Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Self-Refining Video Sampling: inference-time method using a video generator as its own refiner to correct physics and motion. no retraining needed; scores >70% human preference; is validated on Wan2.2 & Cosmos.

Wildminder

9,854 subscribers

32,055 views • 4 months ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

1/ Excited to announce MoCA - a simple yet effective video editing method that accomplishes diverse spatial (style, background, object) and motion video edits using a motion conditioned image animation model.

1/ Excited to announce MoCA - a simple yet effective video editing method that accomplishes diverse spatial (style, background, object) and motion video edits using a motion conditioned image animation model.

Wilson Yan

10,627 views • 2 years ago

Using a range rover to hit a coach bus on motion is mad physics

Using a range rover to hit a coach bus on motion is mad physics

Unique Movie Moments 🐬

31,680 views • 1 year ago

Self-Forcing++ for minute-scale video generation ByteDance's new method generates high-quality videos up to 4 min 15 sec! It scales diffusion models without long-video teachers or retraining, preserving fidelity and consistency.

Self-Forcing++ for minute-scale video generation ByteDance's new method generates high-quality videos up to 4 min 15 sec! It scales diffusion models without long-video teachers or retraining, preserving fidelity and consistency.

DailyPapers

17,720 views • 8 months ago

holy moly 🔥 Pusa 1.0 is a new image-to-video finetune based on Wan2.1-T2V-14B that surpasses Wan-I2V-14B and is: > 200x cheaper to train (only cost $500 🤯) > 5x faster inference > supports start-end frames, video extension and video completion

holy moly 🔥 Pusa 1.0 is a new image-to-video finetune based on Wan2.1-T2V-14B that surpasses Wan-I2V-14B and is: > 200x cheaper to train (only cost $500 🤯) > 5x faster inference > supports start-end frames, video extension and video completion

Dreaming Tulpa 🥓👑

15,688 views • 10 months ago

Withholding weapons validates the false & destructive narrative that Israel is acting unjustly by defending herself. Israel is conducting a war of self-defense more humanely than any army in history. And they're doing so against an enemy using its own civilians as human shields.

Withholding weapons validates the false & destructive narrative that Israel is acting unjustly by defending herself. Israel is conducting a war of self-defense more humanely than any army in history. And they're doing so against an enemy using its own civilians as human shields.

Nikki Haley

160,877 views • 2 years ago

SHOCKING? #Africa is not allowed to be developed and become a self-sufficient in #foodsecurity. Not even allowed to grow its own food. If you try, then you'll be sanctioned & labelled as dictator. Consequently, they'll cook "human rights & regime change" against you.

SHOCKING? #Africa is not allowed to be developed and become a self-sufficient in #foodsecurity. Not even allowed to grow its own food. If you try, then you'll be sanctioned & labelled as dictator. Consequently, they'll cook "human rights & regime change" against you.

ኤርትራ 🇪🇷 ERITREA

26,085 views • 3 years ago

China Foreign Ministry: "The U.S. is using human rights as a pretext to wrongfully impose unilateral sanctions and interfere in Cuba's internal affairs... If the U.S. truly cares about human rights, it should reflect on its own human rights violations at Cuba's Guantanamo Bay"

China Foreign Ministry: "The U.S. is using human rights as a pretext to wrongfully impose unilateral sanctions and interfere in Cuba's internal affairs... If the U.S. truly cares about human rights, it should reflect on its own human rights violations at Cuba's Guantanamo Bay"

COMBATE |🇵🇷

15,442 views • 10 months ago

Excited to open-source GMR: General Motion Retargeting. Real-time human-to-humanoid retargeting on your laptop. Supports diverse motion formats & robots. Unlock whole-body humanoid teleoperation (e.g., TWIST). video with 🔊

Excited to open-source GMR: General Motion Retargeting. Real-time human-to-humanoid retargeting on your laptop. Supports diverse motion formats & robots. Unlock whole-body humanoid teleoperation (e.g., TWIST). video with 🔊

Yanjie Ze

88,183 views • 10 months ago

Using civilians as human shields is a crime against humanity — something Iran’s regime and its proxies do all the time. Islam is a crime against humanity.

Using civilians as human shields is a crime against humanity — something Iran’s regime and its proxies do all the time. Islam is a crime against humanity.

Liza Rosen

28,549 views • 2 months ago

Introducing Core Sandbox Alpha, the first inference-time learning coding assistant. Every interaction affects reasoning immediately. No RL. No retraining. No fine-tuning.

Introducing Core Sandbox Alpha, the first inference-time learning coding assistant. Every interaction affects reasoning immediately. No RL. No retraining. No fine-tuning.

Rei

52,493 views • 6 months ago

A viral video has surfaced showing a man using a metal object to take cashew nuts from stored bags, sparking debate online. Many have questioned whether he was conducting a legitimate sampling process and if the method used was appropriate.

