Another challenging hard surface scan completed! Photogrammetry often struggles... with pixel-perfect alignment on these kinds of objects, but we’ve developed methods to make it work. Still, always room for improvement! #photogrammetry #b3d #3dscanningshow more

We can 3D scan even the hardest objects -... dark, low-contrast, and nearly featureless surfaces are no longer a challenge - all with full PBR material stack. Need this level of precision for your production? Let me know! #photogrammetry #photometryshow more

New scan sample, old wooden container scanned using our... advanced robotic scanner developed at INVOLV3D and inciprocal software for PBR material processing. You can literally smell this wood! #3dscanning #pbr #b3dshow more

#AniMatePro is shapping up to the early access release.... It will likely be broken down in 2 addons. One with the productivity improvements for interacting with the timeline and another one with all the tools to manipulate and interact with the keyframes. Still looking for some animators that are work with Blender or would want to work with Blender but are missing some tools from Maya or Animbot to make the switch. Join discord and request access: #b3dshow more

Model turntable of Tsunami design by Lē Yamamura for... the Wings of fire Japanese release. These designs are a lot of fun to work with, still a bit rough in some places but I'm happy enough with how it turned out. Who should I make next?show more

Tested Claude + Blender today with two real workflows.... Sharing what actually happened. First one was a simple demo. Gave it one sentence to build a cyberpunk room. Geometry, lighting, camera, render. It got there, though not everything in the prompt landed perfectly. Good enough to show the concept, not a stress test. Second one was more interesting. Threw a raw photogrammetry scan at it. Massey Ferguson tractor, 250k faces, grass background everywhere, debris floating all over the place. Asked Claude to clean it up autonomously. It did the work. Removed 106 floating chunks, cut face count by 87%, stripped the grass, flipped the model upright, set up studio lighting and rendered. Technically it worked. But honest answer: the result wasn't clean enough for production. This kind of cleanup still needs the right tools and someone who knows what they're doing. So here's where it actually fits. Not the precise technical stuff. The setup work. The repetitive stuff. The things that take forever if you don't already know Blender. It won't replace your 3D artist. It replaces the boring part before they start.show more

I've been testing out Windows 11's new Haptic Signals... feature on compatible hardware, and I think I love it. I'm someone who yearns for small UX details like this, and I'm so happy to see Microsoft leaning into subtle but quality UX implementations like haptics across the Windows Shell. These new Haptic Signals provide a subtle bump when interacting with certain UI elements and actions, such as snapping app windows or dragging files in File Explorer. The feedback response feels great and intentional, and makes using the OS feel more satisfying. There's still room for improvement, but I'd be lying if I said I wasn't impressed with this first implementation. It's only supported on a handful of devices for now, but that should change over time. I've always loved the haptic feedback layered throughout Google Pixel phones, and now Windows is getting its own version of it. More UX niceties like this, please!show more

Little review on the HUION Kamvas 16 (Gen 3)... by 타블렛 Was quite happy with it as I could make my sketchbook drawings on it without any problem! I normally work on an expensive tablet, over 2000 euros but the arrival of these smaller pen displays with excellent quality open up so much possibilities for starting artists. Without having to compromise much on quality! Thank you for Huion for sending me the tablet! #huionshow more

Boomer dads really built their entire personality around “I... worked hard and figured it out.” Cool. You also: Bought a house on one income at 26. Paid $400 a semester for college. Got a job with benefits straight out of high school. Had a pension waiting at the end. Paid $90 a month for health insurance. Filled up a tank for $18. Raised a family on one salary and still took a vacation every summer. You didn’t just work hard. You worked hard in an economy that was designed to reward hard work. We work hard in an economy designed to extract it. Same effort. Completely different machine. But sure dad. Tell me again about your bootstraps.show more

