How To Use Stable Diffusion X-Large (SDXL) On Google Colab For Free

[FurkanGozukara]

How To Use Stable Diffusion X-Large (SDXL) On Google Colab For Free

Full tutorial: https://www.youtube.com/watch?v=s2MQqmv6yAg

#SDXL is currently in beta and in this video I will show you how to use it on Google Colab for free. Hopefully how to use on PC and RunPod tutorials are coming as well.

Source GitHub Readme File ⤵️

https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Tutorials/How-To-Use-Stable-Diffusion-SDXL-Locally-And-Also-In-Google-Colab.md

Our Discord server ⤵️

https://bit.ly/SECoursesDiscord

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 🥰 ⤵️

https://www.patreon.com/SECourses

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

https://www.youtube.com/playlist?list=PL_pbwdIyffsnkay6X91BWb9rrfLATUMr3

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

https://www.youtube.com/playlist?list=PL_pbwdIyffsmclLl0O144nQRnezKlNdx3

00:00:00 How to use SDXL On Google Colab for free

00:00:18 How to accept SDXL agreement and get weights

00:00:43 How to generate access token of Hugging Face

00:01:07 How to start Colab properly to get GPU and use SDXL

00:02:14 Advanced settings of Google Colab gradio for SDXL

00:02:48 it/s on Google Colab for SDXL

thumb photo taken from twitter : stonekaiju

Video Transcription

• 00:00:00 Greetings everyone.

• 00:00:01 In this video, I will show you how to use latest Stable Diffusion SDXL 0.9 on Google

• 00:00:07 Colab.

• 00:00:08 I have prepared a GitHub readme file.

• 00:00:11 The link of this file will be in the description of the video and also in the comment section

• 00:00:14 of the video.

• 00:00:15 Click this link to open Google Colab.

• 00:00:17 To be able to use SDXL, you need to have a Hugging Face account and accept terms and

• 00:00:24 services for Stable Diffusion SDXL version.

• 00:00:28 First of all, click this link.

• 00:00:30 You need to be logged in in your Hugging Face account.

• 00:00:34 Currently, I am logged in into my account.

• 00:00:37 Then you will see here accept terms and services.

• 00:00:41 You need to accept it.

• 00:00:42 It will automatically verify you and then you will have this screen.

• 00:00:47 Then you need to generate an access token.

• 00:00:49 Click this link.

• 00:00:51 You will get to this screen.

• 00:00:52 Here click new token.

• 00:00:54 Give any name.

• 00:00:55 Give, role read and write.

• 00:00:57 Copy.

• 00:00:58 Then in here, copy paste your access token.

• 00:01:01 Don't worry, this will be only visible to you.

• 00:01:03 No one else will see it.

• 00:01:05 Click connect in here.

• 00:01:07 This is really important.

• 00:01:09 Make sure that you have connected and you have GPU.

• 00:01:12 If you don't have a GPU, click change runtime and get a GPU.

• 00:01:16 It works with a free GPU.

• 00:01:18 These are my notebook settings.

• 00:01:20 Then click play this icon.

• 00:01:22 Click run anyway.

• 00:01:23 Okay, we did set our access token.

• 00:01:25 Click this play icon.

• 00:01:27 It will download and install all of the necessary models, weights and requirements for running

• 00:01:33 SDXL.

• 00:01:35 Once the installation has been completed, you will see the green checkmark here and

• 00:01:40 it will show you a play icon like this.

• 00:01:43 Then run this cell below.

• 00:01:44 This will start a new gradio interface that we will use.

• 00:01:49 Once the files are downloaded and requirements are installed, you will get a running on public

• 00:01:54 URL link like this.

• 00:01:56 Click it.

• 00:01:57 And here now we have our own Stable Diffusion SDXL 0.9 version.

• 00:02:03 This is all for us.

• 00:02:04 We can use it as much as we want.

• 00:02:06 This is running in this Google Colab.

• 00:02:09 As long as we run this, this will run.

• 00:02:12 There are advanced settings.

• 00:02:13 By default, it generates 4 images at the same time with 50 steps.

• 00:02:18 There is refiner steps.

• 00:02:19 I haven't tested them yet.

• 00:02:21 I just got access.

• 00:02:22 So I am just making a video for you.

• 00:02:24 And there is also guidance scale.

• 00:02:25 This is CFG as we know.

• 00:02:28 So one of our discord member told me to use this as a prompt.

• 00:02:32 I will use it.

• 00:02:33 Okay, I have entered my prompt.

• 00:02:35 I will generate an image.

• 00:02:36 And we will see.

• 00:02:38 When the first time you generate an image, it will load all of the weights into the VRAM

• 00:02:43 and memory.

• 00:02:44 Therefore, it will take some time.

• 00:02:46 You will see the generation process here.

• 00:02:49 Currently, it is taking 3.10 seconds per it, but we are generating four images at the same

• 00:02:55 time.

• 00:02:56 So to get the correct it we need to multiply this by 4.

• 00:03:01 So it becomes like 0.75 seconds per it which makes 1.3 it per second for Stable Diffusion

• 00:03:09 SDXL on a free Google Colab notebook.

• 00:03:13 Once you see the process 100%, be patient, the images will appear here.

• 00:03:19 It is taking about 150 seconds to generate four images with 50 steps.

• 00:03:25 I will also hopefully make a tutorial for RunPod and to run it on your computer as well.

• 00:03:31 So it will probably run much faster on those platforms.

• 00:03:34 Okay, these are the images we got.

• 00:03:37 Wow, looking very amazing, very beautiful.

• 00:03:40 This is also decent and cute.

• 00:03:42 Wow, this is also pretty good, pretty amazing.

• 00:03:45 These are native 1k resolution.

• 00:03:48 Finding out the prompts and finding what is possible with Stable Diffusion SDXL is up

• 00:03:53 to you.

• 00:03:54 So stay subscribed to see the new tutorials.

• 00:03:57 Hopefully see you later.