IA Lab

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

0 24 0 Comment on TRELLIS.2 — Microsoft Just Rebuilt Image-to-3D Around One New Idea: A Voxel That Stores Shape and Surface Together, at 1536³, in Seconds
TRELLIS.2 — Microsoft Just Rebuilt Image-to-3D Around One New Idea: A Voxel That Stores Shape and Surface Together, at 1536³, in Seconds TRELLIS.2 — Microsoft Just Rebuilt Image-to-3D Around One New Idea: A Voxel That Stores Shape and Surface Together, at 1536³, in Seconds TRELLIS.2 — Microsoft Just Rebuilt Image-to-3D Around One New Idea: A Voxel That Stores Shape and Surface Together, at 1536³, in Seconds TRELLIS.2 — Microsoft Just Rebuilt Image-to-3D Around One New Idea: A Voxel That Stores Shape and Surface Together, at 1536³, in Seconds

TRELLIS.2 — Microsoft Just Rebuilt Image-to-3D Around One New Idea: A Voxel That Stores Shape and Surface Together, at 1536³, in Seconds

Microsoft Research just dropped TRELLIS.2, and it quietly moves the goalposts for image-to-3D. Feed it one picture…

Issanou 26 August 2026
0 28 1 Comment on Insta360 Just Put Gaussian Splat Capture in Your Pocket — The X6 Turns a $699 Camera Into a One-Loop Radiance-Field Scanner (Powered by the KIRI Engine Team)
Insta360 Just Put Gaussian Splat Capture in Your Pocket — The X6 Turns a $699 Camera Into a One-Loop Radiance-Field Scanner (Powered by the KIRI Engine Team) Insta360 Just Put Gaussian Splat Capture in Your Pocket — The X6 Turns a $699 Camera Into a One-Loop Radiance-Field Scanner (Powered by the KIRI Engine Team) Insta360 Just Put Gaussian Splat Capture in Your Pocket — The X6 Turns a $699 Camera Into a One-Loop Radiance-Field Scanner (Powered by the KIRI Engine Team) Insta360 Just Put Gaussian Splat Capture in Your Pocket — The X6 Turns a $699 Camera Into a One-Loop Radiance-Field Scanner (Powered by the KIRI Engine Team)

Insta360 Just Put Gaussian Splat Capture in Your Pocket — The X6 Turns a $699 Camera Into a One-Loop Radiance-Field Scanner (Powered by the KIRI Engine Team)

The rig that turns a real place into a 3D Gaussian splat just shrank to the size…

Issanou 26 August 2026
0 33 0 Comment on Meshy T2 — The AI That Finally Gives You a Mesh You’d Open in Blender: Artist Topology From One Image in 6 Seconds
Meshy T2 — The AI That Finally Gives You a Mesh You’d Open in Blender: Artist Topology From One Image in 6 Seconds Meshy T2 — The AI That Finally Gives You a Mesh You’d Open in Blender: Artist Topology From One Image in 6 Seconds Meshy T2 — The AI That Finally Gives You a Mesh You’d Open in Blender: Artist Topology From One Image in 6 Seconds Meshy T2 — The AI That Finally Gives You a Mesh You’d Open in Blender: Artist Topology From One Image in 6 Seconds

Meshy T2 — The AI That Finally Gives You a Mesh You’d Open in Blender: Artist Topology From One Image in 6 Seconds

Every AI 3D tool can hand you a shape. Almost none can hand you a shape you’d…

Issanou 22 August 2026
0 24 0 Comment on MotionBricks — NVIDIA Just Killed Motion Matching: One Model That Animates Your Game Character at 15,000 FPS and Drives a Real Robot
MotionBricks — NVIDIA Just Killed Motion Matching: One Model That Animates Your Game Character at 15,000 FPS and Drives a Real Robot MotionBricks — NVIDIA Just Killed Motion Matching: One Model That Animates Your Game Character at 15,000 FPS and Drives a Real Robot MotionBricks — NVIDIA Just Killed Motion Matching: One Model That Animates Your Game Character at 15,000 FPS and Drives a Real Robot MotionBricks — NVIDIA Just Killed Motion Matching: One Model That Animates Your Game Character at 15,000 FPS and Drives a Real Robot

MotionBricks — NVIDIA Just Killed Motion Matching: One Model That Animates Your Game Character at 15,000 FPS and Drives a Real Robot

NVIDIA just dropped a motion model that generates 15,000 frames of character animation per second. Same model,…

Issanou 22 August 2026
0 25 0 Comment on Scenario Just Made Capturing a Real Place a Button — Tencent’s World Model Now Turns a Phone Video Into a Splat, Next to Your Game Art
Scenario Just Made Capturing a Real Place a Button — Tencent’s World Model Now Turns a Phone Video Into a Splat, Next to Your Game Art Scenario Just Made Capturing a Real Place a Button — Tencent’s World Model Now Turns a Phone Video Into a Splat, Next to Your Game Art Scenario Just Made Capturing a Real Place a Button — Tencent’s World Model Now Turns a Phone Video Into a Splat, Next to Your Game Art Scenario Just Made Capturing a Real Place a Button — Tencent’s World Model Now Turns a Phone Video Into a Splat, Next to Your Game Art

Scenario Just Made Capturing a Real Place a Button — Tencent’s World Model Now Turns a Phone Video Into a Splat, Next to Your Game Art

