IA Lab

AI experiments, AI+Art crossovers, and curated external articles about artificial intelligence and creative technology.

0 24 0 Comment on LichtFeld Studio 0.5.4 — The Free, Open-Source Gaussian Splat Workstation That an AI Agent Can Drive Through MCP
LichtFeld Studio 0.5.4 — The Free, Open-Source Gaussian Splat Workstation That an AI Agent Can Drive Through MCP LichtFeld Studio 0.5.4 — The Free, Open-Source Gaussian Splat Workstation That an AI Agent Can Drive Through MCP LichtFeld Studio 0.5.4 — The Free, Open-Source Gaussian Splat Workstation That an AI Agent Can Drive Through MCP LichtFeld Studio 0.5.4 — The Free, Open-Source Gaussian Splat Workstation That an AI Agent Can Drive Through MCP

LichtFeld Studio 0.5.4 — The Free, Open-Source Gaussian Splat Workstation That an AI Agent Can Drive Through MCP

An all-in-one, GPL-3.0 desktop app for 3D Gaussian Splatting: train, edit and export splats, with an MCP interface so an agent can run the pipeline. Version 0.5.4 just shipped. NVIDIA-only.

Issanou 15 October 2026
0 52 0 Comment on AMD Just Bought Fei-Fei Li’s World Labs for $8.2 Billion — The Spatial-AI Lab Behind Marble and Atlas Is Now a Chip Company’s Secret Weapon
AMD Just Bought Fei-Fei Li’s World Labs for $8.2 Billion — The Spatial-AI Lab Behind Marble and Atlas Is Now a Chip Company’s Secret Weapon AMD Just Bought Fei-Fei Li’s World Labs for $8.2 Billion — The Spatial-AI Lab Behind Marble and Atlas Is Now a Chip Company’s Secret Weapon AMD Just Bought Fei-Fei Li’s World Labs for $8.2 Billion — The Spatial-AI Lab Behind Marble and Atlas Is Now a Chip Company’s Secret Weapon AMD Just Bought Fei-Fei Li’s World Labs for $8.2 Billion — The Spatial-AI Lab Behind Marble and Atlas Is Now a Chip Company’s Secret Weapon

AMD Just Bought Fei-Fei Li’s World Labs for $8.2 Billion — The Spatial-AI Lab Behind Marble and Atlas Is Now a Chip Company’s Secret Weapon

We have covered Fei-Fei Li’s World Labs more than any other outfit in this Lab — Marble,…

Issanou 7 October 2026
0 111 1 Comment on AVTR-1 — The First Fully Open Real-Time AI Avatar That Renders Every Pixel of the Face, and Listens Back While You Talk
AVTR-1 — The First Fully Open Real-Time AI Avatar That Renders Every Pixel of the Face, and Listens Back While You Talk AVTR-1 — The First Fully Open Real-Time AI Avatar That Renders Every Pixel of the Face, and Listens Back While You Talk AVTR-1 — The First Fully Open Real-Time AI Avatar That Renders Every Pixel of the Face, and Listens Back While You Talk AVTR-1 — The First Fully Open Real-Time AI Avatar That Renders Every Pixel of the Face, and Listens Back While You Talk

AVTR-1 — The First Fully Open Real-Time AI Avatar That Renders Every Pixel of the Face, and Listens Back While You Talk

Here is a talking head that invents every single pixel of its own face, 25 times a…

Issanou 7 October 2026
0 38 0 Comment on Bake3D — The Image-to-Rigged-Character App That Finally Stops Assuming Everything Has Two Arms and Two Legs
Bake3D — The Image-to-Rigged-Character App That Finally Stops Assuming Everything Has Two Arms and Two Legs Bake3D — The Image-to-Rigged-Character App That Finally Stops Assuming Everything Has Two Arms and Two Legs Bake3D — The Image-to-Rigged-Character App That Finally Stops Assuming Everything Has Two Arms and Two Legs Bake3D — The Image-to-Rigged-Character App That Finally Stops Assuming Everything Has Two Arms and Two Legs

Bake3D — The Image-to-Rigged-Character App That Finally Stops Assuming Everything Has Two Arms and Two Legs

Hook The dead-statue problem in AI 3D is old news: models spit out a beautiful mesh you…

Issanou 30 September 2026
0 42 1 Comment on Lyra 2.0 — NVIDIA’s World Model Stopped Forgetting: Walk a Whole Street From One Photo, and Get Splats a Robot Can Drive Through
Lyra 2.0 — NVIDIA’s World Model Stopped Forgetting: Walk a Whole Street From One Photo, and Get Splats a Robot Can Drive Through Lyra 2.0 — NVIDIA’s World Model Stopped Forgetting: Walk a Whole Street From One Photo, and Get Splats a Robot Can Drive Through Lyra 2.0 — NVIDIA’s World Model Stopped Forgetting: Walk a Whole Street From One Photo, and Get Splats a Robot Can Drive Through Lyra 2.0 — NVIDIA’s World Model Stopped Forgetting: Walk a Whole Street From One Photo, and Get Splats a Robot Can Drive Through

Lyra 2.0 — NVIDIA’s World Model Stopped Forgetting: Walk a Whole Street From One Photo, and Get Splats a Robot Can Drive Through