You can say what you want about the current... Vision Pro hardware, but you won’t change my mind: Personas are the future of telecommunication. They’re resistant to catfishing, always safe for work, and hit a perfect sweet spot: you’re not forced to turn on a camera if you’re uncomfortable, but you still get far more emotion and presence than just voice. With them you always look good, have perfect lighting, and have full control over the background. Over the past year I’ve met over 100 new people with whom I played games, watched movies, worked, and just hung out, always having far more fun than I’ve ever had on a traditional video call.show more

✅ Big version jump. Whats new?! XRP Ledger is... now in version 3.0! with some major introductions. New are the so called 'unsupported' amendments👀: - Lending and Borrowing on XRP, essentially Loans on chain. - DynamicMPTs, similar to Dynamic NFTs. these amendments are introduced, but not completed in development and not available to be voted on yet. in XRP 3.0 there are also a list of normal fix amendments, among them: fix for the TokenEscrow amendment. fix for PriceOracles. fix for MPTs. fix for clawback tokens in AMMs. the changes list is long and i want to thank everyone who contributed to this release. Hard work and effort went in to make the XRP Ledger a robust, reliable and innovative rail for on chain finance ❤️show more

Admin Post: This is a T Puller. It's really... hard to pull a door closed from a wheelchair and these make it much easier. They stick on the door on the hinge side. I just flip it out and use it to pull the door until I can reach the knob. In the past I've used pull ropes but I still had to go through, turn around, wheel back in to get close enough to grab the rope and then back out. Chesnyy could also grab and pull a rope but there needed to be room for her to maneuver and at times she's not at the door with me (especially if I'm taking another dog out) or she might be carrying something. T Pullers make the house a bit easier for me. #Accessibility Video description for inclusivity: and open door that leads out to a deck. There is a small black handle on the inside. A hand reaches out and flips it out and uses it to pull the door almost closed, then grabs the door knob.show more

You can't 3D reconstruct glass from images... ...WRONG! Thanks... for video diffusion, now just about anything is possible! Introducing...Diffusion Knows Transparency (DKT) Transparent and reflective objects usually break robot vision and photogrammetry pipelines because they don't follow the "solid object" rules standard cameras expect. DKT is a new AI model that repurposes the "internal physics engine" found in video generation models to solve this problem. Researchers took a massive video diffusion model (WAN) and fine-tuned it using a custom-built synthetic dataset to turn it into a high-precision depth sensor. To train the AI, they built the first massive synthetic video library of transparent objects, 1.32 million frames of perfectly labeled glass and metal objects in motion. Without ever seeing a "real" labeled video of glass during training, the model (DKT) outperformed all previous specialized systems on real-world benchmarks (ClearPose, DREDS). They created a "lightweight" 1.3B parameter version that runs fast enough (0.17s per frame) to be used on actual robot hardware. Two reasons I find this project important: 1. It further proves that synthetic data will be essential for training the next generation vision models. 2. In real-world robotic tests, using DKT's depth maps nearly doubled the success rate of robot arms trying to pick up objects on tricky reflective or translucent surfaces. At home robots will need to interact with these types of objects on a daily basis. Check out the project page here: Code is LIVE! #Computervision #Robotics #AIshow more

Video diffusion models have strong implicit representations of 3D... shape, material, and lighting, but controlling them with language is cumbersome, and control is critical for artists and animators. GenLit connects these implicit representations with a continuous 5D control signal describing the direction and intensity of a point light source. This enables single-image near-field relighting of an image using a video diffusion model. We use a ControlNet-like approach and show that, with a small amount of synthetic data, GenLit generalizes to complex real-world images. Given a single image and the 5D lighting signal, GenLit creates a video of a moving light source that is inside the scene. It moves around and behind scene objects, producing effects such as shading, cast shadows, secularities, and interreflections with a realism that is hard to obtain with traditional inverse rendering methods. GenLit shows that it is possible to get continuous control over implicit physical processes within a video model. I think this is just the beginning and promises to make such models much more practical for creators. Shrisha Bharadwaj will present today at SIGGRAPH Asia Room: S423/S424, Level 4 @ 13:50 on 15 of Dec.show more

