AI UGC ComfyUI Workflow

What it is

AI UGC ComfyUI Workflow is a node-based image and asset pipeline built in ComfyUI for producing user-generated-style content at volume. It templates scene composition, enforces style consistency, and queues batch runs for social, ads, and product marketing.

Why I built it

UGC-style creative wins on authenticity, but manual creation does not scale for agencies running dozens of client tests per week. I needed a system where briefs map to workflow inputs, outputs land in predictable folders, and style stays on-brand across hundreds of generations.

Workflow architecture

• Input layer — brief fields: subject, setting, mood, aspect ratio, brand colors
• Style lock — LoRA, reference conditioning, or style transfer nodes pinned per client
• Generation core — SDXL / Flux branch depending on fidelity needs
• Post-processing — upscale, face fix, background cleanup
• Output routing — named exports by client, campaign, and variant ID

Key decisions

• ComfyUI over one-click tools because branching logic matters (different clients, different style locks)
• Parameterized prompts stored outside the graph so producers can run batches without editing nodes
• Failed generations log with seed and settings for reproducible fixes

What it demonstrates

• ComfyUI as production infrastructure, not hobby tooling
• Batch creative output with auditability
• Bridge between creative direction and generative execution

Takeaways

UGC at scale is an ops problem disguised as a creative problem. The workflow wins when a producer can rerun variant 47 without asking a developer.