Loading video...

Video Failed to Load

Go Home

Pixal3D Pixel-Aligned 3D Generation from Images

_akhaliq's profile picture

AK

524,731 subscribers

31,427 views • 4 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

3 Comments

AK's profile picture
AK4 months ago

paper: app:

Gains Agent's profile picture
Gains Agent4 months ago

Pixel-aligned generation from images is the right step. The production hurdle: 30fps on consumer hardware. Cross that and you break photogrammetry as the capture standard.

ULWI's profile picture
ULWI4 months ago

You must bring training, then that would be great.

Related Videos

🚨 Pixal3D is now live on fal! 🧊 High-fidelity 3D assets from a single image 🎯 Pixel-aligned back-projection for direct pixel-to-3D correspondence 🎨 Detailed geometry with PBR textures, ready for downstream pipelines
0:38

🚨 Pixal3D is now live on fal! 🧊 High-fidelity 3D assets from a single image 🎯 Pixel-aligned back-projection for direct pixel-to-3D correspondence 🎨 Detailed geometry with PBR textures, ready for downstream pipelines

fal

14,192 views • 4 months ago

Pixel-perfect 3D generation is here. 🧵 Tencent’s new Free Local Pixal3D project every pixel directly into 3D space. No more "AI hallucinations"—just sub-pixel faithful reconstruction. ✅ Open Weights ✅ Local (24GB VRAM) ✅ Matches silhouettes perfectly Is this the new benchmark for open-source 3D? Full workflow and test: 👇 #3DAI #Pixal3D #GameDev #Tencen #OpenSource #3DModeling
1:13

Pixel-perfect 3D generation is here. 🧵 Tencent’s new Free Local Pixal3D project every pixel directly into 3D space. No more "AI hallucinations"—just sub-pixel faithful reconstruction. ✅ Open Weights ✅ Local (24GB VRAM) ✅ Matches silhouettes perfectly Is this the new benchmark for open-source 3D? Full workflow and test: 👇 #3DAI #Pixal3D #GameDev #Tencen #OpenSource #3DModeling

Stefan 3D AI

23,062 views • 4 months ago

Pixal3D by Tencent. 3D models from images with unprecedented accuracy. - Explicit 2D-3D correspondence - Aggregates multi-view feature volumes for global consistency - 93.57% IoU on Toys4K benchmark - Outperforms TRELLIS/Hunyuan3D-2.1 based on Direct3D-S2 + DINOv2
0:53

Pixal3D by Tencent. 3D models from images with unprecedented accuracy. - Explicit 2D-3D correspondence - Aggregates multi-view feature volumes for global consistency - 93.57% IoU on Toys4K benchmark - Outperforms TRELLIS/Hunyuan3D-2.1 based on Direct3D-S2 + DINOv2

Wildminder

22,170 views • 4 months ago

PixWorld Unifying 3D Scene Generation and Reconstruction in Pixel Space
1:01

PixWorld Unifying 3D Scene Generation and Reconstruction in Pixel Space

AK

16,740 views • 2 months ago

Direct3D-S2 Gigascale 3D Generation Made Easy with Spatial Sparse Attention high resolution 3D generation from image
0:28

Direct3D-S2 Gigascale 3D Generation Made Easy with Spatial Sparse Attention high resolution 3D generation from image

AK

59,276 views • 1 year ago

Custom pixel-art-pont-and-click-engine prototype. Testing an idea I had for projection based 3D modeling tool to make real 3D environments that remain pixel perfect and in still images look like drawings. All of the artwork created in engine #adventuregame
2:26

Custom pixel-art-pont-and-click-engine prototype. Testing an idea I had for projection based 3D modeling tool to make real 3D environments that remain pixel perfect and in still images look like drawings. All of the artwork created in engine #adventuregame

Jussi Kemppainen | Drivers of the Apocalypse

21,433 views • 9 days ago

🤯 Gemini 3.1 Pro Google DeepMind generates parametric 3D models straight from images! 🚀Paired with Rodin Gen-2 Edit's Refine, #Gemini effortlessly delivers million-face detailed #3D models. 🔥3D Generation workflows are evolving! See it in action 👇 #CG #Hyper3D
0:39

🤯 Gemini 3.1 Pro Google DeepMind generates parametric 3D models straight from images! 🚀Paired with Rodin Gen-2 Edit's Refine, #Gemini effortlessly delivers million-face detailed #3D models. 🔥3D Generation workflows are evolving! See it in action 👇 #CG #Hyper3D

