ComfyUI hits $500M valuation as creators seek more control over AI-generated media

ComfyUI just closed a $30 million funding round at a $500 million valuation. That’s a big number for a company that, until recently, was mostly known in niche circles for its node-based interface that lets you tweak every little knob in image, video, and audio generation.

The round was led by Accel, with participation from existing investors like a16z and Y Combinator. The money is earmarked for expanding the team, building out enterprise features, and making the whole thing less intimidating for newcomers.

If you’ve ever used Stable Diffusion through the command line or dabbled with Automatic1111’s WebUI, you’ve probably heard of ComfyUI. It’s the tool that took the opposite approach from Midjourney or DALL-E: instead of hiding the complexity behind a polished chat interface, ComfyUI throws a visual graph of nodes at you. You connect a checkpoint loader to a CLIP text encoder, pipe that into a KSampler, wire up a VAE decoder, and out comes an image. It’s powerful, but it’s not exactly beginner-friendly.

That complexity is the whole point. The people who gravitate toward ComfyUI are the ones who want to control the seed, the scheduler, the CFG scale, the denoising strength, and everything in between. They’re not satisfied with typing “a cat in a spaceship” and hoping for the best. They want to load a specific LoRA, set a custom attention mask, and chain multiple models together. This is the audience that’s been underserved by the big players, and ComfyUI has been quietly eating their lunch.

The $500 million valuation feels high for a tool that’s still largely a desktop application, but the company has been moving fast. They launched a cloud-hosted version last year, added real-time collaboration features, and released a mobile app that lets you run workflows on-device. The enterprise tier, which includes team management and private model hosting, has been picking up steam with studios and agencies that need to maintain consistent outputs across projects.

What I find interesting is the timing. The AI image generation market is getting crowded, and the big players are all racing to add more controls. Midjourney has its style tuner and image remix features. Adobe Firefly is pushing its own node-based interface called Firefly Workflows. Even OpenAI’s DALL-E 3 now supports editing with inpainting and outpainting. But ComfyUI has a head start with its open ecosystem and the massive library of community-built workflows.

The downside? ComfyUI’s learning curve is real. I’ve seen friends bounce off it because the node graph looks like a programming language designed by a mad scientist. The company knows this, and they’re investing in templates and guided modes that abstract away the complexity for specific use cases. But there’s a tension between making the tool accessible and preserving the granular control that power users love.

Another thing worth noting: ComfyUI is built on top of open-source models, which means it’s vulnerable to the same licensing headaches that hit Stability AI and others. The company hasn’t disclosed how much they pay for compute or how they handle model distribution at scale. The valuation suggests investors are betting that the creator economy will pay a premium for control, but the unit economics could get nasty if model providers change their terms.

Still, the round is a strong signal. The market for AI media tools is bifurcating: you have the mass-market apps that prioritize ease of use, and the professional tools that prioritize control. ComfyUI is staking its claim on the latter, and $500 million says there’s a real business there.