How to run comfyui with a rented GPU

Rent a powerful GPU, launch comfyui and really enjoy the nonsense guide to your journey.

When you first dip your toes into the world of AI tools, everything feels like a smooth journey. You enter a prompt in chatgpt and you’re OK – instant results. You show it around – applause. You click on Midjourney or the runway and BAM – beautiful images, things are moving and don’t need to be assembled.

But in the end, you reach a point where you want to go a little deeper.

You want not only to use AI, but you also want to build with it. Control it. Learn how magic happens (at least to a certain extent). For me, this idea got me into the rabbit hole of comfyui.

What is comfyui?

COMFYUI is a free open source tool for generating images and videos using AI – built around module-based, node-based interfaces. Not only do you type prompts and want the AI ​​to do it correctly, you can also gradually design your own image generation workflow.

Comfyui is an image generation tool that is free to use. Its node-based UI is not as difficult as you think.

Each block (or “node”) in comfyui represents part of the process – loading the model, interpreting the hints, adjusting the parameters, generating images – you can connect and customize them as you like.

What makes Comfyui stand out:

1. It is free and open source, which means anyone can use, extend or modify it without a subscription.
2. It gives you more control over the generation process.
3. It is powered by a steady diffusion, but it won’t lock you into a way of working.

If you ever wanted to peek under the hood generated by AI Image and actually build your own system – Comfyui lets you do that without paying for some closed black box platforms.

At first, the node-based interface may seem a bit daunting, but you will soon realize that Comfyui is a modular visual sandbox that gives you actual control over what you create – and that’s where it’s exciting.

How to get started with comfyui

Technically, there is nothing that can stop you from downloading Comfyui to your computer and running from there – but be ready to wait. a lot of. AI models like comfyui use the capacity of many graphics cards, and things will be slow if you don’t run ultra-high-end game settings.

Especially when you’re still figuring out what all the settings do, waiting a few minutes to get an image full of artifacts is…unsatisfactory. That’s why we’re going to rent out some GPUs to do the heavy lifting for us. Don’t worry, it’s very simple, no promise.

When I first started, this was what appeared after waiting for 5 minutes. So it’s better to rent a good GPU.

There are usually two ways to solve it:

One is the-template-way that just hits – simple and fast, but Jank is slightly faster.

Another is a manual CLI-based setup – with all the features, and all the responsibilities that come with it.

Ready? let’s go.

Set comfyui with template

This is very simple. After that, we will only rent the GPU and get the link to open the pre-installed version of comfyui. You can start using it right away – no settings, no pressure. It does sound like there are more commitments than before. The provider we chose is Cloudrift.

Cloudrift is a GPU provider that allows you to rent high-performance models like Nvidia’s RTX 4090 and RTX 5090 for a decent hourly price. No long-term commitment – ​​just pay.

When you select a GPU on Cloudrift, the selection may be different from you, depending on what is available, but the RTX 4090 is always a good choice.

Here’s how to start:

1. Go to cloudrift.ai and log in.
2. Add points ($10 is enough to get started).
3. Click New to start setting up your instance.
4. Select container mode.
5. Choose your GPU (start with the cheapest if you aren’t sure – even the low-end workflow will handle basic workflows well).
6. Under the Select Software tab, select Recommended.
7. Find a template called comfyui-flux.1-schnell under image generation. Select and confirm.
8. Click Deploy and wait for one minute while your instance rotates.
9. In the dashboard, you will see a summary of your instance – click Connect and open the link in the pop-up window. There you can also find useful resources on how to use Comfyui.

The new browser window that pops up should look like this.

Setting up COMFYUI using CLI (also known as a more complex way)

As mentioned earlier, using templates allows you to get started right away. Comfyui’s “Manager Menu” makes it easier for you to install updates, other tools and plugins.

But I found that by setting everything manually I can get a better idea of ​​how they work – and my settings are probably more intentional and precise.

So, here is the premium version:

Let’s start with the basics: CLI = Command Line Interface. This is a tool that allows you to interact with a service, such as a rented GPU, using text commands instead of clicking a button. If you are on a Mac, just open the terminal. On Windows, it is CMD or PowerShell.

In our case, we had to install an additional tool before starting: Cloudrift CLI – a lightweight command line tool that connects to the cloud GPU you rented and allows you to deploy applications (such as Comfyui) in a few lines.Here is how to install Cloudrift CLI.

Once this is done, go to Cloudrift.ai Rent a GPU Let’s get our hands dirty.

Step 1: Rent GPU via Cloudrift

Cloudrift is a GPU provider that allows you to rent out models of the RTX 4090 and RTX 5090 at a pretty good price.

