Figma Weave Is ComfyUI for People Who Actually Ship

Node-based AI generation has been a thing for almost three years now — but until recently, it lived exclusively in the ComfyUI-and-terminal crowd. Figma's acquisition of Weavy and the relaunch as Figma Weave puts that same paradigm inside the tool where most product design actually happens. Whether that's a revolution or just good product strategy depends on what kind of team you're on.

#What Weave actually is

Strip away the marketing, and the product is a browser-based node editor for generative AI workflows. You connect model nodes (Flux for images, Kling for video, any LLM for text) with editing nodes — masking, color grading, distortion — on a visual canvas. Each connection is explicit: output of node A feeds into input of node B. If you've used ComfyUI, Houdini, or even Unreal's Blueprint system, the mental model is identical.

The difference is context. The platform runs in the browser, integrates with Figma's collaboration layer, and ships with commercial model access (Sora, Seedance, Ideogram) that you'd normally need separate API keys and billing accounts to use. Same power, repositioned for creative teams rather than solo AI tinkerers running a 4090 under their desk.

#Five workflow patterns worth stealing

Figma's launch blog walked through five workflows using "Epoch," a fictional brand. Three of them generalize well beyond the demo.

Style extraction → application. Feed two reference images into a blend node, tweak influence weights, then pipe the resulting style description through an LLM node to generate a reusable style prompt. Apply it to any new subject. This is "brand guidelines as a workflow" — more precise than handing a designer a PDF with hex codes and vibes. You get a machine-readable definition of your brand's visual language that anyone on the team can execute against.

Multi-aspect ratio recomposition. Build the workflow once, generate one hero asset, let the system handle resizing across aspect ratios. The old way: export at 1:1, 16:9, 9:16, 4:5 manually, or write a script that center-crops and hopes for the best. The node-based approach is generative — each ratio gets model-aware recomposition, not a dumb crop. Backgrounds extend naturally, subjects reposition intelligently.

Motion from stills. Take a static asset, add a motion reference (a looping animation, a 3D node driving camera movement), and generate video through a Kling Element node. Physics-aware motion without opening After Effects. For product teams cranking out social assets, this alone might justify the tool.

#Why design system teams should care

Here's where it gets interesting for people who maintain component libraries and brand assets at scale.

Design systems have always been good at codifying structure — spacing tokens, color primitives, typography scales. They've been terrible at codifying aesthetics. Your system can enforce that a card component uses --spacing-lg padding and --color-surface-primary background. It cannot enforce that the illustration inside that card matches the brand's visual identity.

Weave workflows are, functionally, aesthetic tokens. A workflow that extracts your brand's style, applies it to arbitrary subjects, and outputs at multiple resolutions is a repeatable, shareable, version-controlled recipe for "what our visuals look like." That's the part design systems have always hand-waved with a Notion page titled "Illustration Guidelines" that nobody reads.

The practical shift: instead of commissioning illustrations per feature launch and praying for consistency, a design system team publishes a workflow. Product designers run it with their specific subjects. Output is consistent by construction, not by discipline. The brand team reviews workflows, not individual assets — a fundamentally different review surface that actually scales.

#The ComfyUI comparison is inevitable

ComfyUI is free, open-source, runs locally, and gives you absolute control. If you want to fine-tune a LoRA, swap in a custom checkpoint, or run everything air-gapped, ComfyUI wins and it's not close.

But "free and open source" has a hidden cost: onboarding. Getting a non-technical designer productive in ComfyUI takes days. The node graph is dense, error messages are cryptic, and half the community workflows assume you have local GPU access and a tolerance for YAML.

Figma Weave makes the opposite trade:

For a three-person startup where one founder doubles as designer, ComfyUI makes sense. For a 50-person product org with a brand team that needs consistent output across six designers, the coordination problem is the actual bottleneck — and that's exactly what the Figma integration solves.

#What's still missing

The platform isn't wired into Figma's main editor yet. It lives at weave.figma.com as a separate app — no direct component library connection, no auto-sync with Variables, no way to plug a generated asset into a Make prototype without manual export. Figma says full integration is coming later this year.

#The bet Figma is making

The trajectory is clear enough: Figma wants "design system" to mean structure and aesthetics, both codified and reproducible. Tokens handle the structural half. Weave workflows handle the visual half. Whether teams actually adopt workflow-as-asset-pipeline depends on how fast the integration lands and whether the enterprise tier solves the version control gap — right now there's no branching, diffing, or changelog for workflows, which feels incomplete for teams that version everything else through Git.

But if you produce more than a handful of brand illustrations or marketing assets per month, it's worth trying today. Start with the style extraction pattern. It forces you to articulate your brand's visual language in a form that's actually executable — and that exercise alone is valuable even if you never run the workflow again.