Tafi is building a text-to-3D character generator that plugs into Omniverse

If you have ever built a 3D character from scratch, you know the pain. Even with asset libraries, dialing in facial topologies, rigging skeletons, and fine-tuning weight maps can eat up entire workdays.

Tafi, the company behind the popular Daz 3D platform, just announced it is developing a text-to-3D character engine aimed at turning prompts directly into production-ready 3D models.

Unlike experimental AI tools that spit out unrigged digital blobs, Tafi is betting heavily on structured, exportable character data.

Billions of variations straight into game engines

The secret weapon here is Tafi's proprietary Genesis character platform, which artists have relied on for over two decades.

Tafi is tapping this library to create tens of billions of 3D character variations. Each model carries hundreds or thousands of labeled parameters, giving the foundation model an unusually structured training ground.

Because the outputs inherit standard rigging and topology, they can export in native file formats to major pipelines, including Unreal Engine, Unity, Blender, Maya, 3ds Max, and Cinema 4D.

NVIDIA is also stepping in. Tafi confirmed that an upcoming update will support Universal Scene Description (OpenUSD) for direct integration with NVIDIA Omniverse, backed by NVIDIA developer ecosystem VP Richard Kerris.

Getting artists paid instead of scraping them

Most generative AI tools spark immediate backlash from artists, but Tafi is taking a noticeably different route.

The platform has distributed more than $200 million to creators over the last two decades through its revenue-sharing marketplace. For its new AI engine, Tafi built an attribution system that tracks monetization specifically for artists who opt in to contribute training data.

Melbourne-based 3D artist Mada De Leeuw noted that while creators are understandably wary of AI, Tafi's opt-in protection model is earning genuine community support.

Tafi is currently accepting applications for its beta program and evaluating enterprise use cases.

Why it matters

We think Tafi is attacking the single biggest flaw in the current text-to-3D landscape: pipeline readiness.

Most generative 3D tools create messy meshes that look fine in previews but break the moment you try to rig or animate them. By rooting its engine in the existing Genesis framework and supporting native exports to Unreal, Unity, and Blender, Tafi is targeting actual production utility rather than just novelty demos. For indie game developers and small studios, this could compress character prototyping from days to minutes.

Equally important is Tafi's creator strategy. Generative AI is facing severe legal and cultural pushback over uncredited training data, and Tafi's opt-in revenue-sharing model gives it a clean ethical slate that competitors scraping the open web simply do not have.

The real test will be aesthetic diversity. When a model relies on parametric variations of an existing base mesh, outputs can easily look homogenous. We will have to see how far prompts can push beyond standard Daz character tropes once beta testers get their hands on it.