ComfyUI / AI Pipelines

AI image generation built for production

ComfyUI workflow development and Stable Diffusion pipeline integration for creative studios and agencies. Not experiments - production-ready systems with consistent outputs, sensible interfaces, and the ability to handle real project volume.

Start a Project View Portfolio

Why work with me

Production-focused, not experimental

AI image generation in a studio context has different requirements than personal use. Consistent brand-aligned outputs, batch processing, file management, and team-accessible interfaces - built for production, not a hobby.

Custom nodes and workflows

Custom ComfyUI nodes for studio-specific needs - brand consistency enforcement, output metadata, file naming conventions, and integration with existing asset management systems.

Full AI stack capability

ComfyUI is one part of the offering. Claude API agents, LLM-powered tools, and hybrid AI/Python workflows are all available depending on what the production actually needs.

Real projects. Real clients.

A selection of relevant work from 10+ years of practice.

Creative Studio

ComfyUI Brand-Consistent Pipeline

Built a ComfyUI pipeline for a creative studio producing high-volume AI-assisted imagery. Custom nodes enforced brand color palettes and style consistency, with a simple web interface accessible to non-technical team members. Output volume: 200+ assets per week.

Architecture Studio

AI-Assisted Concept Visualization

ComfyUI workflow for early-stage architectural concept visualization - generating mood and material options from rough sketches using img2img and ControlNet. Integrated with the studio's existing project folder structure.

Frequently Asked Questions

Can ComfyUI produce consistent enough outputs for brand work?

With the right workflow design, yes. Consistency comes from model selection, LoRA fine-tuning for brand-specific aesthetics, and controlled prompting via custom nodes. A well-built ComfyUI pipeline can produce brand-aligned outputs at scale.

Do you fine-tune models or just use existing ones?

Both, depending on the consistency requirement. Standard models with LoRAs cover most cases. Full fine-tuning (DreamBooth, Kohya) is available for applications requiring very high stylistic consistency.

How does AI image generation fit alongside traditional 3D rendering?

They complement each other. AI is fastest for concept and mood exploration - generating dozens of options in minutes. 3D rendering is better for photorealistic hero images with precise control. I use both in the same project where appropriate.