PODCAST

ComfyUI's CEO on Control, Adoption, and Why the Tool Won't Be Acquired

Mar 20, 2026 · 42 min watch

67 passages
LISTEN ONSpotifyApple PodcastsYouTube

The latest Denoised covers ComfyUI's evolution from a two-week open-source project to a 4-million-user platform that's becoming a job requirement across the industry. We got to sit down with Yoland Yan, CEO of ComfyUI, to discuss the tool's architecture, the intelligence agencies using it, and why the company is turning down acquisition offers.

Quick Take

ComfyUI started as a one-person project built in two weeks by someone who'd never written ML code. Now it's approaching 10% of the world's professional creative workforce. The story isn't about a tool that does everything — it's about a tool that gives creatives absolute control, and why that matters more than speed or simplicity.

What We Learned: The Origin of ComfyUI's Control-First Philosophy

Yoland Yan, the CEO, clarifies the naming confusion upfront: he's not comfyanonymous. That's Yanick, his co-founder, who built the original tool. Yanick's background is unusual for an AI company founder. He spent 10 years as a C++ programmer at his father's dental implant factory in Quebec City, automating manufacturing systems. Zero machine learning experience. When he started ComfyUI in early 2023, he wrote it in two weeks.

The motivation was specific: Yanick wanted to chain multiple diffusion models together in a single workflow. Some models excel at aesthetics but can't generate high resolution. Others are strong at upscaling. He wanted to use one model for the first 50% of generation steps, then hand off to another model for the final output. No existing product gave him that control without rewriting code from scratch.

The node-based interface solved a real problem. Once you write custom code, you need visual feedback. You need to see what happens when you change a parameter. A node-based workflow graph became the natural solution — it gives you control, immediate feedback, and a visual representation of what's happening at each step. That architecture is still why ComfyUI is adopted across VFX, animation, and studio pipelines.

What We Discussed: Scale and the Path to Industry Standard

ComfyUI now sees approximately 50,000 downloads per day for local installations, with roughly 4 million total users. Given that the total professional creative workforce is estimated at 50 million globally, ComfyUI is already approaching 10% penetration in that market.

The company's long-term vision is to build ComfyUI into what they call "the operating system for visual AI." That doesn't mean everyone uses the same interface. It means ComfyUI becomes the workflow orchestration layer that powers generative content creation — whether running locally, on Comfy Cloud, or on mobile SDKs. The goal is that more than 50% of generative content flows through a ComfyUI workflow layer within 5 to 10 years.

Netflix listed ComfyUI as a job requirement roughly 1.5 to 2 years ago, and that pattern has accelerated across the industry. Yan sees this as validation of the tool's utility, but also as a signal that the onboarding process needs improvement. That's where App Mode comes in.

What We Explored: App Mode and Simplified Workflows

App Mode is a new feature that lets creators expose only the parameters they want end users to interact with. Imagine a complex workflow with dozens of nodes and connections. In App Mode, you select which parameters matter (resolution, style, intensity, etc.) and hide the rest. The interface becomes simple — just a few controls and an output.

The practical application: a VFX artist builds a complex style-transfer workflow in ComfyUI, then shares it with a client via a cloud link. The client uploads an image, adjusts a few sliders, and gets the output. They never see the "spaghetti" of nodes underneath.

App Mode also handles dependency management. If a workflow requires specific models, ComfyUI now embeds that information and prompts users to download missing models. This is an area the team plans to keep improving, but the foundation is in place.

What We Questioned: Acquisition and Mission

When asked about acquisition prospects, Yan's answer was direct: "Zero." The company has fielded interest from investors and acquirers, but the founding team is mission-driven. They've turned down investors who wanted to shift toward a "Comfy Pro" model with feature-gating.

The business model is usage-based. Comfy Cloud charges based on compute time, not features. If you can render a job in 20 minutes on Comfy Cloud instead of waiting two days locally, that's the value proposition. No artificial feature walls needed.

Yan's broader point: there are two types of companies in Silicon Valley history. One makes marginal improvements on what the world was already moving toward. The other changes the trajectory entirely. ComfyUI's founders believe they're building the second kind — a foundation layer that keeps creative tools in the hands of creatives, not locked behind closed ecosystems.

Bottom Line: Control as a Moat

ComfyUI's story isn't about being the fastest or easiest tool. It's about giving creatives control. That philosophy — control, quality, scalability — has been consistent since Yanick built the first version in two weeks. It's why studios adopted it. It's why Netflix listed it as a job requirement. And it's why the company is turning down acquisition offers.

The tension in the industry right now is between tools that promise to automate everything and tools that give you the power to do exactly what you want. ComfyUI is betting on the latter. Whether that bet pays off depends on whether creatives continue to value control over convenience.

Links from This Episode

Tools & Platforms:

  • ComfyUI — Open-source node-based workflow engine

  • Comfy Cloud — Cloud backend for ComfyUI workflows

VP Land Coverage:

denoised podcast

KEEP READING