Greetings everyone. In this video, I will show you how to use
latest Stable Diffusion SDXL 0.9 on Google Colab. I have prepared a GitHub readme file. The link of this file will be in the description
of the video and also in the comment section of the video. Click this link to open Google Colab. To be able to use SDXL, you need to have a
Hugging Face account and accept terms and services for Stable Diffusion SDXL version. First of all, click this link. You need to be logged in in your Hugging Face
account. Currently, I am logged in into my account. Then you will see here accept terms and services. You need to accept it. It will automatically verify you and then
you will have this screen. Then you need to generate an access token. Click this link. You will get to this screen. Here click new token. Give any name. Give, role read and write. Copy. Then in here, copy paste your access token. Don't worry, this will be only visible to
you. No one else will see it. Click connect in here. This is really important. Make sure that you have connected and you
have GPU. If you don't have a GPU, click change runtime
and get a GPU. It works with a free GPU. These are my notebook settings. Then click play this icon. Click run anyway. Okay, we did set our access token. Click this play icon. It will download and install all of the necessary
models, weights and requirements for running SDXL. Once the installation has been completed,
you will see the green checkmark here and it will show you a play icon like this. Then run this cell below. This will start a new gradio interface that
we will use. Once the files are downloaded and requirements
are installed, you will get a running on public URL link like this. Click it. And here now we have our own Stable Diffusion
SDXL 0.9 version. This is all for us. We can use it as much as we want. This is running in this Google Colab. As long as we run this, this will run. There are advanced settings. By default, it generates 4 images at the same
time with 50 steps. There is refiner steps. I haven't tested them yet. I just got access. So I am just making a video for you. And there is also guidance scale. This is CFG as we know. So one of our discord member told me to use
this as a prompt. I will use it. Okay, I have entered my prompt. I will generate an image. And we will see. When the first time you generate an image,
it will load all of the weights into the VRAM and memory. Therefore, it will take some time. You will see the generation process here. Currently, it is taking 3.10 seconds per it,
but we are generating four images at the same time. So to get the correct it we need to multiply
this by 4. So it becomes like 0.75 seconds per it which
makes 1.3 it per second for Stable Diffusion SDXL on a free Google Colab notebook. Once you see the process 100%, be patient,
the images will appear here. It is taking about 150 seconds to generate
four images with 50 steps. I will also hopefully make a tutorial for
RunPod and to run it on your computer as well. So it will probably run much faster on those
platforms. Okay, these are the images we got. Wow, looking very amazing, very beautiful. This is also decent and cute. Wow, this is also pretty good, pretty amazing. These are native 1k resolution. Finding out the prompts and finding what is
possible with Stable Diffusion SDXL is up to you. So stay subscribed to see the new tutorials. Hopefully see you later.
