In this video, I will address two
questions: What happened to Stable Diffusion in Google Colab? I'll also show
you how to use Stable Diffusion for free, including options with Automatic1111.
Let's get started. So, first things first. What happened
to Stable Diffusion in Google Colab? Well, in reality, not much has changed for
stable diffusion but Google officially banned the use of the most popular interface
for Stable Diffusion, Automatic1111. So, when you run a notebook, it
disconnects after 5 minutes. However, it's important to note that working
with Stable Diffusion is still possible, but in different interfaces,
which we'll discuss shortly. Can your account be banned for attempting
to run a notebook with Automatic1111? Yes, it's possible, but unlikely, and at
the very least, it will give you a warning. Does Automatic work in the paid version?
Yes, it works without any issues. Are there other ways to run Stable Diffusion? With Automatic, the options are limited because
all existing solutions were banned. You can try to find a method through Kaggle, which is a free
alternative to Google Colab, but even it has banned Automatic just before I planned to release
my notebook for you. I managed to find a solution, and now it's once again possible to use
Automatic1111 for free on Kaggle. However, I believe this solution doesn't scale well, and it
may be banned if it becomes widespread. Therefore, I've made it available to my Patreon subscribers,
where you can access the notebook and use the latest Automatic1111 without limitations.
For those who find this solution unsuitable, I'd like to remind you that alternative user
interfaces still work very well. You can use ComfyUI; I'll leave a link in the description.
If you're not a fan of complex setups and nodes, there's even a better solution called
SwarmUI, which I've also discussed before. It works flawlessly for basic user
cases, and you can check the video with a tutorial and the link is also available
in the description under this video. There's also an alternative in the form of Hugin
Face Spaces, which isn't perfect but can still be suitable for occasional solutions. And, of
course, there's ClipDrop from StabilityAI. So, let me show you how to use my Kaggle
notebook with kaggle.com or Automatic1111.
That's quite easy. First thing first, you
need to create your account. Press register and you
can authorize using Google. OK, next you should press
here and press "Settings". We need to pass phone verification in
order to get access to free GPU. And then you can get access to GPU, which
will be shown here in quotas. You have 30 hours a week for free.
That is to my mind quite a lot. Then we should press this button here,
"Create new notebook". Here we go.
As you can see, it looks almost the same like a Google Colab.
And what you need to do next, press "File", "Import notebook"
and upload my notebook here, which have been linked to my Patreon
page, "Browse files" and choose "SD1". Press "Import" and here we go.
In order to use this notebook, you don't need to know anything about coding, which
is very important because a lot of things might look a bit
frightening, but it's easy to use and I say to get you don't need to
know about coding anything. So as it works before in a Google Colab.
So now we just downloaded our notebook and as you can see, it
looks almost the same as it was in a Google Colab.
We have here step 1, step 2 and step 3. In step 1, we are
installing our requirements. In order to do that, just choose this
cell and press "Run" here. I already done this.
In order to hide this code, just click here.
In step 2, we should download our checkpoint, Lora
models and controlnet models. How to do that?
Press "Expand". So here you should just
paste your desired model. To do that, go to your model webpage and
choose "Copy link address". A link for downloading the
model, not a link for model. Just for downloading the
model, I say it once again. Copy link address and paste it here,
in the inverted commas. That's it.
And then just press this button or this button in order to
start downloading process. There is a code to which
looks a bit intimidating. In order to make it look
clearer, just press here. And also you can hide this bar.
In the next cell, you can download your contract models also by
passing your controlnet model here.
This works almost the same. Just in this case, we use "list", so all
models should be with commas and brackets.
When all models will be downloaded, just run stable diffusion.
Here we go. We are in our lovely
automatic 1111 interface. It's the latest version of automatic
which supports the SDXL and the refiner models.
Let's generate our Cat, traditionally. Here we go.
So unfortunately, we lose ability to use Google colab for automatic.
But at the same time, we are able to do that in a Kaggle, which
works basically the same as Google Colab.
So it seems not so bad. I will update that
notebook according to your design. Just let me know in a Patreon page.
And also if you want me to add in extensions, just let me
know and I'll do that. So thanks for watching.
See you next time.
