Figma AI Sites vs Luma AI Dream Machine 2.0

“Figma AI Sites solves the single most painful moment in the design-to-web pipeline: the moment a developer opens the Figma file and re-interprets every spacing decision. The output is AI-generated HTML and CSS, which means the tool is only as good as how faithfully it translates auto-layout constraints, component variants, and responsive breakpoints into actual DOM structure — and from what Figma has shown, the fidelity is genuinely high. The specific design decision that earns this ship is that it respects Figma's own component and constraint system as the source of truth rather than screen-shotting the design and guessing. The risk is edge cases: complex interactions, hover states, and scroll behaviors that Figma's prototyping layer doesn't fully express will still fall through the cracks.”

“The primitive here is: static-site generator with Figma as the schema, AI as the compiler. That's actually not a bad idea, but the DX bet is entirely wrong for developers — it routes complexity into the AI black box rather than into a deterministic build step you can debug, version, or extend. The first 10 minutes for a developer means opening the exported HTML and asking whether you can wire up a real CMS, add a custom component, or run it through a build pipeline — and the answer is almost certainly 'not without re-engineering it.' The weekend alternative is real: Webflow, Framer, and even Builder.io solve this problem with more developer surface area and established component ecosystems. The skip is because 'export to custom hosting' without a documented output schema or CLI is just a ZIP file with AI feelings about your design.”

“The primitive is straightforward: a video generation model with stateful character identity seeded from a reference image and a text-driven camera/lighting control layer exposed over the existing API. The DX bet is correct — they didn't invent a new schema, they extended the existing Luma API so developers already in the ecosystem can adopt character consistency with minimal migration cost. The moment of truth for a developer is whether the character reference endpoint returns consistent results across multiple calls with the same seed, and early API docs suggest it does. This isn't a weekend Lambda script — maintaining character identity across generated frames requires model-level architecture decisions you can't bolt on — so the moat is technical, not just a wrapper around someone else's inference.”

“Category is design-to-code publishing, and the direct competitors are Framer, Webflow, and Builder.io — all of which have been doing this longer, have richer interaction models, and have established CMS pipelines. The specific scenario where Figma AI Sites breaks is the moment a client wants to update a blog post, add a product to an e-commerce section, or change a nav link without asking a designer — the CMS integration claim needs scrutiny because 'CMS integration' on launch usually means 'we support one CMS with five field types.' What kills this in 12 months: Figma's core business is design collaboration, not hosting infrastructure, and the moment this feature requires serious investment in CDN reliability, edge performance, and CMS depth, it either gets spun out or quietly deprioritized. To earn a ship, Figma needs to publish the full capability surface of the CMS layer and show production sites running on it at scale.”

“Character consistency in AI video generation is the real problem — Runway, Kling, and Pika have all fumbled it in different ways — so shipping a model that actually holds a face across cuts is a meaningful technical win, not a feature-flag press release. Where it breaks: complex multi-character scenes with similar appearances, anything requiring precise lip sync, and longer-form sequences where drift accumulates across ten-plus shots. The kill scenario isn't a competitor — it's OpenAI's Sora team or Google's Veo deciding to solve this properly with their compute budgets, at which point Luma's lead evaporates in a single model release.”

“The buyer is clear: design-led teams and agencies who already pay for Figma and want to eliminate the handoff cost to a developer or a separate publishing tool like Webflow. The pricing architecture is clever — burying this inside the Professional plan turns it into a retention feature rather than a new SKU, which means Figma doesn't have to win on publishing, they just have to make switching to Webflow slightly less obvious. The moat is the existing Figma file corpus: every team that publishes a site through AI Sites is now locked into Figma as their design AND publishing layer, which is real workflow lock-in. The stress test is whether Webflow or Framer responds by building a Figma import that's good enough — and the answer is they already have import features, so Figma's window to make this sticky is measured in months, not years. The specific business decision that earns the ship is using hosting as an expansion revenue vector inside an existing enterprise contract rather than trying to build a standalone publishing business.”

“Character consistency is the feature that makes AI video actually usable for storytelling — before this, every cut produced a different version of your protagonist's face, which meant the output was demo reel material, not real content. Dream Machine 2.0's scene control panel goes further by letting you specify camera angle and lighting in plain language, which means a solo creator can actually direct a sequence rather than just roll the dice on motion. The fingerprint is still there in the slightly uncanny smoothness of motion transitions, but it's faint enough now that the output clears the bar for social and short-form without a heavy round of manual fixes.”

“The thesis here is that video generation becomes a viable production primitive only when output is composable — meaning a character in shot 5 is recognizably the character from shot 1, which is the minimum requirement for narrative media. That bet is correct and the dependency is tight: it only pays off if creators adopt multi-shot workflows rather than one-off generations, and that adoption hinges on whether the consistency holds under adversarial conditions like wardrobe changes and lighting variance. The second-order effect that nobody's pricing in is what this does to the stock footage and B-roll industry — consistent AI characters at this quality level make licensed human footage economically unjustifiable for a large slice of commercial use cases within 18 months. Luma is on-time to the consistency trend, not early, but they're executing well enough that timing is not the liability.”