Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

GPT-6 changes 3D modeling forever. It works inside Blender completely autonomously, like an artist: gathers reference photos on its own, builds the scene, renders test frames, checks them against the references, then fixes what's off. When the scene holds up, it ships it to Unreal Engine 5 as a... show more

Stefan 3D AI

14,675 subscribers

161,522 views • 1 month ago •via X (Twitter)

Arts Science & Technology Gaming

Anya Rossi• Live Now

Private livecam show

34 Comments

Is this a 3D model?1 month ago

it will not change modeling literally at all lmao

Stefan 3D AI1 month ago

The Palace of Fine Arts in San Francisco, rebuilt by Astra in Blender. It found hundreds of photos of the real building, rendered intermediate frames, and compared them against the references until the scene matched. 1.5x speed.

Stefan 3D AI1 month ago

A house modeled in Blender from a design drawing, then imported into Unreal Engine 5 as an explorable scene. Game studio Playco ran Astra on a single greybox and got three playable prototypes, with about half the manual fixes of their usual pipeline.

Stefan 3D AI1 month ago

The kart racer from the launch page, playable right in the browser. More from day one: a voxel London shifting between medieval, Tudor and modern eras (@chetaslua), and Pietro Schirano's underwater exploration game built from pictures, with sound and gameplay.

OS'TA1 month ago

It's not 3D modelig

thasxzoo1 month ago

This is so sad honestly

Aeris Vahn Ephelia - 𝑹𝑬𝑽1 month ago

You don't fear Stefan that even your channel and current work will be obsoletes by 2027 ?

Webster | JARVIS1 month ago

holy shit imagine shelling out for a 3d artist and the intern just installs gpt-6 instead blender + unreal pipeline in one prompt is wild though. the whole "checks its own renders and fixes them" part is what actually scares me

Majid Manzarpour1 month ago

excited to see what you cook with it!

Den Bersus1 month ago

AI bros have been claiming every single ChatGPT/Claude update changes 3D modeling forever since GPT-3.

mrMatiox ⚡1 month ago

Please @OpenAI don't destroy artist 💀💀

Gregor1 month ago

the 'checks against references, then fixes' loop is the part i'd want to see uncut. does it converge in 2-3 passes or is it still drifting at iteration 8?

Princess Mononoke1 month ago

Can someone explain how it works? Can I input an image into GPT-6 and get the model? Does it work on character model?

Abhimanyu ARYAN1 month ago

So GTA 6 will be released sooner now?

Paca amarilla1 month ago

🤮🤮🤮

JesusIsGodAndKing1 month ago

@TokenGremlin

Alan Tobin1 month ago

Amazing those days of having to be a button wizard at ever piece of software is nearly over. Natural language Interface gets better and better every 2 to 3 months now.

Thom Aster1 month ago

Now couple this with the Lidar in your router... "It'S aLl FoR yOuR sAfEtY/EnTeRtAiNtMeNt"

Leozera1 month ago

Viu isso Vô? @lgriao

Pearce1 month ago

我觉得许多美术师看到这个快失眠了🤫

Pockets Full ⭕f NFT's1 month ago

Wow who thought of this

Ricardo Guia1 month ago

@grok did they use blender? And what else?

@OmnisVibe1 month ago

Wow

Jennie1 month ago

gpt 6 automation is wild, but for manual control i rough scenes in krea since image to 3D object you can rotate before rendering keeps me directing. different use cases

Crimson Skald1 month ago

That's gonna run like arse, without optimization.

ClipStudio1 month ago

the greybox to warm lit room cut is so satisfying

Singulith1 month ago

すごい...😲

Hatsuki J Graetz1 month ago

i cant wait to see you test it

LustRazor1 month ago

Looking forward to trying it. Anthropic models were not great at modeling.

Loong🐉1 month ago

The 'render → check → fix' loop is the part that actually matters here. Most 'autonomous 3D' demos skip the second pass and ship whatever the first prompt gives back. If Astra closes that loop without human QA, indie art-pipelines just collapsed into one afternoon.

