Hello everyone, are you enjoying your day?
Today, I’m excited to show you how to use SDXL Turbo on ComfyUI. If you’ve previously used Stable Diffusion and are curious about it, you can check the link below for more information!
–>Go to Post (What is SDXL Turbo and How to Use)

SDXL Turbo
SDXL Turbo is an innovative feature for real-time image generation from text, eliminating waiting or loading times. It’s a game-changer for AI artists, leveraging a special version of SDXL tuned for speed and quality. Remember my post about LCM with LoRA? SDXL Turbo can be downloaded and used similarly.
–> Go to Post (What is LCM LoRA and How to Use)
1. Preparation
- ComfyUI
- SDXL Turbo Model
1) Install ComfyUI
Today’s feature requires ComfyUI, so make sure it’s installed. For those new to ComfyUI, check the link below, then come back after installing.
2) Prepare SDXL Turbo Model
Once ComfyUI is installed, download the SDXL Turbo model and place it in the right directory. Download from the link below, then put it in the ComfyUI>Models>Checkpoints folder and run ComfyUI.
–> Download Link (SDXL Turbo Model)
2. How to use SDXL Turbo
Now, let’s set up the nodes for SDXL Turbo in the ComfyUI interface. Here’s a list of the nodes you’ll need.
Double-click on the empty screen, type in the search bar, and create each node!
- Load Checkpoint
- CLIP Text Encode (Prompt) X2
- Sampler Custom
- K SamplerSelect
- SD TurboScheduler
- Empty Latent Image
- VAE Decoder
- Preview Image
I’ll show an image for reference, so you can set it up yourself!

Now, let’s go through the settings.
1) Load Checkpoint
Select the SDXL Turbo Model you downloaded earlier.

2) CLIP Text Encode (Prompt)
We have two CLIP Text Encode (Prompt) boxes: one for Positive and another for Negative.
We’ll only use Positive for now. Write a simple prompt and move to the next step.

3) Sampler Custom
Customize the Sampler’s features. Connect the boxes you created earlier, then create nodes for the sampler, sigmas, and latent.

4) K Sampler Select
Choose your preferred sampler from the Sampler Custom node. We’ll pick the basic Euler.

5) SDTurboScheduler
Connect this to Sigmas from Sampler Custom, setting the number to 1.

6) Empty Latent Image
Set the resolution for the image you want to generate. We’ll go for 512×512 with a batch size of 1.

7) VAE Decoder
This is the final step, connecting Sampler Custom to the VAE node from Load Checkpoint.
8) Preview Image
Preview the images generated from the connected nodes.
3. The most important step before starting the generating process!
After setting everything, write a suitable prompt in the Positive Prompt box, activate the Extra Options checkbox below the Queue Prompt button, then set the Batch Count to the number 1. Finally, activate the Auto Queue checkbox and press Queue Prompt.

Since this creates new images in real-time, you’ll notice immediate changes as you slightly modify the prompts.
I’ll quickly show you how the preview images change in real-time based on the prompts you experiment with.



Isn’t it incredible? You’ve seen how the prompts you write are instantly reflected in the images, in unbelievably short amounts of time.
I’m constantly amazed at how such fantastic technologies are emerging day by day. If your computer performance is a bit low or it takes too long to generate a single image, I highly recommend using the SDXL Turbo model to get various results faster than anyone else.
If you don’t have ComfyUI and only have WebUI Stable Diffusion, you can use it in the same way by following the link below!