🔥Real-time path tracing for million-polygon PBR 3D Asset generated by AI—in your browser.
🤖Built by @pissang1, this WebGPU renderer combines progressive path tracing, OpenPBR materials, and denoising.
🔍We integrated it into Hyper3D.AI to reveal every detail.
New #AI #3D model?🤔❌ Nope.
Still Rodin Gen-2.5.
🤖Now with Agentic Mode.
🔍A whole new level of detail.
Try it out at hyper3d.ai
Hyper3D by Deemos retweeted
Hyper3D just launched WorldGen, and it takes 3D generation from single assets to full scenes.
From one image, it builds a structured 3D scene where every important object is its own editable asset. You can move it, replace it, edit it, and interact with it.
How it works:
1. Upload a scene image
2. WorldGen identifies the objects
3. Generate the full scene with Max Gen and walk around with WASD
4. Select any object and move, delete, replace, or regenerate it with Hyper3D Rodin
5. Click "Interactive" to see physical properties
6. Export to Blender, Unity, or UE
It pairs Rodin's independent Mesh assets with 3D Gaussian Splatting for the environment, and predicts position, scale, collision, mass, and friction.
Built for embodied AI training, film pre-viz with video models like Seedance 2.5, game dev, and XR.
From DeemosTech, whose CAST research won a Best Paper Award at SIGGRAPH 2025.
Try Hyper3D WorldGen:
hyper3d.ai/?utm_source=X&utm…
Get updates @DeemosTech:
linktr.ee/DeemosTech
14-day Free Trial Code: 35TLO1MS
#Hyper3D #DeemosTech #Rodin #WorldGen #AI3D
Hyper3D by Deemos retweeted
What if 3D modeling could plan itself?
Instead of mapping out every step in advance, simply give it a reference. Hyper3D Agentic Mode will analyzes your inputs, including complex reference images, understands the relationships between views, parts, and assemblies, then automatically plans and generates the asset.
If it’s not quite right, just adjust the datails what you’d like to change. Or keep refining it in natural language until it matches your vision.
From analysis to planning, generation, and refinement, this is what agentic 3D modeling looks like.
Hyper3D by Deemos retweeted
I bottled my VFX workflow into an AI Skill: video in, editable Blender scene out—with camera solving, UV-mapped buildings mesh（with @DeemosTech mcp）, alignment, lighting。V1 is alive 👀 Next goal: one clip + “flood this city” → AI-generated CG VFX. Let the chaos begin.
Go 🎣
Available soon
Paper Alert🚨
🤔What if 🦴rigging and 🎬animation could be generated together?
Introducing ACT(SIGGRAPH 2026) — a unified framework for rigging, skinning, and animating 3D
「with any given skeletion topologies」
Soon available at HYPER3D.AI
Details👇
Hyper3D by Deemos retweeted
Could Codex take an AI-generated 3D mesh, use a point cloud as a guide, and scale, stretch, or locally deform it to perfectly match a tracked camera?
I had a day off, so I decided to find out.
As usual, Codex handled image-to-3D generation through @DeemosTech MCP. Then I let it handle the entire alignment process on its own.
And… it worked. Perfectly. 👀
See the video for the results.
I’m packaging this workflow into a skill soon. The idea: shoot a quick video, and automatically get all of this in Blender:
A solved camera
An aligned environment mesh (yes, an actual mesh!)
An environment HDRI and simulated lighting
AI-powered 3D has so much potential for integrating CG into real footage. This is exactly the kind of automation I want: less time on tedious scene setup, more time creating what actually goes into the shot.
Hyper3D by Deemos retweeted
We open-sourced OhMyGame 48 hours ago. Today we have an amazing partner. 🤯
Say hi to Hyper3d 👋 Their 3D model is now built right into OhMyGame, and it's **FREE** for everyone, sponsored by @DeemosTech 🎁
This is the power of open source ❤️
github.com/WhiteTowerAI/ohmy…
Hyper3D at U of Cambridge! ⚡️
5 countries. 8 time zones. 200+ hackers. 24 hours to ship something wild. 🚀
Best part? Seeing OG Hyper3D users bring AI-powered 3D into real workflows.
Less friction. More creation. The future is being hacked into reality. 🔥
#Hyper3D #Hackathon #AI3D
Hyper3D by Deemos retweeted
不会写代码的我，用 AI 不到一天给家人做了一台免费的 3D 推币机《金潮号》。
真实物理推币、九宫格老虎机、会说话的鹦鹉，手机浏览器打开就能玩。
开发全过程在文章里，手把手教程（9 步 + 全部提示词）在第一条回复 👇
Hyper3D by Deemos retweeted
I’ve been messing around with AI 3D tools lately, and I honestly didn’t expect this to work as well as it did.
I gave Hyper3D a single reference with information of the same object and basically told it to figure out the 3D workflow for me.
What surprised me was that it didn’t just generate a model from one image. It actually understood the different views and used them together to build a complete asset.
And then you can keep working on it.
With the Low-Poly N-GONS workflow, I could ask it to add custom controls for different parts, change dimensions, create animations, and even change specific materials just by describing what I wanted.
So it feels less like “generate a 3D model and download it” and more like having something you can actually keep iterating on.
Generate something → look at it → tell it what to change → keep going.
I’ve been having a lot of fun experimenting with this.
Try Agentic Mode on Hyper3d.ai
@DeemosTech
#Hyper3D #DeemosTech #Rodin #AI3D #AgenticMode #3DModeling #AI
Paper Alert🚨
🤔What if 🦴rigging and 🎬animation could be generated together?
Introducing ACT(SIGGRAPH 2026) — a unified framework for rigging, skinning, and animating 3D
「with any given skeletion topologies」
Soon available at HYPER3D.AI
Details👇
A sneak peek to Rodin Rigging👀
Finger Cross🙏Finger-Level✌️
Run🏃♂️Swim🛟Fly🪽, or any motion you can imagine
Coming this month🔥
Made this trial with @OpenAIDevs GPT-6.1 Sol
(4/5)
Once the character has a rig, describe the motion.
“A snake was wriggling along with its head raised.”
“A Tyrannosaurus rex walks forward and opens its mouth.”
ACT generates topology-aware motion for very different characters.
(5/5)
ACT can generate:
• skeleton
• skinning
• text-driven motion
Rigging and motion are generated together as one unified process.
📄:dl.acm.org/doi/abs/10.1145/3…
