Doug Traill

#VisionPro Vs #Quest interaction paradigms. This is what it took 👉 https://lnkd.in/dxif3aEF In this interview with Daniel Sproll we discussed their journey of porting Puzzling Places to #VisionPro and the value #mixedreality added to this relaxing puzzling experience.

123

8 Comments

Raspberry Pi 5 is getting a first party AI accelerator add on, Hailo 8L accelerator which can do 13 trillion operations at int8 , enough to run quantized LLM's fast, processing videos in real time What would you build if you get access to it?

33

2 Comments

👀 Customize #NVIDIACosmos world foundation models and video foundation models with new #NVIDIANeMo framework capabilities✨ 1️⃣High-throughput video data curation 2️⃣Efficient #multimodal data loading 3️⃣Scalable model training 4️⃣Accelerated in-framework inference Read our blog to learn more ➡️ bit.ly/3W3GHsj

13

New real-time technologies announced at #CES2025 with the launch of #NVIDIA #Blackwell. - Neural Materials - Neural Texture Compression, - Mega Geometry - allowing - Curve primitives for ray-traced hair (LSS) - New BRDFs for real-time pathtraced hair and SSS for skin - part of the upcoming "RTX Character Rendering SDK" so anyone can adopt photoreal hair and skin shading in their real-time environment. https://lnkd.in/gnMBn_C2 #realtime #rendering #3drendering #airendering #digitalhuman #3dtexture #ai #machinelearning #3dart #photorealistic #3davatar #gameart #gamedevelopment #gamedeveloper

36

very nice use of AI

53

Another attempt as converting CGI into photorealism via generative video, note the output is only 480p@24fps it reseeds after 78 frames, I will work on a fix for this today. No adhesion to the mouth poses in the original video, so the audio track has no sync at all. #text2video #script2video

13

4 Comments

Microsoft GASP and Cafca are revolutionizing 3D and AI! 🚀 Here’s why it’s exciting: 🎥 Lifelike Avatars: Human faces and expressions with stunning realism. 💡 Tech Mashup: NERF, Gaussian splats, volumetric capture—next-level innovation. 📊 Data Challenges: Still need high-quality datasets (multi-point sources, phased lighting). 🌍 Beyond Entertainment: This tech is expanding into industries everywhere. The pace of breakthroughs is wild. Where will this take us next? Let’s dive in! #NERF #GaussianSplats #VolumetricCapture #DigitalTwins #AIInnovation #FutureTech

4

Wondering when projection mapping will make a come back so I've started to experiment again with #RF3 web-based projection mapping. Full experience -> https://lnkd.in/eQd-YK9b #reactthreefiber #3d #nike #projectionmapping

10

How cool is websim.ai??? 😅🕶️🎉 By using Anthropics Claude 3 models (choose between Opus, Sonnet and Haiku) websim.ai generates a website for you by just providing a URL and optional parameters. I tried https://lnkd.in/g54nQrTN and it created a website with a searchable database of movies shot on celluloid 🤯 I mean how cool is that! And I like the fact that a lot of the generated websites look a lot like they are straight from the first days of the world wide web 😅 #websimai #ai #websites #llm #anthropic #claude3

8

2 Comments

LTX with STG vs Luma Dream Machine I2V comparison study... Spatiotemporal Skip Guidance for Enhanced Video Diffusion Sampling https://lnkd.in/guvGeFNy LTX video was generated in #comfyui locally on my 3080 / 12gb in less than a minute.

46

15 Comments

Google Cloud and Anthropic made some exciting announcements: 📣 Claude 3 Opus, Anthropic’s most intelligent model, is now generally available on Vertex AI ⚙️ Tool use is available for the Claude 3 model family on Vertex AI, enabling Claude to act as an autonomous agent 📈Provisioned throughput is available for the Claude 3 model family on Vertex AI

6

Distributed simulations in SimBricks are another one of the fun and somewhat counter-intuitive technical pieces in SimBricks: instead of also implementing efficient network communication in each simulator (something I know a little something about), we add separate proxies that translate between local shared-memory message passing and TCP or RDMA. This obviously increases latency but surprisingly for synchronized simulations we have not found cases where this small increase is a problem. But instead, the huge benefit of not paying thousands of cycles for sending network messages on the critical path in each simulator, especially for already slow simulators, means that we do not slow down simulations at all. For more detail, see Marvin's post below!

2

Step-by-step reference app ➡️ Discover how to generate a multimodal AI agent with Meta Llama 3.2 and build a Streamlit app that seamlessly processes text, images, PDFs, and PPTs. Integrates Llama 3.2 models, NIM microservices, Milvus, & LlamaIndex.

12

It’s all about optimizing your #production #pipeline and giving artists full #creative control and extra iteration time! #art #2d #workflow #AI #mattepainting #realtime #games #film #tv #animation #interactive #media #entertainment

22

1 Comment