Comfy Cloud Is Out of Beta — and It's Just Getting Started
The Comfy Cloud private beta aimed to deliver ComfyUI's complete capabilities globally without installation complexities.
We're excited to share that Comfy Cloud has now graduated from beta. This landmark is attributable to the numerous creators, developers, and artists in our community who tested rigorously, identified flaws, reported issues, and encouraged us to develop a product ready for launch. We are truly grateful. Next, we'll detail the offerings now available.
Cloud now hosts your preferred custom nodes
Approximately 90% of local workflow custom nodes are accessible on the cloud, requiring no configuration or installation. Transfer your frequently utilized pipelines to the cloud in their original form.
Charges apply solely to active workflow execution, not idle periods
Typically, cloud services invoice users starting at machine activation, regardless of whether they are actively processing or merely contemplating node connections. This inactive period accumulates swiftly, discouraging trial and error. Comfy Cloud operates uniquely: fees are incurred exclusively during active GPU usage when workflows are processing. Constructing your graph? No cost. Adjusting node parameters? Complimentary. Billing commences upon execution and ceases immediately upon workflow completion. Exceptional workflows emerge from repeated refinement, not initial attempts. Our pricing model is designed to align with this principle.
Current Models, Immediately Accessible
Locally executing AI workflows often involves time-intensive tasks like locating models, verifying compatibility, and overseeing storage capacity. ComfyCloud provides immediate availability to premier open-source models (such as Qwen, LTX 2, and Hunyuan) and proprietary models (including Nano Banana 2, Grok, and Kling 3.0). While these models are potent upon deployment, achieving distinctive results aligned with your artistic intent may necessitate LoRA-based fine-tuning. Comfy Cloud accommodates custom LoRA uploads, fully integrated with HuggingFace and Civitai. Additionally, we offer widely-used community-trained LoRAs, enabling boundless experimentation and output customization. Whether you've developed a unique character, a signature lighting technique, or a brand-aligned visual style, you can incorporate it into the cloud and utilize it across workflows.
High-Performance GPUs Accessible via Browser
Comfy Cloud operates on NVIDIA Blackwell RTX 6000 Pro GPUs — identical hardware utilized in advanced AI research facilities and production studios. Each unit features:
96GB of VRAM — sufficient for intensive video generation, upscaling, and multi-model workflows
180GB of system RAM — enabling seamless execution of the most resource-intensive pipelines
Experience It at No Cost
To minimize obstacles for inquisitive creators not yet prepared for commitment, we recently launched a complimentary tier — providing a genuine preview of Comfy Cloud's capabilities. Upon readiness for deeper exploration, our premium plans offer extended workflow durations, increased credits for API nodes (including Sora, Veo, Nano Banana 2, etc.), and team billing arrangements.
Future Developments
Graduating from beta marks a significant achievement, yet our journey continues. Preview upcoming developments:
Workflow API deployment — operate ComfyUI workflows as production APIs
Parallel workflow execution — process several tasks concurrently
Team plans — facilitate collaboration within studios or agencies
Expanded model library — continuously incorporating new open and closed-source models
Continuous expansion of custom node support — community-developed nodes that enhance creative infrastructure beyond any single entity's capacity
Introducing the #comfyrewind Challenge
Celebrating ComfyCloud's beta graduation, we are distributing 100,000 complimentary Cloud credits to creators capable of transporting us to past eras. The Challenge entails producing a 30-second reel with ComfyUI in a pre-1990s vintage aesthetic, authentically evoking a bygone decade (1960s, 1970s, 1980s, or 1990s).
Participation Guidelines
Generate your reel utilizing ComfyUI
Share it as an Instagram Reel
Mention @Comfycreators and include #comfyrewind
Submissions conclude on March 31
Selection Criteria
The top three creators will be selected based on reel view counts and innovative ComfyUI application. Prizes: first place receives 100,000 credits; second place, 50,000; third place, 30,000 ComfyCloud credits.