You point your phone at a room. You take a short walk-around video. A few minutes later,…

Issanou 22 August 2026
0 45 1 Comment on Tripo’s Makers Just Won SIGGRAPH Best in Show — A Room Full of Strangers Texted Photos and Watched Them Become Rigged, Animated 3D Characters in Seconds
Tripo’s Makers Just Won SIGGRAPH Best in Show — A Room Full of Strangers Texted Photos and Watched Them Become Rigged, Animated 3D Characters in Seconds Tripo’s Makers Just Won SIGGRAPH Best in Show — A Room Full of Strangers Texted Photos and Watched Them Become Rigged, Animated 3D Characters in Seconds Tripo’s Makers Just Won SIGGRAPH Best in Show — A Room Full of Strangers Texted Photos and Watched Them Become Rigged, Animated 3D Characters in Seconds Tripo’s Makers Just Won SIGGRAPH Best in Show — A Room Full of Strangers Texted Photos and Watched Them Become Rigged, Animated 3D Characters in Seconds

Tripo’s Makers Just Won SIGGRAPH Best in Show — A Room Full of Strangers Texted Photos and Watched Them Become Rigged, Animated 3D Characters in Seconds

At SIGGRAPH 2026’s Real-Time Live!, the audience became the demo. People pulled out their phones, sent in…

Issanou 7 August 2026
0 50 0 Comment on Mage-Flow — Microsoft Just Shipped a 4B Image Model That Generates AND Edits at Native Resolution, Beats a 20B Rival, and It’s MIT-Licensed
Mage-Flow — Microsoft Just Shipped a 4B Image Model That Generates AND Edits at Native Resolution, Beats a 20B Rival, and It’s MIT-Licensed Mage-Flow — Microsoft Just Shipped a 4B Image Model That Generates AND Edits at Native Resolution, Beats a 20B Rival, and It’s MIT-Licensed Mage-Flow — Microsoft Just Shipped a 4B Image Model That Generates AND Edits at Native Resolution, Beats a 20B Rival, and It’s MIT-Licensed Mage-Flow — Microsoft Just Shipped a 4B Image Model That Generates AND Edits at Native Resolution, Beats a 20B Rival, and It’s MIT-Licensed

Mage-Flow — Microsoft Just Shipped a 4B Image Model That Generates AND Edits at Native Resolution, Beats a 20B Rival, and It’s MIT-Licensed

Microsoft just dropped a text-to-image and image-editing model that fits on a single GPU, generates a 1024px…

Issanou 6 August 2026
0 49 0 Comment on OmniFaceRig — Meta Just Automated the One Job AI 3D Kept Skipping: Giving Your Character a Face That Can Actually Talk
OmniFaceRig — Meta Just Automated the One Job AI 3D Kept Skipping: Giving Your Character a Face That Can Actually Talk OmniFaceRig — Meta Just Automated the One Job AI 3D Kept Skipping: Giving Your Character a Face That Can Actually Talk OmniFaceRig — Meta Just Automated the One Job AI 3D Kept Skipping: Giving Your Character a Face That Can Actually Talk OmniFaceRig — Meta Just Automated the One Job AI 3D Kept Skipping: Giving Your Character a Face That Can Actually Talk

OmniFaceRig — Meta Just Automated the One Job AI 3D Kept Skipping: Giving Your Character a Face That Can Actually Talk

Every AI 3D pipeline has a dirty little secret. It hands you a gorgeous mesh with no…

Issanou 4 August 2026
0 43 0 Comment on PixWorld — One Model That Both Reconstructs a Real Place and Dreams Up a New One, in 0.6 Seconds, With No VAE in Sight
PixWorld — One Model That Both Reconstructs a Real Place and Dreams Up a New One, in 0.6 Seconds, With No VAE in Sight PixWorld — One Model That Both Reconstructs a Real Place and Dreams Up a New One, in 0.6 Seconds, With No VAE in Sight PixWorld — One Model That Both Reconstructs a Real Place and Dreams Up a New One, in 0.6 Seconds, With No VAE in Sight PixWorld — One Model That Both Reconstructs a Real Place and Dreams Up a New One, in 0.6 Seconds, With No VAE in Sight

PixWorld — One Model That Both Reconstructs a Real Place and Dreams Up a New One, in 0.6 Seconds, With No VAE in Sight

For two years the AI 3D world has quietly split into two tribes: the reconstruction people who…

Issanou 4 August 2026
0 42 0 Comment on FLUX 3 — The Lab Behind Half of ComfyUI Just Quit Image Models: One Backbone for Video, Sound, and a Robot Arm
FLUX 3 — The Lab Behind Half of ComfyUI Just Quit Image Models: One Backbone for Video, Sound, and a Robot Arm FLUX 3 — The Lab Behind Half of ComfyUI Just Quit Image Models: One Backbone for Video, Sound, and a Robot Arm FLUX 3 — The Lab Behind Half of ComfyUI Just Quit Image Models: One Backbone for Video, Sound, and a Robot Arm FLUX 3 — The Lab Behind Half of ComfyUI Just Quit Image Models: One Backbone for Video, Sound, and a Robot Arm

FLUX 3 — The Lab Behind Half of ComfyUI Just Quit Image Models: One Backbone for Video, Sound, and a Robot Arm

The lab that gave the open-source world FLUX.1 and FLUX.2 just stopped making image models. On July…

Issanou 4 August 2026