LightVAE + ComfyUI node: High-performance video VAE; runs 2–3x... faster using 50% less memory; LightTAE offers a 10+x speedup on just ~0.4GB VRAMshow more

Microsoft just a 1-bit LLM with 2B parameters that... can run on CPUs like Apple M2. BitNet b1.58 2B4T outperforms fp LLaMA 3.2 1B while using only 0.4GB memory versus 2GB and processes tokens 40% faster. 100% opensource.show more

Introducing Steerable Motion, a ComfyUI custom node for steering... videos using batches of images. You provide images & instructions on how you want to travel between them and it creates smooth video to fill in the gaps. See attached + more here:show more

Introducing #Ray2 Flash—3x faster, 3x cheaper new model. Flash... brings Ray’s frontier production-ready Text-to-Video, Image-to-Video, audio, and control capabilities with high quality and speed to all subscribers—so you can create more, faster, and without limits. Available now.show more

Our video editing model Lucy-Edit just got way better.... 2× faster, half the price (and open source). More control, faster performance, and seamless results.show more

🎉 The built-in node help menu is now available... to everyone. 🔁 Just update your ComfyUI to the latest version and make sure the frontend is at least v1.22.2show more

A little side-by-side comparison, one using the Wan model... in ComfyUI, the other using Luma Ray 2.show more

✨ Free ComfyUI Workflow for Clarity Upscaler! Finally, we... have a fully ComfyUI workflow that runs locally on your own GPU and uses the same technology as Clarity To use it, just download the free-workflow.json file from the ComfyUI-ClarityAI repo 😀 Ad: If you're looking for a more convenient way to upscale, try my app: 👩‍💻 App: 🐰 ComfyUI:show more

Free Blender AI Motion Capture Plugin: Runs a 1-billion-parameter... model on a GPU with 8GB VRAM, supporting real-time capture and video uploads — no cost required.show more

We just launched a 3x performance improvement for our... order book interface. Pre-batched messages (12x fewer msgs to parse), less GC pressure, lower CPU, leading to a much snappier ActiveTrader experience (old on left, new on right). More improvements incoming.show more

I ported DepthCrafter to ComfyUI! 🔥 Now you can... generate super stable depthmap videos from any input video. The VRAM requirement is pretty high (>16GB) if you want to render long videos in high res (768p and up) It pairs well with Depthflow! Repo link in comments below👇show more

MiniMax just dropped M2.5 and it's on par with... Opus 4.6 while being 20x cheaper and 3x faster???show more

See how Half-Life 2 looks like running on just... 8 MB of VRAM, which is 8 times less than the GPUs of its time, brought to you by Budget Builds. It's 60 FPS:show more

The first natively trained 1-bit model: BitNet 2B. Trained... on 4 trillion tokens. that can run on CPUs like Apple M2 Native 1.58-bit weights and 8bit activations W158A8 Outperforms LLaMA &close to Qwen 2.5 1.5B in while using only 0.4GB memory versus 2GB and processes tokens 40%show more

MTP speedup Qwen by 2.5x in Atomic Chat Dense... vs MoE models on 2x RTX 5090 Qwen3.6 27B: 51 → 117 tps +137% Qwen3.6 35B-A3B: 218 → 267 tps +25% MTP drafts several tokens ahead and verifies them in one pass. The speedup depends on memory moved per pass. Dense 27B reads all 27B params per token, MoE 35B-A3B only reads 3B active. Dense had way more to save by batching. The baseline tps also differ (218 vs 51) for the same reason from the other side. Token generation is memory-bandwidth bound, and MoE moves ~8x less memory per token, so its baseline is already 4x ahead. ~80% draft acceptance. Zero accuracy loss. ~1 GB extra VRAM. Open-source code and local AI app – in the comments 👇show more

OK the Nexus CLI has some fixes deployed: -... Given a Node ID, it runs end-to-end Node ID generation on the Nexus OS is still getting fixed.show more

Griffain is now 3x faster

Did a quick and dirty implementation of a spatial... hash structure to speedup RTAO, ray results are stored in cells indexed by pos/normal/cell size and after storing a few rays occlusion can be queried from the cell instead of raytracing it. 3x faster RTAO with no denoising.show more

Bytedance just dropped realtime voice translation 3x faster than... before, with only a ~3s lag! Seed LiveInterp 2 is a full duplex speech-to-speech model with >70% correctness. When this makes it to video calls, it'll open up previously impossible connections.show more