Hyper3D🔜SIGGRAPH26

77,998 views • 6 months ago

Apple drops a new 3D model generator, LiTo > Learns geometry + appearance together in a unified 3D latent space > Turns RGB-D images into a compact surface light field > Recreates specular highlights under complex lighting > Generates 3D objects from a single image with input-consistent materials Big step toward more realistic 3D generation
0:33

Apple drops a new 3D model generator, LiTo > Learns geometry + appearance together in a unified 3D latent space > Turns RGB-D images into a compact surface light field > Recreates specular highlights under complex lighting > Generates 3D objects from a single image with input-consistent materials Big step toward more realistic 3D generation

⚡AI Search⚡

13,033 views • 3 months ago

Nano Banana 2, new state of the art in image generation and editing combined with Gemini’s real-world knowledge! You can simulate 3D CAD models purely through images. From sketch to real object!
0:22

Nano Banana 2, new state of the art in image generation and editing combined with Gemini’s real-world knowledge! You can simulate 3D CAD models purely through images. From sketch to real object!

koray kavukcuoglu

87,809 views • 6 months ago

Good. ComfyUI just got native MoGe support - 3D geometry from monocular images.
0:35

Good. ComfyUI just got native MoGe support - 3D geometry from monocular images.

Wildminder

16,196 views • 4 months ago

GPT4-V can describe the location via text, but can’t accurately output the coordinate of each word. Introducing: Pixel Aligned Language Models. It generates image captions along with the aligned pixel coordinates of the image. (1/n
1:34

GPT4-V can describe the location via text, but can’t accurately output the coordinate of each word. Introducing: Pixel Aligned Language Models. It generates image captions along with the aligned pixel coordinates of the image. (1/n

Jiarui Xu

65,304 views • 2 years ago

NeuralRemaster Phase-Preserving Diffusion for Structure-Aligned Generation
3:32

NeuralRemaster Phase-Preserving Diffusion for Structure-Aligned Generation

AK

39,909 views • 9 months ago

📢📢GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction📢📢 Reconstructing high-fidelity 3D scenes from sparse RGB input is hard. It needs a strong 3D prior! We reformulate multi-view scene reconstruction as conditional 3D generation over overlapping spatial chunks, lifting posed image features into a generative shape prior via 3D conditioning. As an example prior, we build on Trellis2, and train it such that its reconstruction is pixel aligned and matches from all views. GenRecon achieves unprecedented reconstruction quality from any sparse RGB input sequence, even from a phone capture. The reconstruction also includes PBR materials which facilitates relighting and virtual object insertion. Amazing work by Katharina Schmid, Nicolas von Lützow, Jozef, Angela Dai
0:26

📢📢GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction📢📢 Reconstructing high-fidelity 3D scenes from sparse RGB input is hard. It needs a strong 3D prior! We reformulate multi-view scene reconstruction as conditional 3D generation over overlapping spatial chunks, lifting posed image features into a generative shape prior via 3D conditioning. As an example prior, we build on Trellis2, and train it such that its reconstruction is pixel aligned and matches from all views. GenRecon achieves unprecedented reconstruction quality from any sparse RGB input sequence, even from a phone capture. The reconstruction also includes PBR materials which facilitates relighting and virtual object insertion. Amazing work by Katharina Schmid, Nicolas von Lützow, Jozef, Angela Dai

Matthias Niessner

18,321 views • 3 months ago

SceneDreamer: Unbounded 3D Scene Generation from 2D Image Collections abs: project page:
0:58

SceneDreamer: Unbounded 3D Scene Generation from 2D Image Collections abs: project page:

AK

317,635 views • 3 years ago

RTFM (Real-Time Frame Model) explores real-time, interactive frame generation from images, producing new views on the fly without explicit 3D. It differs from Marble, which targets high-fidelity persistent worlds (3DGS). We’re interested in how ideas from both might connect 🌎
2:21

RTFM (Real-Time Frame Model) explores real-time, interactive frame generation from images, producing new views on the fly without explicit 3D. It differs from Marble, which targets high-fidelity persistent worlds (3DGS). We’re interested in how ideas from both might connect 🌎

World Labs

74,757 views • 7 months ago

With every generation of Pixel, Google delivers innovation to consumers, and our latest devices are no different. Hear about how Pixel uses AI to bring helpful experiences to life from VP of Product Management at Pixel, Venkat Rapaka. Learn more →
2:04

With every generation of Pixel, Google delivers innovation to consumers, and our latest devices are no different. Hear about how Pixel uses AI to bring helpful experiences to life from VP of Product Management at Pixel, Venkat Rapaka. Learn more →

Made by Google

22,391 views • 1 year ago