Give NVIDIA’s new model a single photo and a whisper of a prompt, and it doesn’t just…

Issanou 30 September 2026
0 40 0 Comment on Blender 5.3 Just Made Gaussian Splats Native — Import, Render and Geometry-Node Them Across Workbench, EEVEE and Cycles, No Add-On Required
Blender 5.3 Just Made Gaussian Splats Native — Import, Render and Geometry-Node Them Across Workbench, EEVEE and Cycles, No Add-On Required Blender 5.3 Just Made Gaussian Splats Native — Import, Render and Geometry-Node Them Across Workbench, EEVEE and Cycles, No Add-On Required Blender 5.3 Just Made Gaussian Splats Native — Import, Render and Geometry-Node Them Across Workbench, EEVEE and Cycles, No Add-On Required Blender 5.3 Just Made Gaussian Splats Native — Import, Render and Geometry-Node Them Across Workbench, EEVEE and Cycles, No Add-On Required

Blender 5.3 Just Made Gaussian Splats Native — Import, Render and Geometry-Node Them Across Workbench, EEVEE and Cycles, No Add-On Required

For two years, getting a Gaussian splat into Blender meant an add-on, a workaround, and a lot…

Issanou 26 September 2026
0 50 0 Comment on LocalMesh 2.0 — The Photo-to-3D App That Stopped Inventing the Back of Your Object and Started Reading It From Your Photos (Local, Offline, $39)
LocalMesh 2.0 — The Photo-to-3D App That Stopped Inventing the Back of Your Object and Started Reading It From Your Photos (Local, Offline, $39) LocalMesh 2.0 — The Photo-to-3D App That Stopped Inventing the Back of Your Object and Started Reading It From Your Photos (Local, Offline, $39) LocalMesh 2.0 — The Photo-to-3D App That Stopped Inventing the Back of Your Object and Started Reading It From Your Photos (Local, Offline, $39) LocalMesh 2.0 — The Photo-to-3D App That Stopped Inventing the Back of Your Object and Started Reading It From Your Photos (Local, Offline, $39)

LocalMesh 2.0 — The Photo-to-3D App That Stopped Inventing the Back of Your Object and Started Reading It From Your Photos (Local, Offline, $39)

You point your camera at a thing on your desk. Twenty seconds later you are spinning a…

Issanou 26 September 2026
0 42 1 Comment on Beeble SwitchX 2.0 — The Video-to-Video AI That Rebuilds Your Whole Shot in 4K and Never Touches the Actor
Beeble SwitchX 2.0 — The Video-to-Video AI That Rebuilds Your Whole Shot in 4K and Never Touches the Actor Beeble SwitchX 2.0 — The Video-to-Video AI That Rebuilds Your Whole Shot in 4K and Never Touches the Actor Beeble SwitchX 2.0 — The Video-to-Video AI That Rebuilds Your Whole Shot in 4K and Never Touches the Actor Beeble SwitchX 2.0 — The Video-to-Video AI That Rebuilds Your Whole Shot in 4K and Never Touches the Actor

Beeble SwitchX 2.0 — The Video-to-Video AI That Rebuilds Your Whole Shot in 4K and Never Touches the Actor

Most AI video tools ask you to give up control. You type a prompt, you cross your…

Issanou 23 September 2026
0 90 0 Comment on ComfyUI Just Made Image-to-3D Native — TRELLIS.2 and Pixal3D Ship Inside the Graph, No Custom Nodes, No CUDA Hell
ComfyUI Just Made Image-to-3D Native — TRELLIS.2 and Pixal3D Ship Inside the Graph, No Custom Nodes, No CUDA Hell ComfyUI Just Made Image-to-3D Native — TRELLIS.2 and Pixal3D Ship Inside the Graph, No Custom Nodes, No CUDA Hell ComfyUI Just Made Image-to-3D Native — TRELLIS.2 and Pixal3D Ship Inside the Graph, No Custom Nodes, No CUDA Hell ComfyUI Just Made Image-to-3D Native — TRELLIS.2 and Pixal3D Ship Inside the Graph, No Custom Nodes, No CUDA Hell

ComfyUI Just Made Image-to-3D Native — TRELLIS.2 and Pixal3D Ship Inside the Graph, No Custom Nodes, No CUDA Hell

Hook You want to turn a picture into a 3D model on your own GPU. Until now…

Issanou 23 September 2026
0 113 0 Comment on KaiNinja — The First AI 3D Model Trained on Parts an Agent Authored, So Your Generated Object Finally Comes Apart
KaiNinja — The First AI 3D Model Trained on Parts an Agent Authored, So Your Generated Object Finally Comes Apart KaiNinja — The First AI 3D Model Trained on Parts an Agent Authored, So Your Generated Object Finally Comes Apart KaiNinja — The First AI 3D Model Trained on Parts an Agent Authored, So Your Generated Object Finally Comes Apart KaiNinja — The First AI 3D Model Trained on Parts an Agent Authored, So Your Generated Object Finally Comes Apart

KaiNinja — The First AI 3D Model Trained on Parts an Agent Authored, So Your Generated Object Finally Comes Apart

Hook Every AI 3D model you have ever generated came out as one fused blob. The wheels…

Issanou 23 September 2026