The quality, cost, and control you can achieve for upscaling + fixing plastic AI skin with open source models still amazes me... Models used: → Z-image-turbo for image gen (~3s) → SDXL + Lora for skin texture (~15s) → SeedVR2 for upscaling (~40s)

"Local Video Editing" is officially dead 🤯 VideoSOS can run 100+ AI models directly in your browser without cloud processing. It uses a massive stack including Veo 3.1, FLUX, Gemini 2.5 Flash, and Imagen 4, running locally on your hardware. 100% Open Source.

🚨 ✨ New 3D pose estimation method from M- Lab of Adaptive Intelligence @EPFL! #FMPose3D allows for monocular (i.e. single camera) 2D ➡️ 3D 🔥 Led by Ti & w/xiaohang #FMPose3D is SOTA on human & animal 3D benchmarks, & will be integrated into DeepLabCut 🦄 ⬇️👀 📝 arxiv.org/abs/2602.05755

x.com/i/article/2019…

Flux 2 Klein 4B param, dropped at 2 steps with much higher FPS! I also added a couple of LORA's, I'll mess around with. Such a good diffusion model!

Pretty cool unified AI VFX production pipeline. - Cinema Prompt Engineering with ~110 film and animation presets; - an infinite Storyboard Canvas + distributed rendering across multiple ComfyUI nodes; - 13+ LLM providers. github.com/NickPittas/Dir…

i built an app that converts any space into a digital clone in minutes as the founder of Teleport - the only iphone app that can capture high-quality 360° panoramas - i already had the perfect input when World Labs released their 3d reconstruction api 📍 first test - a

This is crazy! With LTX 2 in ComfyUI i was able to keep my original camera movement + it did a full lip-sync on the singing haha. All running locally. I made the original video back in 2021 when Metahumans first came out and we were working in Unreal Engine 4. This music video

seedance 2.0 will be the official killer of Sora 2 + VEO workflow

ComfyUI Action Director. Interactive 3D viewport for ControlNet; - loads FBX/GLB; batch-renders OpenPose, Depth, Canny, Normal; - manual near/far depth control and infinite UI scaling. github.com/yedp123/ComfyU…

NVIDIA just dropped PersonaPlex-7B 🤯 A full-duplex voice model that listens and talks at the same time. No pauses. No turn-taking. Real conversation. 100% open source. Free. Voice AI just leveled up. huggingface.co/nvidia/persona…

In some of my tests, Sparc3D produced the results from a single photo that are, frankly, quite crazy.

Pooya Moradi M. presented a demo made with Meshcapade, an AI-powered, markerless motion capture software. It was recently acquired by Epic Games: 80.lv/articles/ai-mo…

While AI-based tools like Sparc3D or Hunyuan 3D can generate genuinely impressive results from a single photo, they often rely on credits, paywalls, and non-so-transparent processing of data. In this tutorial we strip it down to basics: a smartphone, 100% free software,

FlowRVS - segmentation as a continuous deformation, mapping video latents directly to masks via an ODE. Built on Wan’s T2V. - complex semantic understanding with temporal consistency. - no flickering github.com/xmz111/FlowRVS

We are impressed by this new Z-Image-Turbo LoRA from the community! By utilizing Flow-DPO, it effectively eliminates "washed-out" artifacts and brings cinematic, physically accurate lighting to our ultra-fast distilled model. 🔹 The Magic: Stunning photorealism in just 8

A 1-billion-parameter AI real-time motion model that connects to a 1080P camera or uploaded videos, supports UE/Unity/Blender, and requires 8GB of VRAM for real-time processing.

Many tried, most failed. But this is the first skin enhancer I've used that actually makes characters look real. Meet Vellum from OpenArt It's now a staple in my workflow. I won't start a project without it:

#NVIDIA just released a whole ecosystem for human(oid) motion and robot learning from human data. 🚀🦾 Data, as we all know, is the key to scaling AI models. To accelerate the field of Embodied AI, we have open-sourced a full stack of models and tools to capture, generate,

There are many human body models, SMPL, MHR, Anny… which one should you use? Answer: all of them. At GTC 2026, we release SOMA, a unified body layer that takes any model's shape and pose, and gives you one canonical mesh and rig.