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.

Presenting: Voxel global illumination demo in three.js (Link in comments) Uses voxelization in a worker thread, blue-noise powered stochastic sampling, and A-torus denoising to create a realtime GI solution that supports completely dynamic scenes. It runs at 60-80fps (w/out recording software) on an M1 mac! #threejs #webgl #gi #vxgi

N8 Programs

10,236 subscribers

34,378 views • 2 years ago •via X (Twitter)

Arts Science & Technology Education

Anya Rossi• Live Now

Private livecam show

10 Comments

N8 Programs2 years ago

Link:

N8 Programs2 years ago

The demo is currently broken. This will be fixed soon.

N8 Programs2 years ago

It has been rolled back to an earlier version - will work on improving it for pixel compat soon.

N8 Programs2 years ago

Some clarifications about the benefits of this technique: - Works in world space, things that go offscreen still create illumination - Completely dynamic, supports all objects and lights moving each frame and cons: - Expensive both CPU and GPU side - Requires lots of work to adapt to new materials and scenes, not very customizable.

mrdoob2 years ago

@threejs Looking great! 🔥 Some artifacts on Pixel 7:

Cody Bennett2 years ago

Getting some weird stuff on Windows/DirectX via Chrome. Will look more into this later. CORS issues prevent me from testing on MacOS Safari but works great there on Chrome.

N8 Programs2 years ago

This could be because I have a function called sampled1d... just pushed a fix - let me know if it helps.

herbst2 years ago

Starts to look really good and actually runs at ~28 fps on my phone, pretty nice!

N8 Programs2 years ago

TYSM!

mark2 years ago

amazing work

Related Videos

Jolt multi-threaded is crazy🤯 Stresstest using our new ECS library, Three.js BatchedMesh and Jolt Physics. 1000+ dynamic bodies, 120 fps on a low-end PC *while* recording. But try it for yourself! Demo Link + Code in the comments. #gamedev #webgamedev

Jolt multi-threaded is crazy🤯 Stresstest using our new ECS library, Three.js BatchedMesh and Jolt Physics. 1000+ dynamic bodies, 120 fps on a low-end PC while recording. But try it for yourself! Demo Link + Code in the comments. #gamedev #webgamedev

Brian Breiholz

17,993 views • 1 year ago

A first look at Surface Cache GI in Unity! Get robust, fully dynamic global illumination in real time - meaning no more tedious authoring or waiting on lightmap bakes. Surface Cache Gl is ready to preview in 6.7 alpha today, and will become broadly available with Unity 6.7 LTS. Watch how it brings Sakura Rabbit 樱花兔’s Mecha Repair Hangar to life ⬇️

A first look at Surface Cache GI in Unity! Get robust, fully dynamic global illumination in real time - meaning no more tedious authoring or waiting on lightmap bakes. Surface Cache Gl is ready to preview in 6.7 alpha today, and will become broadly available with Unity 6.7 LTS. Watch how it brings Sakura Rabbit 樱花兔’s Mecha Repair Hangar to life ⬇️

Unity

21,854 views • 8 days ago

It looks like #HaloCampaignEvolved relies exclusively on Hardware Lumen for its Global Illumination and Reflections, with no Software Lumen fallback. I did a quick test on the 3060 Ti by disabling the “Raytracing (DXR)” flag in NVIDIA Profile Inspector, which basically disables the GPU's ray tracing support. This gives a significant performance boost, but it completely disables the game’s Lumen GI and Reflections. So even though the game uses Hardware Lumen, it can still run on non-RT GPUs, albeit with substantially degraded visuals.

BenchmarKing

148,402 views • 4 days ago

You can now vibecode your own WisprFlow or Monologue alternative that runs completely locally on Apple Silicon using MLX-Audio-Swift 🔥 Check out this live transcription of Dwarkesh Patel interview with Andrej Karpathy using Qwen3-ASR-0.6B quantized to 4bit on a M3 Max. It also runs in realtime on a iPhone 15 Pro and iPad Pro M1. No cloud. No API keys.

You can now vibecode your own WisprFlow or Monologue alternative that runs completely locally on Apple Silicon using MLX-Audio-Swift 🔥 Check out this live transcription of Dwarkesh Patel interview with Andrej Karpathy using Qwen3-ASR-0.6B quantized to 4bit on a M3 Max. It also runs in realtime on a iPhone 15 Pro and iPad Pro M1. No cloud. No API keys.

Prince Canuma

60,974 views • 5 months ago

Video on how to set up a pre release #Toccata $KAS node for Mac users. It is so simple to do we should have 10,000 nodes running… IzioDev -Download the software for Mac at the link in Michael’s tweet -Open the folder. Copy the file path for Kaspad -Paste it in a terminal window, press space, then type “- - utxoindex” (no space between the two dashes). -Press enter and let it run. Use a mini pc or something without a display that can run 24/7. I am using an M1 Mac Mini.