Michael Swengel1 month ago

lol. With terrible topology, embarrassing UVs and borderline unusable assets?

Field Marshal1 month ago

Big if true

MARUF1 month ago

blender viewport rendering and mesh hierarchy look ridiculously clean rn

Elliott1 month ago

I’m team Grok, but I will say, this is quite the accomplishment, and it is a testament to the power of world modeling. Manipulating an environment in true three dimensions to match a desired outcome is a big step.

Related Videos

Astra is very good at 3D modeling, and I can't wait for all of you to experience it, for now here is a little walkthrough on how I built the demo house for our launch blog post. From a Blender scene to a Unreal Engine 5 walkable experience. 🧵

Astra is very good at 3D modeling, and I can't wait for all of you to experience it, for now here is a little walkthrough on how I built the demo house for our launch blog post. From a Blender scene to a Unreal Engine 5 walkable experience. 🧵

Thomas Ricouard

2,955,123 views • 2 months ago

here’s how to build a game-ready room in blender and move it into unreal engine 5 with claude opus 5.5... nobody opened a 3d app by hand, and now it runs in real time the trick is letting opus decide how each object gets made, smart way to set it up: - step 1: connect opus 5.5 to blender through blender mcp - step 2: give it one reference image for the room: layout, mood, light - step 3: simple objects get modeled directly in blender (walls, desk, shelves, boxes, blinds) - step 4: complex props get a reference image first, then tripo turns it into 3d (chair, monitors, figurines) - step 5: opus cleans up the tripo props in blender and places them in the room - step 6: light the scene in blender until the stills look right - step 7: have opus port the whole scene from blender to unreal engine 5 - step 8: the room now runs live in the engine, ready for gameplay the agent works, you direct. he still checked every result and gave notes along the way

here’s how to build a game-ready room in blender and move it into unreal engine 5 with claude opus 5.5... nobody opened a 3d app by hand, and now it runs in real time the trick is letting opus decide how each object gets made, smart way to set it up: - step 1: connect opus 5.5 to blender through blender mcp - step 2: give it one reference image for the room: layout, mood, light - step 3: simple objects get modeled directly in blender (walls, desk, shelves, boxes, blinds) - step 4: complex props get a reference image first, then tripo turns it into 3d (chair, monitors, figurines) - step 5: opus cleans up the tripo props in blender and places them in the room - step 6: light the scene in blender until the stills look right - step 7: have opus port the whole scene from blender to unreal engine 5 - step 8: the room now runs live in the engine, ready for gameplay the agent works, you direct. he still checked every result and gave notes along the way

rewind

84,372 views • 15 days ago

🚨 STOP learning Blender !! Learning Blender in 2026..... Is officially obsolete. MiniMax Design can now automate Blender to create ANY 3d SCENE !! I just gave it a PROMPT.... And MiniMax Design used GPT-6 Astra to handle everything like: 1️⃣ Opening Blender 2️⃣ Generating the 3D elements 3️⃣ Fixing lighting 4️⃣ Camera Tracking Data etc.... It built the entire scene inside Blender. Finally, I used MiniMax H3 Max to add colors to the 3d SCENE !! Then I just rendered the 3d scene !! Best part? Using GPT-6 Astra through MiniMax Design is way more affordable than running it directly.

🚨 STOP learning Blender !! Learning Blender in 2026..... Is officially obsolete. MiniMax Design can now automate Blender to create ANY 3d SCENE !! I just gave it a PROMPT.... And MiniMax Design used GPT-6 Astra to handle everything like: 1️⃣ Opening Blender 2️⃣ Generating the 3D elements 3️⃣ Fixing lighting 4️⃣ Camera Tracking Data etc.... It built the entire scene inside Blender. Finally, I used MiniMax H3 Max to add colors to the 3d SCENE !! Then I just rendered the 3d scene !! Best part? Using GPT-6 Astra through MiniMax Design is way more affordable than running it directly.