A viral video has surfaced showing a man using a metal object to take cashew nuts from stored bags, sparking debate online. Many have questioned whether he was conducting a legitimate sampling process and if the method used was appropriate.

THE STATE NEWS

104,006 views • 1 month ago

New Video Series: Statistics & Data Analysis! 35 videos, 10 hours: Random sampling, Central limit theorem, Distribution estimation, Method of moments, Maximum likelihood estimation, Hypothesis testing, Monte Carlo sampling, Bayesian statistics, and more!

New Video Series: Statistics & Data Analysis! 35 videos, 10 hours: Random sampling, Central limit theorem, Distribution estimation, Method of moments, Maximum likelihood estimation, Hypothesis testing, Monte Carlo sampling, Bayesian statistics, and more!

Steven Brunton

115,161 views • 9 months ago

I'm excited to share our new work Diffusion as Shader (DaS), a versatile controllable video generation method for various tasks: object manipulation, camera control, mesh-to-video, and motion transfer. Project page: Github:

I'm excited to share our new work Diffusion as Shader (DaS), a versatile controllable video generation method for various tasks: object manipulation, camera control, mesh-to-video, and motion transfer. Project page: Github:

Yuan Liu

33,363 views • 1 year ago

Our new video on using textures in Figma is now live. We cover everything from the basics to developing practical skills by designing three unique cards using the Noise & Texture, Pattern Hero, and Noise Grid Generator plugins.

Our new video on using textures in Figma is now live. We cover everything from the basics to developing practical skills by designing three unique cards using the Noise & Texture, Pattern Hero, and Noise Grid Generator plugins.

Alex Barashkov

58,210 views • 1 year ago

VLA policies learn generalist robot behaviors from massive teleoperation datasets, hoping that the right behavior emerges. But they rarely use perception during training or inference: powerful foundation models of 3D geometry, semantics, or human motion are ignored. TimSong, Long Le, and our GRASP Laboratory team introduce Omniguide, based on a simple idea: Instead of retraining policies, let perception guide action at inference time. We express diverse guidance sources as attractive and repulsive energy fields in 3D space and inject their gradients into the generative process of VLA policies. This lets perception modules steer actions without retraining the policy: A modular path toward composable robot intelligence. Project: Paper:

VLA policies learn generalist robot behaviors from massive teleoperation datasets, hoping that the right behavior emerges. But they rarely use perception during training or inference: powerful foundation models of 3D geometry, semantics, or human motion are ignored. TimSong, Long Le, and our GRASP Laboratory team introduce Omniguide, based on a simple idea: Instead of retraining policies, let perception guide action at inference time. We express diverse guidance sources as attractive and repulsive energy fields in 3D space and inject their gradients into the generative process of VLA policies. This lets perception modules steer actions without retraining the policy: A modular path toward composable robot intelligence. Project: Paper:

Kostas Daniilidis

46,883 views • 2 months ago

🎉 ComfyUI now natively supports NVIDIA’s Cosmos-Predict2 model family! Cosmos-Predict2 brings high-fidelity, physics-aware Image generation and Video2World (Image-to-Video) generation. Another reality inside ComfyUI!

🎉 ComfyUI now natively supports NVIDIA’s Cosmos-Predict2 model family! Cosmos-Predict2 brings high-fidelity, physics-aware Image generation and Video2World (Image-to-Video) generation. Another reality inside ComfyUI!

ComfyUI

16,925 views • 11 months ago

Physics teachers. Just a quick demo using #repurposedphysics kit. This time for sound lessons. Just a slinky and a cup needed 🙂

Physics teachers. Just a quick demo using #repurposedphysics kit. This time for sound lessons. Just a slinky and a cup needed 🙂

Dr Dan Jones

19,445 views • 3 years ago

Time for another AI video comparison! Today we're looking at Kling 1.6 vs Adobe's new Firefly video generator. While Firefly shows some cool dynamics in its motion, it struggles a bit with distortion. Kling 1.6 continues to impress with its consistency and better control over the final output. It's fascinating to see how different tools approach video generation. Which results catch your eye more?

Time for another AI video comparison! Today we're looking at Kling 1.6 vs Adobe's new Firefly video generator. While Firefly shows some cool dynamics in its motion, it struggles a bit with distortion. Kling 1.6 continues to impress with its consistency and better control over the final output. It's fascinating to see how different tools approach video generation. Which results catch your eye more?

Curious Refuge

37,866 views • 1 year ago

Here’s a fun STEM project: make a paper horse that walks on its own using physics. Thanks to the incline, gravity and their structure, the horses walk by rocking back and forth on their curved legs.

Here’s a fun STEM project: make a paper horse that walks on its own using physics. Thanks to the incline, gravity and their structure, the horses walk by rocking back and forth on their curved legs.

Massimo

566,783 views • 2 years ago

Learning to fight is retraining your baseline human instinct

Sensitive content

Learning to fight is retraining your baseline human instinct

TOPG MERCH

116,860 views • 6 months ago