Video on how to set up a pre release #Toccata $KAS node for Mac users. It is so simple to do we should have 10,000 nodes running… IzioDev -Download the software for Mac at the link in Michael’s tweet -Open the folder. Copy the file path for Kaspad -Paste it in a terminal window, press space, then type “- - utxoindex” (no space between the two dashes). -Press enter and let it run. Use a mini pc or something without a display that can run 24/7. I am using an M1 Mac Mini.

₿RT 𐤊 🐈📈

12,858 views • 1 month ago

1/ Introducing web2mcp: a toolchain that auto-generates an MCP interface for any website, powered by @stagehanddev + GPT-5. It uses computer-use-preview to map out all actions in a web app, then compiles them into an MCP spec. Demo on a sample app:

1/ Introducing web2mcp: a toolchain that auto-generates an MCP interface for any website, powered by @stagehanddev + GPT-5. It uses computer-use-preview to map out all actions in a web app, then compiles them into an MCP spec. Demo on a sample app:

Neel Somani

51,223 views • 11 months ago

Few know this yet, but Hyperfy has built an insanely well-performing WebXR engine. 🔹 Almost 100Hz feel on my Quest 3 🔹 Runs buttery smooth in a browser, standalone 🔹 Even screen recording doesn’t cause noticeable frame drops Hard to believe this is WebXR. Three.js #WebXR 🔗 Link to try it yourself in the comments!

Few know this yet, but Hyperfy has built an insanely well-performing WebXR engine. 🔹 Almost 100Hz feel on my Quest 3 🔹 Runs buttery smooth in a browser, standalone 🔹 Even screen recording doesn’t cause noticeable frame drops Hard to believe this is WebXR. Three.js #WebXR 🔗 Link to try it yourself in the comments!

𝚟𝚘𝚡𝚟𝚒𝚎𝚗𝚗𝚎

19,750 views • 1 year ago

you don’t need a robot to see physical ai in action, you just need a browser. this week I finally tried something I've wanted to do for a while: design an SO-101 scene in three.js, connect it to a real leader arm, collect teleoperation data, train an ACT model, then port it to run in the browser. the result is the first ever demo of ACT that runs locally on the web. try it yourself now at

you don’t need a robot to see physical ai in action, you just need a browser. this week I finally tried something I've wanted to do for a while: design an SO-101 scene in three.js, connect it to a real leader arm, collect teleoperation data, train an ACT model, then port it to run in the browser. the result is the first ever demo of ACT that runs locally on the web. try it yourself now at

Binh Pham

120,373 views • 1 month ago

Excited to announce that I've shipped my first game! Written from the ground up in Three.js in 3 months and published on VIVERSE Official Some tech stuff that I'm really proud of: - fully custom spherical harmonics prt lighting. Giving a baked diffuse look while still allowing for dynamic objects - Runs completely inside a WebWorker (rendering & simulation) using OffscreenCanvas - Supports all input types: desktop, mobile, gamepad (non-trivial if your app runs in a worker!) - No Postprocessing! No bloom - everything you see is faked via mesh based shaders for performance reasons Huge thanks to VIVERSE Official who have made this possible by letting me into their Creator Program. It's not easy to make a living off gamedev on the web yet, so this is great for the ecosystem. All links below: #VIVERSE #threejs #IndieGameDev

Excited to announce that I've shipped my first game! Written from the ground up in Three.js in 3 months and published on VIVERSE Official Some tech stuff that I'm really proud of: - fully custom spherical harmonics prt lighting. Giving a baked diffuse look while still allowing for dynamic objects - Runs completely inside a WebWorker (rendering & simulation) using OffscreenCanvas - Supports all input types: desktop, mobile, gamepad (non-trivial if your app runs in a worker!) - No Postprocessing! No bloom - everything you see is faked via mesh based shaders for performance reasons Huge thanks to VIVERSE Official who have made this possible by letting me into their Creator Program. It's not easy to make a living off gamedev on the web yet, so this is great for the ecosystem. All links below: #VIVERSE #threejs #IndieGameDev

Brian Breiholz

14,587 views • 1 year ago

So i vibe coded a threejs webgpu splat renderer from scratch in an evening with gpt 5.4 Introducing Spark++ Give it a week or two, performance is slightly worse on fast camera movements but 60 fps on my toaster in less dense areas. Had to reduce overall density for screen recording, but it works on webgpu! And I have absolutely no clue how Splats work, zero Demo: Repo:

So i vibe coded a threejs webgpu splat renderer from scratch in an evening with gpt 5.4 Introducing Spark++ Give it a week or two, performance is slightly worse on fast camera movements but 60 fps on my toaster in less dense areas. Had to reduce overall density for screen recording, but it works on webgpu! And I have absolutely no clue how Splats work, zero Demo: Repo:

robot

32,580 views • 4 months ago

how many of you send things to yourself on whatsapp and never find them again? i tried fixing it. built a mac app that lives in your notch. - drag and drop any media. - ask it anything later. - uses local llm and runs locally. - pay once, keep it forever. ~ ~ built this at build something wonderful w/ Devesh.

how many of you send things to yourself on whatsapp and never find them again? i tried fixing it. built a mac app that lives in your notch. - drag and drop any media. - ask it anything later. - uses local llm and runs locally. - pay once, keep it forever. ~ ~ built this at build something wonderful w/ Devesh.