Darshal Jaitwar

86,087 views • 1 month ago

How about one image straight to an editable UE5 or Blender scene? Lumera just dropped paper - Engine-Native Editable 3D World Reconstruction with Objects & Lighting. You give it an image. It figures out what's in the scene, rebuilds every object as a separate mesh, then recreates the lighting as actual engine lights you can move and tweak. Even pulls an HDR environment out of the shot. The lighting part is the big deal here. Every other image-to-3D approach bakes it in. Lumera keeps it a parameter, like the scene was built in the engine from the start. Code and weights marked as "Soon". Paper -

How about one image straight to an editable UE5 or Blender scene? Lumera just dropped paper - Engine-Native Editable 3D World Reconstruction with Objects & Lighting. You give it an image. It figures out what's in the scene, rebuilds every object as a separate mesh, then recreates the lighting as actual engine lights you can move and tweak. Even pulls an HDR environment out of the shot. The lighting part is the big deal here. Every other image-to-3D approach bakes it in. Lumera keeps it a parameter, like the scene was built in the engine from the start. Code and weights marked as "Soon". Paper -

Stefan 3D AI

47,223 views • 2 months ago

Gemini 4 Argon turned 10 photos into a walkable 3D apartment. It built the whole thing in Blender and Unreal Engine. Guests could tour the place before they even book.

Gemini 4 Argon turned 10 photos into a walkable 3D apartment. It built the whole thing in Blender and Unreal Engine. Guests could tour the place before they even book.

Higgsfield AI 🧩

30,271 views • 19 days ago

This is much worse for 3D artists than another AI image generator. Claude takes one architectural reference, connects directly to Blender, and starts rebuilding it as an actual editable scene. It blocks out the buildings, matches the camera, adds materials, vegetation, lighting, and then renders the result. The interesting part is not the first generation. Claude inspects its own render, notices the composition and lighting are wrong, then goes back into Blender and keeps changing the scene. The final result is still rough. The materials look basic, the geometry needs cleanup, and an experienced artist would make it far better. But AI is no longer producing a flat image and pretending it understands 3D. It is operating the software, judging its own work, and iterating inside the real project. The “AI cannot do their actual work” argument is getting harder to defend.

This is much worse for 3D artists than another AI image generator. Claude takes one architectural reference, connects directly to Blender, and starts rebuilding it as an actual editable scene. It blocks out the buildings, matches the camera, adds materials, vegetation, lighting, and then renders the result. The interesting part is not the first generation. Claude inspects its own render, notices the composition and lighting are wrong, then goes back into Blender and keeps changing the scene. The final result is still rough. The materials look basic, the geometry needs cleanup, and an experienced artist would make it far better. But AI is no longer producing a flat image and pretending it understands 3D. It is operating the software, judging its own work, and iterating inside the real project. The “AI cannot do their actual work” argument is getting harder to defend.

Rina

35,870 views • 3 months ago

OpenAI removed the Astra blog in less than 2 minutes. Here are the two outputs I captured: 1. A playable web game 2. Modeled a house in Blender and turned into a walkable scene in Unreal Engine 5

OpenAI removed the Astra blog in less than 2 minutes. Here are the two outputs I captured: 1. A playable web game 2. Modeled a house in Blender and turned into a walkable scene in Unreal Engine 5

Shikhar

73,101 views • 2 months ago

I kept seeing #GPT-6 Astra modelling results on here and got curious enough to build the thing myself: an agentic pipeline that turns a plain RGB video into 3D assets. The living room for robots MIT CSAIL. Input: one handheld phone walkthrough of our lab kitchen (~20s). Nothing else — no depth sensor, no CAD, no asset library. ~1 day (I actually slept overnight), including human-in-the-loop. Live: It is still not perfect — thin and shiny things are still weak, a few objects are drafts, the room shell needs another pass. The loop, roughly: - Monocular video → metric scan (ViPE): camera poses + depth. This is the measuring instrument, not the output. - GPT-6 lists what should exist as separate objects, then open-vocabulary detection + tracking gives per-object masks; each object is fused and measured in metres. - For every asset, GPT-6 works in its own sandbox with tools: it writes the object as a program in a small Blender DSL (closed solids, PBR colours, hinges/drawers), builds it, renders it over the original video frames, compares it with the scan points in 3D, and iterates. - A separate GPT-6 session is the verifier — it can render the model on any frame it wants and must justify every complaint with a frame. The modeller never grades its own work. - A completeness pass renders the whole modelled scene from the video's own cameras, puts it next to the real frames, and says what's still missing. What the detector keeps missing (a row of identical cabinets) gets placed geometrically instead. - Out comes MJCF/URDF with joints. Now, adding simulation to this env.

