Running ComfyUI workflows on Amazon SageMaker AI processing jobs

2026-06-22 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Software Development & Engineering · Depth: Intermediate, long

Summary

This post details how to automate large-scale content generation by deploying ComfyUI workflows on Amazon SageMaker AI processing jobs. The solution leverages AWS Cloud Development Kit (AWS CDK) to set up the necessary infrastructure, including Amazon S3 for output, Amazon VPC for network isolation, and AWS Lambda to trigger SageMaker processing jobs. It utilizes GPU-accelerated ml.g5.xlarge instances running a custom Docker container with ComfyUI and the 6B-parameter Z-Image Turbo model for text-to-image diffusion. This setup enables generating hundreds of high-quality images in a single batch, with pay-per-second billing and automatic job termination. The process involves pulling model components from HuggingFace, populating workflow templates with prompts and seeds, and streaming generated images to S3 in real-time.

Key takeaway

For MLOps Engineers or AI Architects tasked with scaling generative AI content pipelines, this solution offers a robust framework. You should consider deploying ComfyUI workflows on Amazon SageMaker AI processing jobs to automate high-volume content creation. This approach provides GPU-accelerated inference, pay-per-second billing, and secure, isolated processing, allowing you to rapidly generate diverse assets like ad creatives or localized designs while optimizing costs and freeing creative teams.

Key insights

Automating ComfyUI workflows on Amazon SageMaker AI processing jobs enables scalable, cost-effective, and secure AI-powered content generation for enterprises.

Principles

• GPU-accelerated instances reduce inference times.
• Pay-per-second billing optimizes compute costs.
• Queue-based architecture scales naturally with workload.

Method

Deploy AWS CDK stacks (DataStack, SecurityStack, ComfyUISmStack), configure a custom ComfyUI Docker container on SageMaker ml.g5.xlarge instances, and trigger jobs via Lambda to process prompts and generate images.

In practice

• Generate hundreds of ad variants for A/B testing.
• Create locale-specific packaging designs.
• Produce dynamic video narratives for games.

Topics

• ComfyUI
• Amazon SageMaker AI
• Generative AI
• Text-to-Image Diffusion
• AWS CDK
• Z-Image Turbo
• Content Automation

Code references

• Comfy-Org/ComfyUI
• aws-samples/sample-comfy-to-sagemaker-processing-job

Best for: AI Engineer, MLOps Engineer, AI Architect

Related on AIssential

• How to Build, Run, and Scale High-Quality Creator Workflows in ComfyUI — ComfyUI on NVIDIA RTX GPUs enables local, accelerated generative AI workflows for creative professionals.
• How Ricoh built a scalable intelligent document processing solution on AWS — Combining generative AI, serverless architecture, and standardized frameworks can overcome document processing scaling limits.
• ComfyUI hits $500M valuation as creators seek more control over AI-generated media — Granular, node-based control over diffusion models addresses limitations of prompt-only generation, enhancing creative precision.
• The next chapter for ComfyUI — ComfyUI's creator launched Comfy Org to advance open-source AI inference, focusing on usability and performance.
• Generate Unlimited AI Films and Shots Locally— No Internet, No Limits — Local AI generation offers greater control and consistency for cinematic video production.
• Amazon SageMaker AI now supports optimized generative AI inference recommendations — Automated generative AI inference optimization significantly reduces deployment time and cost.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.