Akhil

25,846 views • 28 days ago

we are working on OpenControl - a library that uses AI to finally solve the headache of internal tooling idea is simple - define a set of tools in code that can reference functions from your codebase or anywhere else these tools get deployed to your infra (runs anywhere, lambda, worker, containers, etc) and exposes a protocol that allows clients like cursor, claude-code, etc to use them they can orchestrate tasks like "can you reset the password of anyone who hasn't logged in a year" demo below and details on how it works in the thread

we are working on OpenControl - a library that uses AI to finally solve the headache of internal tooling idea is simple - define a set of tools in code that can reference functions from your codebase or anywhere else these tools get deployed to your infra (runs anywhere, lambda, worker, containers, etc) and exposes a protocol that allows clients like cursor, claude-code, etc to use them they can orchestrate tasks like "can you reset the password of anyone who hasn't logged in a year" demo below and details on how it works in the thread

dax

109,046 views • 1 year ago

APPLE BUILT A $10,000 COMPUTER THAT RUNS WHAT USED TO REQUIRE AN ENTIRE SERVER ROOM 06:40 he shows the number that breaks everything - one Mac Studio is more powerful than four Dell or AMD machines combined and uses half the energy DeepSeek R1 doesn't run on even two competitors - you need four - one Mac Studio does it alone runs quieter than a fan in the room and uses less energy than a kettle in the kitchen four Mac Studios on a desk connected through a regular cable run a 1 trillion parameter model - what used to require an entire data center anyone can put this at home and Jeff Geerling proved it - run the same models that big companies use without cloud, without subscriptions, without data leaving the room

APPLE BUILT A $10,000 COMPUTER THAT RUNS WHAT USED TO REQUIRE AN ENTIRE SERVER ROOM 06:40 he shows the number that breaks everything - one Mac Studio is more powerful than four Dell or AMD machines combined and uses half the energy DeepSeek R1 doesn't run on even two competitors - you need four - one Mac Studio does it alone runs quieter than a fan in the room and uses less energy than a kettle in the kitchen four Mac Studios on a desk connected through a regular cable run a 1 trillion parameter model - what used to require an entire data center anyone can put this at home and Jeff Geerling proved it - run the same models that big companies use without cloud, without subscriptions, without data leaving the room

Noisy

26,153 views • 1 month ago

NEW: Axios w/ a first look at RNC’s video response to Biden It is also the committee’s first video that is 100% generated by AI software meant to create images that look and feel real (in this case an imagined dystopia look at the future of Biden wins)

NEW: Axios w/ a first look at RNC’s video response to Biden It is also the committee’s first video that is 100% generated by AI software meant to create images that look and feel real (in this case an imagined dystopia look at the future of Biden wins)

Alex Thompson

4,269,064 views • 3 years ago

MEET BLUE: NVIDIA’S REAL-LIFE WALL-E Blue is an AI robot that moves, reacts, and talks like something straight out of a Pixar film. Blue runs on 2 NVIDIA computers and uses a physics engine for ultra-realistic tactile feedback, rigid and soft body simulation, all in real time. This isn’t a concept, this is how NVIDIA plans to train robots for the real world. And yes, it casually stole the show. Source: entrepreneurialogy

MEET BLUE: NVIDIA’S REAL-LIFE WALL-E Blue is an AI robot that moves, reacts, and talks like something straight out of a Pixar film. Blue runs on 2 NVIDIA computers and uses a physics engine for ultra-realistic tactile feedback, rigid and soft body simulation, all in real time. This isn’t a concept, this is how NVIDIA plans to train robots for the real world. And yes, it casually stole the show. Source: entrepreneurialogy

Mario Nawfal

38,388 views • 1 year ago

GOOGLE'S NEW AI FITS IN YOUR POCKET AND THINKS LIKE A NESTING DOLL Google just dropped Gemma 3n, an AI that runs right on your phone with no Wi-Fi needed. It reads text, watches video, hears audio, and even talks back all at once. Powered by “MatFormer,” it’s basically an AI inside an AI inside an AI. Oh and it runs at 60 frames per second while using less memory than your favorite mobile game. Source: rowancheung

