A duo of Ubisoft and Unity veterans just shipped the world-building pipeline AAA studios have been dreaming about for a decade. Dreamlands turns a plain-text conversation into thousands of stylistically coherent 3D meshes, assembled into a complete game-ready world — in minutes. Not months.
The Story
Thomas Tobin spent years as a Technical Art Director at Ubisoft — shipping titles like Far Cry and Rainbow Six Siege Mobile. He watched world artists burn through month-long sprints generating environments that got scrapped when the creative direction shifted. His co-founder Jas Rowinski built the GPU infrastructure at Unity that scaled to 1,000+ machines for Apple Vision Pro content delivery. Together, they spent four years tackling the same bottleneck from both ends — creative and infrastructure — before founding Dreamlands.
The premise is simple and brutal: “The bottleneck was never creativity — it was the time between having an idea and being able to see it.” Dreamlands compresses that gap from months to minutes.
Here’s how the pipeline works: you describe an environment in plain language — biome, aesthetic, era, mood, density. Dreamlands feeds that into a Gemini-powered research layer that validates architectural and environmental authenticity (it actually checks whether that building style existed in that historical period, whether that flora grows in that climate). Then a Vertex AI orchestration layer runs concept art generation → textured 3D mesh generation → automatic world placement in parallel. Output: thousands of stylistically coherent meshes, assembled into a complete scene, ready to export one-click to Unity or Unreal Engine.
At GDC 2026, Google Cloud placed Dreamlands alongside Capcom, Sony PlayStation, and 10Six Games as a flagship example of where AI-driven game creation is headed. That is not an accidental co-billing — it signals that major studios are already evaluating (and using) this pipeline in production.
Why You Should Care
World-building is the most expensive, most wasteful phase of AAA game development. Typical cost: six figures and three to six months for a single biome. And that’s before the creative director says “let’s try a different tone.” Studios historically have two options: invest in the full pipeline and pray the vision doesn’t change, or prototype in gray box hell and lose the visual coherence when you upres.
Dreamlands breaks that constraint entirely. Because the generation is conversational and fast, you can iterate on world feel the same way you iterate on a character design in Midjourney — generate a direction, react to it, refine, generate again. The difference is that what comes out the other end is production-grade geometry, not a mood board.
For 3D artists: this is not a threat to your job. The interesting work — hero assets, character geometry, unique architectural landmarks — still needs human craft. What Dreamlands kills is the repetitive scatter work: that 400-tree forest, those 200 background buildings, the endless ground cover variations. You stop generating filler and start designing landmarks.
For indie studios and small teams: the platform offers a free tier on Google Cloud Marketplace and has committed to affordable subscription plans for studios under $1M revenue. You can now prototype a world that looks like a AAA environment without a world art team.
For creative technologists: the Gemini + Vertex AI stack powering this is the same stack you can build on. The architectural pattern — LLM-as-research-layer + orchestrated generation pipeline + engine export — is a blueprint worth studying for any domain that needs coherent large-scale creative output.
Try It / Follow Them
- Platform: dreamlands.ai — free tier available on Google Cloud Marketplace
- Engine plugins: One-click Unity and Unreal Engine integrations; engine-agnostic architecture for custom engines
- Launch trailer: YouTube — Dreamlands AI World Generation Demo
- Founders on LinkedIn: Thomas Tobin and Jas Rowinski
- Context: Google Cloud GDC 2026 announcement
IK3D Lab Take
What makes Dreamlands different from every other “AI 3D generation” tool we’ve covered isn’t the technology — it’s the pipeline philosophy. Most tools stop at the asset. Dreamlands starts from world intent and delivers assembled environments. That’s a categorically different thing.
The Gemini research layer is the detail that nobody talks about but that makes the whole thing work at production quality. Generating 3D meshes is solvable. Generating 3,000 meshes that all feel like they belong in the same world, are historically plausible, and have the right visual density for a given gameplay context — that requires a layer of world knowledge that pure generative models don’t have. Dreamlands built that layer.
The Google Cloud Marketplace positioning is also smart. It means studios can run their Dreamlands pipeline on the same infrastructure where they’re already running their game backends, CI, and playtesting. No new vendor relationships, no new security review — just a new tool in the same environment. That removes the biggest enterprise adoption friction.
Keep this one on your radar. The team has the credibility (Ubisoft + Unity), the infrastructure chops (1,000-machine GPU farms), and the right backers (Google Cloud partnership at GDC). If the free tier opens up broadly, this could be the world-building tool that finally makes “solo dev, AAA-scale environments” a real thing — not just a promise.