I kept seeing #GPT-6 Astra modelling results on here and got curious enough to build the thing myself: an agentic pipeline that turns a plain RGB video into 3D assets. The living room for robots MIT CSAIL. Input: one handheld phone walkthrough of our lab kitchen (~20s). Nothing else — no depth sensor, no CAD, no asset library. ~1 day (I actually slept overnight), including human-in-the-loop. Live: It is still not perfect — thin and shiny things are still weak, a few objects are drafts, the room shell needs another pass. The loop, roughly: - Monocular video → metric scan (ViPE): camera poses + depth. This is the measuring instrument, not the output. - GPT-6 lists what should exist as separate objects, then open-vocabulary detection + tracking gives per-object masks; each object is fused and measured in metres. - For every asset, GPT-6 works in its own sandbox with tools: it writes the object as a program in a small Blender DSL (closed solids, PBR colours, hinges/drawers), builds it, renders it over the original video frames, compares it with the scan points in 3D, and iterates. - A separate GPT-6 session is the verifier — it can render the model on any frame it wants and must justify every complaint with a frame. The modeller never grades its own work. - A completeness pass renders the whole modelled scene from the video's own cameras, puts it next to the real frames, and says what's still missing. What the detector keeps missing (a row of identical cabinets) gets placed geometrically instead. - Out comes MJCF/URDF with joints. Now, adding simulation to this env.

Zhiyang (Frank) Dou

11,516 views • 1 month ago

here's how to build a game like this with Claude Opus 5.5 + Blender + Unreal Engine + Meshy... the trick is giving every tool one job, but there's a smart way to set it up: - step 1: install Claude Code, Unreal Engine 5 and Blender 5.1+, all free except the Claude plan - step 2: connect Blender through its MCP add-on, and Meshy through Meshy MCP with an API key - step 3: brainstorm 5 game ideas with Opus first, pick one, have it expand the idea before anything gets built - step 4: tell it your GPU, so it pushes the graphics as far as your machine can run - step 5: Meshy generates the props and characters, Opus imports them into Blender and cleans them up - step 6: Opus brings everything into Unreal: the level, lighting, HUD, minimap - step 7: play it, then add one feature per prompt: combat, enemies, quests who does what: > Meshy: the 3D models > Blender: cleanup, placement, materials > Unreal: lighting, physics, the actual game > Opus: runs all three from one chat and the real cost... one full first build took about 4 hours and ~19% of a weekly Max plan build the first level well, then the next ones go much faster because the assets already exist

here's how to build a game like this with Claude Opus 5.5 + Blender + Unreal Engine + Meshy... the trick is giving every tool one job, but there's a smart way to set it up: - step 1: install Claude Code, Unreal Engine 5 and Blender 5.1+, all free except the Claude plan - step 2: connect Blender through its MCP add-on, and Meshy through Meshy MCP with an API key - step 3: brainstorm 5 game ideas with Opus first, pick one, have it expand the idea before anything gets built - step 4: tell it your GPU, so it pushes the graphics as far as your machine can run - step 5: Meshy generates the props and characters, Opus imports them into Blender and cleans them up - step 6: Opus brings everything into Unreal: the level, lighting, HUD, minimap - step 7: play it, then add one feature per prompt: combat, enemies, quests who does what: > Meshy: the 3D models > Blender: cleanup, placement, materials > Unreal: lighting, physics, the actual game > Opus: runs all three from one chat and the real cost... one full first build took about 4 hours and ~19% of a weekly Max plan build the first level well, then the next ones go much faster because the assets already exist

rewind

194,759 views • 25 days ago

From the Flux 3 character reference to a fully rigged character in Unreal Engine! Generated the high-poly model with Tripo 3.1, then used the Blender Agent inside the workflow to bake it down to low poly. Atlas rigging tools handled the rig and animations. The fingers came out especially well! Added cloth physics in Unreal Engine to give the generated cape mesh more movement. Check out the full workflow in the comments.

