Shimon Vainer

We initially cared about local LLMs, but KV caches appear in more than text. So we also investigated OCTOPUS for autoregressive video and audio transformers. Joined work with @VikramVoleti, Simon Donné, @esx2ve

@Eran_Efrat היי ערן, היכן אפשר לתרום?

תודה. (רשימה חלקית מאד בזמן קצר) מעריך את האמון. לא באמת היה לי כח להוביל עוד סשן כזה אבל אם באמת נצליח לעזור, היה שווה הכל. מבטיחים לעשות מעל ומעבר. וכמובן תודה גדולה לכל מי שמעורב- איזה כיף זה להביא מלא מוחות כאלה לאותו שולחן ועוד בחינם 😂 ניפגש בשטח בקרוב...

Current hobby: vibecoding tiny tools that eliminate annoying tasks. I open-sourced them in case they’re useful to others.

The first lab to open-source a solid pixel-space video diffusion model will be cited a bazillion times by everyone working on inverse problems esp. related to 3D/4D. With JiT/Simple(r) Diffusion the tech is mostly there, someone with more GPUs than me should make it happen (plz)

We are hiring! ☺️

We need something like full of interactive visualization and hover documentation, like distill pub like visualization, for everything. Math, physics, chemistry, etc etc LLMs can totally make epic interactive diagrams and visualizations at massive scale, we need entire textbooks like this.

Introducing our new work, "Foley Control: Aligning a Frozen Latent Text-to-Audio Model to Video" It's a control model for Stable Audio to generate aligned audio from an input video. Project Page: stability-ai.github.io/foleycontrol.g… Paper: tinyurl.com/4929p8bh 🧵 @StabilityAI

@mariyaivasileva Heya - we're hiring for some really cool opportunities at #StabilityAI. DM me if sounds relevant

I was impacted by the Meta PAR/FAIR layoffs today. I’m a research scientist passionate about computer vision, generative models, multimodal understanding, and AI safety. Over the past decade, I’ve contributed to advancing a broad range of applied machine learning problems — from multimodal generative models and visual recommender systems, to 2D-to-3D human modeling, large-scale data generation for foundation model training, mechanistic interpretability, and rigorous evaluations for trust & alignment of large-scale AI systems. I’m actively seeking new opportunities — please reach out if you have any openings!

Read more about the EA/SAI announcement here: stability.ai/news/stability…

Working on 3D, video, or image generation? We’re hiring at @StabilityAI — building next-gen creative AI tools with EA to reimagine world-building. If you’ve been affected by recent layoffs (Meta/FAIR or elsewhere), DM me. Around #ICCV2025? Let’s connect.

@torchcompiled It's from omages.github.io. Also check out concurrent work unity-research.github.io/Geometry-Image…

3D polygon diffusion is one of the coolest things I've ever seen

New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵

this guy's channel is so small with only a couple K views here and there if you're interested in GPU programming and still a beginner, he's worth a look (Simon Oz on yt)

Congratulations to @gintszilbalodis and the entire Flow film crew for the Academy Award win! Flow is the manifestation of Blender’s mission, where a small independent team is able to create a story that moves audiences worldwide. Thank you for the shout out! 🧡 #b3d

@jon_barron Try generating a simple checkerboard - it's surprisingly impossible. Many questions...

Every image generator I've tried is surprisingly bad at visualizing a 2-layer neural network. What does it mean?

@tinotibaldo gafferongames.com/categories/gam… is a great place to start losing your sanity

@tinotibaldo Short story: back in the day, as a self taught game-dev, I've tried to code Uniball from scratch. It's basically 2D rocket league. It took me a few months to nail the game look and feel. And two years to fail miserably at the multiplayer physics. The rabbithole goes insanely deep

Part two We're Improving the game's physics Let's build this #RocketLeague game together; what would make the game EXTRA fun? #threejs #browsergame

The post below (learnable lambda has to do with skip connections) reminded me of some ResNet and Transformer architecture-related lore that I thought would be fun to write up! Blogpost: lb.eyer.be/s/residuals.ht…

I’m currently exploring new job opportunities🧑‍🔬 My work revolves around text-to-3D with Geometry Images, generative UV mapping, multi-view models for texturing, and other genAI applications in graphics. I’d love to discuss how I can contribute to your research efforts!