GOOGLE'S NEW AI FITS IN YOUR POCKET AND THINKS LIKE A NESTING DOLL Google just dropped Gemma 3n, an AI that runs right on your phone with no Wi-Fi needed. It reads text, watches video, hears audio, and even talks back all at once. Powered by “MatFormer,” it’s basically an AI inside an AI inside an AI. Oh and it runs at 60 frames per second while using less memory than your favorite mobile game. Source: rowancheung

Mario Nawfal

179,423 views • 1 year ago

Following ☄︎ work, I'm experimenting with some react three fiber with #threejs webGPU. We no longer need an EffectComposer to use post-processing effects in #r3f. Now, everything is done by composing TSL nodes. I created a small component that helps in rendering post-processing effects. All you need to do is import the effects you want and add them to the final composition. I will share the link to the demo and source code on Monday 😍. Another thing I want to highlight is how amazing the Screen Space Reflection effect is. It generates reflections not only on the ground, as we had before in r3f in webgl. It is able to calculate the reflections between the meshes. Look how beautiful the reflection of the lightsaber on Darth Vader's mask is. This is not something that is easy to calculate and this is exactly where WebGPU in threejs shines! Congratulations to everyone on the threejs team. I am increasingly impressed with WebGPU and TSL.

Following ☄︎ work, I'm experimenting with some react three fiber with #threejs webGPU. We no longer need an EffectComposer to use post-processing effects in #r3f. Now, everything is done by composing TSL nodes. I created a small component that helps in rendering post-processing effects. All you need to do is import the effects you want and add them to the final composition. I will share the link to the demo and source code on Monday 😍. Another thing I want to highlight is how amazing the Screen Space Reflection effect is. It generates reflections not only on the ground, as we had before in r3f in webgl. It is able to calculate the reflections between the meshes. Look how beautiful the reflection of the lightsaber on Darth Vader's mask is. This is not something that is easy to calculate and this is exactly where WebGPU in threejs shines! Congratulations to everyone on the threejs team. I am increasingly impressed with WebGPU and TSL.

Anderson Mancini

26,339 views • 1 year ago

HAO WANG BUILT AN AI SERVER FARM OUT OF 100 MAC MINIS. ONE $599 BOX KILLS A $200/MONTH CLAUDE CODE BILL FOR $3 IN ELECTRICITY a developer posted his $170 claude code bill from 10 days on reddit. someone replied "i bought a mac mini m4. haven't paid anthropic since" the m4 chip has 120 gb/s memory bandwidth and unified memory. a $599 mac mini runs ai faster than a $1,500 windows pc with a discrete gpu ollama now supports the anthropic messages api. claude code connects to your local mac mini with one environment variable. zero api costs, same interface a heavy developer pays $459 a month across claude code, chatgpt pro, gemini, cursor and copilot. the mac mini pays off in 3 months and runs on $3 after that uber rolled out claude code to 5,000 engineers and burned $3.4 billion in 4 months. the people who own the hardware in 2026 will look very far ahead in 2028 bookmark this and read the article below

HAO WANG BUILT AN AI SERVER FARM OUT OF 100 MAC MINIS. ONE $599 BOX KILLS A $200/MONTH CLAUDE CODE BILL FOR $3 IN ELECTRICITY a developer posted his $170 claude code bill from 10 days on reddit. someone replied "i bought a mac mini m4. haven't paid anthropic since" the m4 chip has 120 gb/s memory bandwidth and unified memory. a $599 mac mini runs ai faster than a $1,500 windows pc with a discrete gpu ollama now supports the anthropic messages api. claude code connects to your local mac mini with one environment variable. zero api costs, same interface a heavy developer pays $459 a month across claude code, chatgpt pro, gemini, cursor and copilot. the mac mini pays off in 3 months and runs on $3 after that uber rolled out claude code to 5,000 engineers and burned $3.4 billion in 4 months. the people who own the hardware in 2026 will look very far ahead in 2028 bookmark this and read the article below

starmex

50,679 views • 1 month ago

I love prototyping. It’s the most important thing that sets great software above the merely good. And to that end, here’s a project a couple of us have been hacking away on, and the team is ready to add people in mass to the TestFlight. It’s an open source version of a visual programming environment, similar to Origami or Quartz Composer. In addition to being able to build anything those two can, it also supports AR, CoreML, and native 3D support. As well as being built in SwiftUI, so it runs on basically any Apple device and is easy to hack on and improve.

I love prototyping. It’s the most important thing that sets great software above the merely good. And to that end, here’s a project a couple of us have been hacking away on, and the team is ready to add people in mass to the TestFlight. It’s an open source version of a visual programming environment, similar to Origami or Quartz Composer. In addition to being able to build anything those two can, it also supports AR, CoreML, and native 3D support. As well as being built in SwiftUI, so it runs on basically any Apple device and is easy to hack on and improve.

Adam Menges

101,412 views • 1 year ago