From the Flux 3 character reference to a fully rigged character in Unreal Engine! Generated the high-poly model with Tripo 3.1, then used the Blender Agent inside the workflow to bake it down to low poly. Atlas rigging tools handled the rig and animations. The fingers came out especially well! Added cloth physics in Unreal Engine to give the generated cape mesh more movement. Check out the full workflow in the comments.

Atlas

140,511 views • 9 days ago

Looking right in 2D doesn’t really mean the scene holds together in 3D. That's why we open-sourced BlenderWorld last week and wanted to give it a closer look here. It’s a benchmark with 223 tasks where agents use Blender to build editable 3D scenes from text, images or video references. We evaluate the actual projects through scene inspection and renders from multiple viewpoints. Can a visual agent turn a real-world brief into a complete, editable 3D scene? BlenderWorld is here to evaluate that You can grab the benchmark below (Links in the comment) and plug in your own 3D agent to see how it does

Looking right in 2D doesn’t really mean the scene holds together in 3D. That's why we open-sourced BlenderWorld last week and wanted to give it a closer look here. It’s a benchmark with 223 tasks where agents use Blender to build editable 3D scenes from text, images or video references. We evaluate the actual projects through scene inspection and renders from multiple viewpoints. Can a visual agent turn a real-world brief into a complete, editable 3D scene? BlenderWorld is here to evaluate that You can grab the benchmark below (Links in the comment) and plug in your own 3D agent to see how it does

Qwen Developers

58,478 views • 10 days ago

The bike, the ruins, the ramps, the Highlands sky. All of it started as one sentence typed into Atlas. Concept art, 3D assets, a staging pass in Blender, an Unreal scene you can ride through. One canvas, so the world keeps one look from first image to engine. Try Atlas today.

The bike, the ruins, the ramps, the Highlands sky. All of it started as one sentence typed into Atlas. Concept art, 3D assets, a staging pass in Blender, an Unreal scene you can ride through. One canvas, so the world keeps one look from first image to engine. Try Atlas today.

Atlas

8,790,944 views • 2 months ago

I gave GPT Astra a clip from Inception, it reconstructed the entire scene in Blender.

I gave GPT Astra a clip from Inception, it reconstructed the entire scene in Blender.

Xiao Tan

150,841 views • 1 month ago

GPT-6 coded an entire world: a busy harbor you can step inside. From a single prompt and an asset pack, it assembled a container port with towering cranes, docked ships, rows of containers, and forklifts across the yard, all within an explorable 3D scene. A first look at our upcoming Code-for-3D benchmark, featuring the text-to-scene setting.

GPT-6 coded an entire world: a busy harbor you can step inside. From a single prompt and an asset pack, it assembled a container port with towering cranes, docked ships, rows of containers, and forklifts across the yard, all within an explorable 3D scene. A first look at our upcoming Code-for-3D benchmark, featuring the text-to-scene setting.

SimWorld

25,313 views • 1 month ago

GPT-6 Astra turned a single floor plan into a full 3D walkthrough. It reconstructed the layout in Blender, furnished every room, lit the scene with natural light, and rendered the tour with ray tracing. Made end to end on Higgsfield Supercomputer.

GPT-6 Astra turned a single floor plan into a full 3D walkthrough. It reconstructed the layout in Blender, furnished every room, lit the scene with natural light, and rendered the tour with ray tracing. Made end to end on Higgsfield Supercomputer.

