This is nuts.
We're thrilled to release & open-source Hunyuan3D World Model 1.0! This model enables you to generate immersive, explorable, and interactive 3D worlds from just a sentence or an image.
It's the industry's first open-source 3D world generation model, compatible with CG pipelines for full editability & simulation. Set to transform game development, VR, digital content creation and so on. Get started now
Project Page:
https://3d-models.hunyuan.tencent.com/world/
Try it now:
https://3d.hunyuan.tencent.com/sceneTo3D
Github:
https://github.com/Tencent-Hunyuan/HunyuanWorld-1.0
Hugging Face:
https://huggingface.co/tencent/HunyuanWorld-1
Clearly smoke and mirrors.
The potential for tech in this space is industry re-defining. However this incarnation is fake as all fuck.
It seems to demonstrate such a severe perspective skew as the camera rotates, that it's as if it's not generating a 3D landscape at all but a simple spherical skybox.
There seems to be only partial scene elements generated into 3D objects but the rest seems to get baked flat into the skybox.
This seems almost entirely useless for a proper games development pipeline.
I think they'd be far better placed devoting time into developing the models to generating fully realised individual 3D props from reference images or text.
If they can get to the point where you can do that with fully realised rendering into a stable 3D model, with fully decomposed materials (albedo, specular, roughness, lightmaps, normals etc), complete with generated reference shaders, and could also cope with a very wide variety of visual styles (from super hi-fidelity UE5 PBR assets, to stylised Genshin style assets etc) then I think you'd have something truly epic indeed.
Imagine a production process that skips 3D modelling & rigging entirely, and jumping straight from concept artist output > AI 3D asset generation + unwrapping + rigging > a combo of AI-supported + hand-crafted/finessed animation work > designer placement and setup in the game > playtesting.
You'd literally save studios from having to pay for teams of 200-400 3D artists (in-house + offshore), in favour of a handful of "touch up" 3D artists to quality control + fine tune what's coming out of the AI pipeline. You'd also save incredible amounts of time, & allow for wildly more creative iteration, without significant costs for rework (since "pre-vis" asset work basically goes away, and you can master assets at a fraction of the time+cost of today).
Cost + time of 3D art + animation is probably the biggest, most influential factor to the reason modern games average ~5-8 year production cycles.
Imagine being able to collapse that entirely.
Finally, think about the control and empowerment this puts into the hands of indies...
A small independent team could make titles at the same or higher level of quality we see in the AAA arena today, providing them with far more leverage, more scope for risk taking and creative freedom, more opportunity to grow their businesses without selling their souls to greedy, malicious and risk averse publishers, in favour of $200m production financing contracts. They can retain IP control which means more control over profits, which will ultimately lead us to I hope another golden age of games made almost entirely by gamers for gamers (less IAPs, less gatcha, loot boxes, less GaaS and other bullshit purely financial data-lead trend-chasing), with significantly less influence from talentless bean counters in the equation.
A man can dream...