pixel surgeon ART | ML | VFX
latentpark
Joined June 2009
- Tweets1.4K
- Following958
- Followers1.9K
- Likes8.7K
Pixel-space 🚀
Thanks for sharing! We open-sourced code and models.
🤗Huggingface Paper: huggingface.co/papers/2609.3…
💻 Code: github.com/nv-tlabs/PixelUMM
🤗 Model: huggingface.co/nvidia/PixelU…
This Anime is full of beautiful images …. made me want to draw again
👽映画『我々は宇宙人』公式👽
＼本編映像公開🏊／
プールの底へ潜り、
ダイブボールを掴む暁太郎。
水の中で揺れる、夏の記憶です…
映画『我々は宇宙人』全国公開中
@KadowakiKouhei
a neat path tracer in a browser #WebGPU
The next version of three-gpu-pathtracer with WebGPU support is published! Drop path tracing into any project for product viz, architectural renders, or generative art.
Check it out here & share what you're making ✨
github.com/gkjohnson/three-g…
#threejs #webgpu
Glad the future is not token-per-second
#TokenPerSecondFatigue
This is NOT Jev.
Open source. Runs on your laptop. Decides in ~27 ms, about 200× faster than waiting on a hosted LLM.
Here it is playing Tetris by itself 👇
brainfunctioncollapse.com/la…
坂本龍一 + 高谷史郎 Async immersion is being shown in Tokyo and when I saw it in M+ Hong Kong (sharing some images @ M+) I have an impression that this work has a challenging soundscape to reproduce and every venue would impose its limits, wondering what the Tokyo one is like.
I saw @benhouston3d MaterialX working in browsers, so there might be a @threejs pathtracer app with MatX support ?
Getting depth of field working with none other than the @pixarsrenderman teapot! All rendered with three-gpu-pathtracer in a browser 🫖✨
Back in the old days, people follow RenderMan Spec. to build renderman renderers, and the catch was always the "Interpretation" of the spec. Nowadays, different DCCs write Alembic in their own ways too. Love to see this sort of initiative, an Open Interpretation.
Typhoon is now publicly available on GitHub🌀 It builds on Linux, macOS, and Windows.
github.com/NVIDIA-Omniverse/…
Presentation slides:
hosted-files.sched.co/osd202…
a clip generated 2 years ago by Runway Gen3. What if the model vendor chose to focus on some specific visuals and generation instead of going after "World" ....
unseen.16
It’s quite often the vidgen model, (here #gen3alpha from @runwayml) is unable to allocate proper duration to each transformation you want especially for the difficult ones. It sometimes just focus on One only.
But Nice Surprise 😍
#vfx #ML #neural #generative
Have been refining a deeper UNet for a higher resolution training while maintaining a small footprint (50MB). Wanna turn this art-engineering work into an iOS app to share how the sky above Victoria Harbour looks like in Summer in this humid city.
⚡️
Fast H3 implementation for Metal. Enjoy, modify, and so forth: github.com/antirez/h3.c Contains code from @liuliu which is welcomed in taking back whatever parts he likes for @drawthingsapp in case there are H3 plans there.
[ Teapot Bounding Box ]
Unable to grab the new one but I have recently found a near optimal Bounding Box for @pixarsrenderman Walking Teapot
Pixel-Space 👏
New paper: AsymFlow🔥
JiT x0-prediction is not enough for pixel generation. Better keep velocity in a low-rank subspace:
- 1.57 FID on ImageNet (best pixel flow model)
- Finetunes FLUX.2 klein into pixel space, beats the original on HPSv3/DPG/GenEval (#1 overall on HPSv3)
1/7
Mike Wong retweeted
Preparing a major PR to improve @ThreeJS MaterialX support significantly. In order to perfect it, I've written an open source @MaterialXcg Fidelity Suite inspired by the @thekhronosgroup / @glTF3D fidelity tool:
material-fidelity.ben3d.ca
Captured a vid of sampling a tiny diffusion model with the neural engine (A17 Pro) on my phone. Not very fast but pretty acceptable given everything happened on the iPhone locally