Higgsfield AI 🧩

59,217 views • 2 months ago

Someone tested Union-Alpha, the new stealth model on OpenRouter, with their usual shooter prompt and it built a full playable FPS. The menu alone gives it away: huge title, orange arrow button, dark 3D scene, the exact same look as GPT-6 Astra. Feels like OpenAI is behind.

Someone tested Union-Alpha, the new stealth model on OpenRouter, with their usual shooter prompt and it built a full playable FPS. The menu alone gives it away: huge title, orange arrow button, dark 3D scene, the exact same look as GPT-6 Astra. Feels like OpenAI is behind.

ALICE ⚡ | AI Growth Mentor

108,338 views • 1 month ago

🚨 A single image → a fully editable UE5 world. Introducing Lumera: a turnkey pipeline that reconstructs an engine-native 3D scene directly from ONE image — not just pixels, point clouds, or Gaussians. It gives you: → object-level 3D meshes → editable 3D layouts → engine-native lights → HDR environment lighting → a scene ready for Unreal Engine 5 In other words: Image → 3D world → UE5. We believe the future of image-to-3D is not “generate something that looks 3D.” It is generate the actual world project. 🌐 📄

🚨 A single image → a fully editable UE5 world. Introducing Lumera: a turnkey pipeline that reconstructs an engine-native 3D scene directly from ONE image — not just pixels, point clouds, or Gaussians. It gives you: → object-level 3D meshes → editable 3D layouts → engine-native lights → HDR environment lighting → a scene ready for Unreal Engine 5 In other words: Image → 3D world → UE5. We believe the future of image-to-3D is not “generate something that looks 3D.” It is generate the actual world project. 🌐 📄

Hao Zhao

52,630 views • 2 months ago

i made AnyScene to build any scene in an hour. i built AnyShot to render those scenes in a few minutes. on my RTX 5070 Ti, a single walkthrough like this takes 6 hours to render. AnyShot fans it out across up to 100 RTX 5090s on fal and renders it in 3 minutes. but it’s not just a render farm. point it at any Blender scene, give it a brief, and an AI cinematographer plans every camera move, checks it never clips through the set, pulls focus, and shoots it. like AnyScene, AnyShot is open source

i made AnyScene to build any scene in an hour. i built AnyShot to render those scenes in a few minutes. on my RTX 5070 Ti, a single walkthrough like this takes 6 hours to render. AnyShot fans it out across up to 100 RTX 5090s on fal and renders it in 3 minutes. but it’s not just a render farm. point it at any Blender scene, give it a brief, and an AI cinematographer plans every camera move, checks it never clips through the set, pulls focus, and shoots it. like AnyScene, AnyShot is open source

apek

16,246 views • 1 day ago

GPT-6 Astra just vibe-coded the Oval Office in 3D. We described the set design > GPT-6 Astra generated the scene code > Higgsfield built the location model in Blender and rendered it with Cycles. The whole pipeline ran end-to-end on Higgsfield Supercomputer.

GPT-6 Astra just vibe-coded the Oval Office in 3D. We described the set design > GPT-6 Astra generated the scene code > Higgsfield built the location model in Blender and rendered it with Cycles. The whole pipeline ran end-to-end on Higgsfield Supercomputer.

Higgsfield AI 🧩

1,251,067 views • 2 months ago

THIS STARTUP FORKED BLENDER AND REBUILT IT AROUND AI AGENTS. it's called Mixar. you open a 3D scene and can: → tell an agent what to change directly inside the scene → generate 3D assets from references → retopologize, texture and clean assets → run multiple agents across different scenes → use 400+ AI models from one place they forked Blender itself and are turning it into an agentic 3D platform.

THIS STARTUP FORKED BLENDER AND REBUILT IT AROUND AI AGENTS. it's called Mixar. you open a 3D scene and can: → tell an agent what to change directly inside the scene → generate 3D assets from references → retopologize, texture and clean assets → run multiple agents across different scenes → use 400+ AI models from one place they forked Blender itself and are turning it into an agentic 3D platform.

matt

85,625 views • 1 month ago