your Hugging Face token so that Gradio will work
4:45 Continuing the manual installation
5:19 Manual installation is completed. How to start using SDXL
6:17 How to delete cached model and weight files
6:44 How the app will download weight files showing live
7:20 Advanced settings of the Gradio APP of SDXL
8:11 Speed of image generation with RTX 3090 TI
8:39 Where are the generated images are saved
9:44 8 GB VRAM settings - min VRAM settings for SDXL
10:06 How to see file extensions on Windows

SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis

Abstract
We present SDXL, a latent diffusion model for text-to-image synthesis. Compared
to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet
backbone: The increase of model parameters is mainly due to more attention blocks
and a larger cross-attention context as SDXL uses a second text encoder. We design
multiple novel conditioning schemes and train SDXL on multiple aspect ratios.
We also introduce a refinement model which is used to improve the visual fidelity
of samples generated by SDXL using a post-hoc image-to-image technique. We
demonstrate that SDXL shows drastically improved performance compared the
previous versions of Stable Diffusion and achieves results competitive with those
of black-box state-of-the-art image generators. In the spirit of promoting open
research and fostering transparency in large model training and evaluation, we
provide access to code and model weights.

The last year has brought enormous leaps in deep generative modeling across various data domains,
such as natural language [50], audio [17], and visual media [38, 37, 40, 44, 15, 3, 7]. In this report,
we focus on the latter and unveil SDXL, a drastically improved version of Stable Diffusion. Stable
Diffusion is a latent text-to-image diffusion model (DM), which serves as the foundation for an
array of recent advancements in, e.g., 3D classification [43], controllable image editing [54], image
personalization [10], synthetic data augmentation [48], graphical user interface prototyping [51], etc.
Remarkably, the scope of applications has been extraordinarily extensive, encompassing fields as
diverse as music generation [9] and reconstructing images from fMRI brain scans [49].
User studies demonstrate that SDXL consistently surpasses all previous versions of Stable Diffusion
by a significant margin (see Fig. 1). In this report, we present the design choices which lead to this
boost in performance encompassing i) a 3× larger UNet-backbone compared to previous Stable
Diffusion models (Sec. 2.1), ii) two simple yet effective additional conditioning techniques (Sec. 2.2)
which do not require any form of additional supervision, and iii) a separate diffusion-based refinement
model which applies a noising-denoising process [28] to the latents produced by SDXL to improve
the visual quality of its samples (Sec. 2.5).
A major concern in the field of visual media creation is that while black-box-models are often
recognized as state-of-the-art, the opacity of their architecture prevents faithfully assessing and
validating their performance.

thumb photo taken from twitter : stonekaiju

[SECourses]

If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on 🥰 ⤵

Technology & Science: News, Tips, Tutorials, Tricks, Best Applications, Guides, Reviews ⤵

Playlist of StableDiffusion Tutorials, Automatic1111 and Google Colab Guides, DreamBooth, Textual Inversion / Embedding, LoRA, AI Upscaling, Pix2Pix, Img2Img ⤵

[matthallett]

Got it working. Thank you. The controls are limited, but now that I can see results, I'll switch over the ComfyUI and continue.

[yashwanthnedunoori]

there is an issue with my comfi ui i only have 512 gb in vram but it identifies as whole ram space ( = 7800mb) then disables the use of v ram which ultimately uses some 300 mb left over ram to process.. how to solve this issue ?

[walidflux]

you are the best please keep providing colab notebooks in the future

[elisabethday]

I advise you wait until the full release, what is out now is a leak.

Gerçekten harikasın! Lokalde bilgisayarımın hızına bağlı olarak pek iyi sonuçlar alamadım. Bir de clipdrop'ta yüklü olan çevrimiçi üretimle, lokaldeki üretim arasında dağlar kadar fark var. Aynı şeyi senin Colab'ın ile denedim ve sonuçlar neredeyse birbirine benziyor. Herhalde clipdrop görüntü üretirken farklı ayarlarmalar, ağırlıklar felan kullanıyor. Eline sağlık tekrardan.

[minimalfun]

Thank you man! Your videos are awesome! çok teşekkürler.

[victorwijayakusuma]

I got this error
size mismatch for copying a param with shape torch.Size([320]) from checkpoint, the shape in current model is torch.Size([640]).
please help

[ASeale]

be aware that 2 of the links on hugging faces to manually download the model are youtube videos showing a man's exit hole. The middle link is a 91gb torrent download. I don't know if it's valid or not. I get the feeling this isn't a true SDXL model.

[sierraonshop_english]

thank you for all the help! if we host all this project into a strong webhosting server we will be able to use it on our website to make prompt and generate image or its not possible ?

[acexa]

how does this compare to comfy ui SDXL

[kevinehsani]

any one getting this error with controlnet recently " RuntimeError: mat1 and mat2 shapes cannot be multiplied (154x2048 and 768x320)"

[sierraonshop_english]

can you do a tuto on mac if possible pleasse ? thanks

[MegaBlindy]

Should add in your tutorial that people need to enable script execution as its disabled by default

[kevinehsani]

Thanks for the great video. I have 8 GB of GPU and seem to have around 16GB of free Ram, is that considered low VRAM?

[snakesea-it]

Can the Lora models trained based on the SD1.5 model be compatible with the SDXL model?

[EthanNollmeyer]

the links to the hugging faces do not work... do you plan on updating the Patreon?

[zozuh]

Hello! During the install torch command I get:

WARNING: There was an error checking the latest version of pip.
after some packages download then it stops

[TheMemeSong]

I'm using 4070 12 GB, do you think it can handle it?

[Sandeepan]

Hi,

Is it possible to use this with ROCm HIP on AMD ?
Or is it specific to NVIDIA/CUDA ?