Ready to deploy
Start with a maintained runtime built around working production workflows.
Production AI Runtime for ComfyUI on RunPod
Your Workflows.
Your API.
Your Cost Control.
Run production-ready ComfyUI workflows in your own RunPod account — scale workers with demand while keeping control over infrastructure, data and operating costs.
Your Workflow / App
WORKFLOW · PROMPT · API
Comfy Rail
WORKFLOW · RUNTIME · DEPLOYMENT
Deploy Runtime
VERSIONED · VERIFIED · MAINTAINED
Generate
WORKFLOW · GPU · RESULT
Return Output
YOUR APP · API · WEB SERVICE
Production value
Maintained over time
Receive versioned releases, documented changes and a stable integration contract.
Stable in production
Run verified runtime images that can add parallel workers as request volume grows.
One maintained production stack
Three specialised runtime profiles.
Choose the package that fits the workflow complexity, startup behaviour and GPU capacity your product needs.
01 · GENERATE
Generate at Speed
Run production-ready ComfyUI generation workflows without rebuilding your infrastructure.
02 · EDIT
Edit Without Limits
Deploy advanced editing workflows with the models and custom nodes your product depends on.
03 · UPSCALE
Upscale for Production
Process demanding, high-resolution outputs with a runtime built for production workloads.
RunPod Serverless, in your account
Scale with demand.
Stop when idle.
Deploy Comfy Rail as a RunPod Serverless endpoint in your own account. Start with one request, add parallel Flex workers as demand grows, and scale back to zero when traffic stops. You keep control of the GPU profile, supported deployment location, endpoint configuration and billing.
Throughput grows with worker count, GPU choice and workflow efficiency rather than a fixed image limit imposed by Comfy Rail.
RunPod bills Flex workers per second while they initialize and run. Configured idle timeout and storage are billed separately. See how Serverless billing works or compare current GPU pricing.
Scale to Zero
Flex workers can scale down completely when idle.
Pay Per Second
Compute is billed while workers initialize and run.
RTX 5090
32 GB VRAM for fast image-generation and editing workloads.
RTX PRO 6000
Blackwell · 96 GB VRAM for larger workflow and model stacks.
Parallel Workers
Add workers automatically as queued request volume increases.
GPU Priority
Select multiple compatible GPU types to improve availability.
RunPod remains the cloud provider and data processor. GPU availability, maximum workers and throughput depend on your RunPod account, endpoint settings, selected hardware and workflow. Data protection depends on your selected location, storage, DPA and application configuration.
How it works
Your ComfyUI workflow
to production runtime.
- 01
Bring your workflow
Start with a working ComfyUI workflow or a clearly scoped production output we can map to a compatible workflow.
- 02
Launch with RunPod Serverless
We package, configure and deploy the right maintained runtime profile in your RunPod account.
- 03
Connect your app, website or SaaS
Call the documented API from your product backend and return results to the experience your users already use.
- 04
Scale with demand
Add parallel workers for request spikes, use multiple GPU priorities for availability, and scale back to zero when idle.
Built for a clear use case
For products built around visual AI.
A strong fit
Apps, websites, SaaS products and public-facing services that need repeatable image generation, editing or upscaling behind their own API.
Not the product
One-off internal image production, campaign-only creative work, unlimited workflow R&D, 24/7 operations or redistribution of runtime images.
See exactly where Comfy Rail fits
Product use cases
End-user image tools, product visualisation, automated editing, high-resolution delivery and visual features inside an app, website or SaaS.
Workflow scope
Compatible ComfyUI workflows can be reviewed, selected and adapted for a defined production result. Models and custom nodes are checked for runtime and licence fit.
Launch & maintenance
We configure the endpoints, integrate and test agreed workflows, choose GPU profiles and verify API paths. Maintained releases and compatibility updates keep the runtime profiles current within the support scope.
FAQ
What stays in your control?
Do you host the API for us?
No. You deploy in your own RunPod account and connect the documented API to your own product.
Do we need a finished ComfyUI workflow?
Not necessarily. A proven workflow is the fastest starting point, but a clearly scoped production use case can also be matched to and adapted from compatible ComfyUI workflows during onboarding. Open-ended workflow development, model training and major custom-node development are separate work.
What is included?
Three maintained deployment packages: a lean Generate profile, a larger Edit profile for advanced multi-model workflows, and the highest-capacity Upscale profile for demanding high-resolution processing. The service also includes versioned releases, agreed workflow integration and testing, endpoint and GPU-profile setup, deployment documentation, API Contract v1, reference material, onboarding and maintenance within the defined support boundary.
How far can it scale?
RunPod Serverless can add parallel workers as request demand grows. We configure worker limits, scaling behaviour and compatible GPU priorities around the workload. Actual throughput depends on workflow execution time, selected GPUs, regional availability and your RunPod account quota.
Who pays for the GPU infrastructure?
RunPod bills your account directly. Flex workers are billed per second while they initialize and run; configured idle timeout, storage, GPU tier and worker count affect the total cost. View current RunPod Serverless GPU pricing.
Your path to production
Bring your workflows.
We will show you the path.
Your email opens with a ready-to-send message. No workflow files or secrets are required for the first conversation.
Request a Demo