1. Go to cloudrift.ai and log in.
2. Add points
3. Select New to start setting up your instance.
4. Select container mode.
5. Select your GPU
6. In the Select Software tab, this time select No container. This gave us a clean slate.
7. Click Deploy and wait for a moment.

Renting a GPU should be simple. Under the Software tab, select No container.

Step 2: Start Comfyui with a command

Now your GPU is running somewhere in the cloud, let’s access it.

First, open the terminal and make sure the Cloudrift CLI is installed – otherwise, the Rift command we are going to use will not work.

First, check if your GPU can be typed by:

Rift Valley Cluster Information

And click Enter.

This is what your terminal will show you. Here I rented two RTX 4090s. The machines they run have their own name and IP address as shown in the first and second columns.

Here is the command you will run in the terminal next:

Rift Docker -X Run -p 8188 -e web_enable_auth = false –name comfyui -it –rm ghcr.io/ai-dock/comfyui:latest -cuda

Let’s break it down:

1. Rift Docker: You are running Docker on a cloud GPU.
2. -x : As long as you have only one instance running – optional – replace it with the name of the Rift Cluster Info (such as MY-GPU-01).
3. -p 8188: Open port 8188, so you can access comfyui through your browser.
4. -e web_enable_auth = false: Disable login screen for simplicity.
5. –NAME COMFYUI: Optional – Make your container easier to reference later.
6. – It: Run containers in interactive mode.
7. –RM: Remove the container when stopped, so it does not take up space.
8. ghcr.io/ai-dock/comfyui:latest-cuda: This is the docker image of comfyui (basically the application we are installing).

You will see some log flights – messages about pulling images, connecting tunnels, etc. This is normal. Give it a minute or two. Congratulations, you just set up the first container and installed Comfyui in it!

Quick Note: Keep it running

If you close the terminal window now, your container stops running. Just keep it and open a new terminal window to continue.

Step 3: Confirm that it has been launched

In the new terminal window, use the following command:

Rift Docker PS

This lists all active containers. If Comfyui is running, you will see something like this:

Then, open the browser and go to:

Replace {node-ip-address} with the IP of the Rift Cluster Info (or the IP listed in your NeuralRack rental details).

This is what you should see now.

Congratulations! You have officially launched Comfyui on your rented GPU. Now let’s do something.

Step 4: Download the “Checkpoint”

Before generating the image, you need a checkpoint – a trained model that tells Comfyui how to convert the prompt to an image. Think of it as the “brain” of surgery. If comfyui is the body, the checkpoint is actually knowing what to draw.

We will use DreamShaper, a beginner-friendly model that is easy to use out of the box. You can find other checkpoints on Civitai.

Run the following to connect to your container:

Rift Docker Exec -it Comfyui Bash

Now you are in the container – just like opening a terminal inside a rented machine.

Next, navigate to the correct folder:

CD/OPT/COMFYUI/Model/Checkpoint/

Then run this to download Dreamshaper V8:

WGET – Includes allocation

(Obviously you can use this for any other checkpoint download link for exchange)

This can take a minute or two, depending on the size of the model and the speed of the internet. You will see something similar:

Dreamshaper_8.safetensors 100%[====================>] 1.99g

After you complete this, exit the container:

exit

Step 5: Start Generation!

Open the browser (if not already), and go to:

You should see the interface of comfyui. If it does not exist yet, load the default text to the image workflow.

I went and updated comfyui through the manager menu, see here:

Once done, make sure to select your Dreamshaper V8 checkpoint in the Load Checkpoint node.

Then there is magic. Click the top box that displays “Tips” and enter your prompt. The box below is a negative prompt, which means you are put there, something you don’t want to put in the image.

Click Queue Prompt, wait a few seconds, and then: the first custom image of the GPU rented using Comfyui+.

I’m a fantasy and video game nerd, and naturally, when I get the chance to create crazy AI-generated artwork, my mind turns to something you see below. But of course, you can put anything you like together (no judgment). You can also try loading different checkpoints from sources like Civitai. In this case, when you use this command, like you did before:

wget – Content breakdown

You just need to replace the link with any model of download link you find.

When done, don’t forget: Close it

One more thing: Once done, go to your neural dashboard and stop the instance here to avoid charges.

tl; dr:

1. COMFYUI is an AI image generation tool based on free nodes that you can run in the cloud using Cloudrift CLI.
2. Rent GPU through Cloudrift Provide you with power without heat (or hardware costs) and significantly improve the speed of creating content.
3. You can start from scratch with a single command, download the model from Civitai, and then generate the first image in minutes.
4. The process is surprisingly touched – even if the terminal looks horrible at first.

That’s all today. Stay curious, stay messy, and don’t be afraid to enter weird stuff into the black box.

That’s where good things happen.