Jimmy Kimmel admits it’s “hard not to agree” with... Spencer Pratt on LA’s problems… but still says don’t vote for him. In his Wednesday night monologue on Jimmy Kimmel Live!, Kimmel acknowledged “this city is a mess” and that Pratt is tapping into real frustrations over crime, homelessness, and failed leadership. Yet he slammed the Hills star as a “narcissist looking for attention” with zero experience to actually fix it. Kimmel compared it to Trump’s rise — agreeing on the problems but rejecting the messenger. Does this make sense? Can you agree with someone’s diagnosis of the issues but completely reject them as the solution? (Video: AI)show more

Tomorrow is a big day for me. I’m going... to do an activity in my daughter’s class (she’s 5), and I’ll be teaching them what AI is. She’s excited about it, but honestly, I think I’m even more excited. I’ll tell them a little bit of history, we’ll play some games… I’ll try to make it as fun as possible, but always with a focus on ethical and responsible use. Of course, it won’t be a tutorial on how to use any tools, lol. They’re still very young, and in my opinion, it’ll be years before they actually need to start using these tools. But I think it’s important for them to start getting some awareness so they can be prepared for the future that’s coming their way.show more

My first red carpet moment as a designer! (Or... in this case, purple carpet) 💜💜💜 #MeganTheeStallion wearing a custom gown designed by me, for the premiere of her documentary #INHERWORDS All of these moments are really special, because it’s always been a dream of mine to bring my designs to life & see them be worn at parties, on red carpets, & for on stage performances & each year, it’s all becoming more & more of a reality after over a decade of hard work, sharing my sketches online & establishing myself. Thank you to all of the talents I’ve collaborated with that have helped make these dreams a reality. Here’s to many many more 🫶🏾show more

It’s difficult to put a finger on exactly when... a nation falls. It is hard to know the very moment, but you can feel it in your bones when the president gleefully admits that his election was purchased with money from another country. He does this knowing that moves like this will usher in untold Republican loss in 2026 and 2028. And yet, he still does it. We’ve moved from nuance to bragging about the theft of American tax dollars and delivery over to another people in another place. It is not done to uphold US interests, but is a tithe to unbelievers who hold inexplicable power and authority over our government. And since it is done despite its devastating impacts on the upcoming elections, one must imagine it’s done under some sort of fear or duress. What other explanation could there be? It is one thing to know what goes on behind the scenes, and quite another for the president to admit it. This admission signifies the full transition from a strong lobby to absolute power, with the full sanction of the president and the full weight of government to enforce speech codes and stand up 100 task forces. This is to Republicans as BLM was to Democrats; this is them on one knee ushering in a time when you too will be expected to kneel.show more

What if you kept asking an LLM to "make... it better"? In some recent work at FAIR, we investigate how we can efficiently use RL to fine-tune LLMs to iteratively self-improve on their previous solutions at inference-time. Training for iterated self-improvement can be costly. The naive approach to training for K self-improvement steps leads to K times the number of rollout steps per episode. We introduce Exploratory Iteration (ExIt), an RL-based automatic curriculum method that bootstraps diverse training distributions of self-improvement tasks by upcycling the LLM's own responses at previous turns as the starting points for both self-improvement and *self-divergence.* In order to decide what task to train on next, the curriculum prioritizes sampling of partial turn histories that led to higher return variance in its GRPO group (a learnability score that comes for free). This automatic curriculum over the bootstrapped task space teaches the model how to perform iterated self-improvement while only ever training the model on single-step self-improvement tasks. We look at ExIt's impact in both single-turn (contest math problems) and multi-turn (BFCLv3 multi-turn tasks), as well as MLE-bench, where the LLM is run in a search scaffold to produce solutions to real Kaggle competitions. Across these eval settings, we find ExIt produces models with greater capacity for inference-time self-improvement compared to GRPO. Notably, ExIt models can self-improve on test tasks for many more steps than the typical solution depth encountered during training, including a 22% improvement in MLE-bench performance compared to GRPO.show more