Your Options for Running ComfyUI in the Cloud
ComfyUI is a powerful graphical user interface for AI image generation and processing. ComfyUI supports a variety of Stable Diffusion models (such as SD1.x, SD2.x, SDXL, and Stable Video Diffusion) and offers many optimizations for performance.
It also comes with an asynchronous queue system, dynamic node composition, model merging, prompt conditioning and editing, and a command line option for low-VRM usage.
ComfyUI also offers a node-based interface, allowing users to create complex workflows without any coding required.
Despite its numerous features and benefits, running ComfyUI locally can be resource-intensive, especially for users with limited hardware resources.
Here’s how you can run ComfyUI in the cloud to alleviate these resource constraints and enhance your AI image generation experience.
Done For You Services
Several cloud-based options are available for running ComfyUI, providing an easy way to access the software without the need for a powerful GPU or local installations. These services offer different features, pricing models, and user experiences to cater to various needs and budgets.
When choosing a cloud-based ComfyUI service, consider the following factors:
- Pricing model: Some services charge based on usage (e.g., GPU time), while others offer monthly subscription plans with a set amount of credits or workflow runs.
- GPU performance: The type and power of the GPU can impact the speed of your workflow execution. Some services offer a choice of GPUs with different VRAM capacities.
- Parallelism: The ability to run multiple workflows simultaneously can significantly reduce overall execution time, especially for complex projects or batch processing.
- Customization: Depending on your requirements, you may need a service that allows you to install custom nodes, extensions, or models.
- User experience: Some services provide a user-friendly interface, pre-built workflows, or template galleries, which can be particularly helpful for beginners or those looking for a streamlined experience.
- Community and support: Having access to an active community of creators and comprehensive support can be valuable for learning, troubleshooting, and inspiration.
Comparison Table
When selecting a cloud-based ComfyUI service, carefully evaluate your specific requirements and budget.
Consider the trade-offs between pricing, performance, customization, and user experience to find the best fit for your needs.
Build Your Own
If you prefer more control over your ComfyUI setup, you can build your own instance using cloud computing platforms. \
This approach allows for greater customization and flexibility compared to the “done for you” services.
Here are a few popular options:
Runpod
Runpod offers a simple way to get started with ComfyUI in the cloud. They provide a pre-configured template that makes it easy to spin up a ComfyUI instance with just a few clicks.
Runpod recommends selecting a pod with at least 16GB of RAM for experimentation, and you can easily scale up for generating larger images.
Their blog post “How to get Stable Diffusion Set Up With ComfyUI on Runpod” provides a detailed walkthrough of the process.
Vast.ai
Vast.ai is another cloud platform that makes it simple to run ComfyUI. They offer a one-click template that gets you up and running quickly.
Just head to their Templates page, select ComfyUI, choose your desired machine (GPU) and storage, and click the “Rent” button to start your instance.
Personally, I’ve found Vast.ai to be a great option as its really cheap and intuitive. For example, to rent a RTX 3090 for an hour will only cost about $0.25! Not bad considering a brand new RTX 3090 will cost about $1,000.
Vultr
Vultr is a cloud computing platform that offers a ready-to-use ComfyUI image in their Marketplace.
This pre-configured image allows you to launch and start using ComfyUI instantly, without any setup hassles. Vultr’s cloud infrastructure provides scalability, reliability, and global accessibility for your ComfyUI projects.
Regardless which option you choose, running ComfyUI in the cloud can greatly enhance your AI image generation experience by providing access to powerful hardware resources without the need for local installations.
But be sure to consider factors such as GPU performance, memory, storage, and pricing.
Each platform has its own pricing structure and available resources, so compare the options to find the best fit for your needs and budget. Building your own instance provides the most flexibility and control, but may require more technical knowledge compared to the “done for you” services.
ChatGPT, Anthropic’s Claude, and other LLMs can be super helpful in creating the essential scripts and configurations for spinning up my ComfyUI instances in the cloud. They have saved me countless hours of research and trial and error.
ComfyUI Workflows
ComfyUI Workflows .json or PNG files that allow you to share pre-configured workflows with others or easily reproduce your own creations.
👉 We’ve got a bunch of workflows for free over on the Prompting Pixels website.
These workflows encapsulate all the nodes, settings, and connections required for a specific image generation task, making it simple to replicate or adapt the process for different inputs or outputs.
Users can drag and drop existing workflows into ComfyUI and start generating art right away, or they can use these workflows as a base and modify them to fit their own needs.
Workflows can be created to support a wide variety of models and extensions, such as Hires Fix, Img2Img, Inpainting, Lora, Hypernetworks, Embeddings/Textual Inversion, Upscale Models, and many more.
