Loading video...

Video Failed to Load

Go Home

Most motion papers tailor one controller to one specific task. This year at SIGGRAPH, our research team asks: can motor control itself be pretrained and reused? Generative Pretrained Controllers, or GPC, turn motor skills into a vocabulary of discrete tokens and train a transformer-based generative controller through next-token prediction....

NVIDIAAI's profile picture

NVIDIA AI

330,699 subscribers

175,697 views • 1 month ago •via X (Twitter)

Science & Technology
LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

NEW RESEARCH: You can now create a new robot optimized for any given task! I love this new project by Huy Ha, Shuran Song, and others. Called "Transformer Transformer: A Unified Model for Motion-Conditioned Robot Co-design", it generates a robot's physical design and its controller together from a task spec. DEFINITIONS: - Reward function: A scoring rule that assigns a number to how well a behavior achieves the task. Here, it is the objective the generated design is pushed to maximize (e.g., track the target motion with low error). - Tokenizing: dividing continuous or structured data (a robot's links, joints, motor specs, states, actions) into a discrete vocabulary of symbols a transformer can process, the same step that turned pixels and audio into "language" for these models. - Diffusion transformer (DiT): A transformer trained to turn random noise into structured output through iterative denoising. Here, it generates robot bodies and trajectories instead of images. - MuJoCo: The standard fast physics simulator for robotics research (DeepMind-maintained). The Menagerie is its curated zoo of ready-to-use robot models. - CMA-ES: Covariance Matrix Adaptation Evolution Strategy, the workhorse black-box optimizer: it evolves a population of candidate designs, keeps the best, and needs thousands of simulator rollouts. - Bimanual multi-trajectory optimization: Finding one design/controller that performs well across several target motions for a two-armed robot at once, harder than optimizing for a single arm and a single motion. - BERT/MAE masked-modeling trick: Train one model to fill in whatever parts of the input you hide (words for BERT, image patches for MAE); at inference, choosing what to mask chooses the task, so masking the body makes it a designer and masking the actions makes it a controller. In practice, you give it a target end-effector motion and a reward function, and it outputs a complete embodiment (link, joint, motor, and inertial property), as well as a controller to drive it. It works by tokenizing both the body (links/joints/motors) and the dynamics (states/actions) into a compact scheme called RoboTokens, training a diffusion transformer (DiT) over them. The same model predicts dynamics using those predictions ("Dynamics Self-Guidance") to push generated designs toward higher reward at inference time. Masking different token types (using the BERT/MAE masked-modeling trick) lets the one model do three jobs: generate an embodiment, control an arbitrary embodiment, or design one conditioned on a motion. It is trained on 11 robots from the MuJoCo Menagerie (0.65 kg hand to 67.5 kg quadruped, 6–35 joints), and validated in sim and on a physical ALOHA doing cloth flinging. I like the fact that this approach inverts the entire recent robotics ideas: designing a policy for a fixed robot -> designing the robot for a fixed task. Every other approach assumes the body is given and learns a controller. Transformer Transformer takes the task (target motion + reward), then generates the body and controller jointly. In practice, it is a ~180× speedup over the standard optimizer at equal-or-better quality. It reaches "CMA-ES-level quality in seconds" and finishes bimanual multi-trajectory optimization in that is worth underlining nowadays! Also worth mentioning: this is the lab behind UMI and Handroid, that I mentioned here previously! The team seems extremely creative, i love these out-of-the-box approaches. Enjoy watching the demo of robot optimization in 3D, data acquisition, then real-life testing:
0:28

NEW RESEARCH: You can now create a new robot optimized for any given task! I love this new project by Huy Ha, Shuran Song, and others. Called "Transformer Transformer: A Unified Model for Motion-Conditioned Robot Co-design", it generates a robot's physical design and its controller together from a task spec. DEFINITIONS: - Reward function: A scoring rule that assigns a number to how well a behavior achieves the task. Here, it is the objective the generated design is pushed to maximize (e.g., track the target motion with low error). - Tokenizing: dividing continuous or structured data (a robot's links, joints, motor specs, states, actions) into a discrete vocabulary of symbols a transformer can process, the same step that turned pixels and audio into "language" for these models. - Diffusion transformer (DiT): A transformer trained to turn random noise into structured output through iterative denoising. Here, it generates robot bodies and trajectories instead of images. - MuJoCo: The standard fast physics simulator for robotics research (DeepMind-maintained). The Menagerie is its curated zoo of ready-to-use robot models. - CMA-ES: Covariance Matrix Adaptation Evolution Strategy, the workhorse black-box optimizer: it evolves a population of candidate designs, keeps the best, and needs thousands of simulator rollouts. - Bimanual multi-trajectory optimization: Finding one design/controller that performs well across several target motions for a two-armed robot at once, harder than optimizing for a single arm and a single motion. - BERT/MAE masked-modeling trick: Train one model to fill in whatever parts of the input you hide (words for BERT, image patches for MAE); at inference, choosing what to mask chooses the task, so masking the body makes it a designer and masking the actions makes it a controller. In practice, you give it a target end-effector motion and a reward function, and it outputs a complete embodiment (link, joint, motor, and inertial property), as well as a controller to drive it. It works by tokenizing both the body (links/joints/motors) and the dynamics (states/actions) into a compact scheme called RoboTokens, training a diffusion transformer (DiT) over them. The same model predicts dynamics using those predictions ("Dynamics Self-Guidance") to push generated designs toward higher reward at inference time. Masking different token types (using the BERT/MAE masked-modeling trick) lets the one model do three jobs: generate an embodiment, control an arbitrary embodiment, or design one conditioned on a motion. It is trained on 11 robots from the MuJoCo Menagerie (0.65 kg hand to 67.5 kg quadruped, 6–35 joints), and validated in sim and on a physical ALOHA doing cloth flinging. I like the fact that this approach inverts the entire recent robotics ideas: designing a policy for a fixed robot -> designing the robot for a fixed task. Every other approach assumes the body is given and learns a controller. Transformer Transformer takes the task (target motion + reward), then generates the body and controller jointly. In practice, it is a ~180× speedup over the standard optimizer at equal-or-better quality. It reaches "CMA-ES-level quality in seconds" and finishes bimanual multi-trajectory optimization in that is worth underlining nowadays! Also worth mentioning: this is the lab behind UMI and Handroid, that I mentioned here previously! The team seems extremely creative, i love these out-of-the-box approaches. Enjoy watching the demo of robot optimization in 3D, data acquisition, then real-life testing:

Léo

25,896 views • 15 days ago

MotionGPT: Human Motion as a Foreign Language paper page: Though the advancement of pre-trained large language models unfolds, the exploration of building a unified model for language and other multi-modal data, such as motion, remains challenging and untouched so far. Fortunately, human motion displays a semantic coupling akin to human language, often perceived as a form of body language. By fusing language data with large-scale motion models, motion-language pre-training that can enhance the performance of motion-related tasks becomes feasible. Driven by this insight, we propose MotionGPT, a unified, versatile, and user-friendly motion-language model to handle multiple motion-relevant tasks. Specifically, we employ the discrete vector quantization for human motion and transfer 3D motion into motion tokens, similar to the generation process of word tokens. Building upon this "motion vocabulary", we perform language modeling on both motion and text in a unified manner, treating human motion as a specific language. Moreover, inspired by prompt learning, we pre-train MotionGPT with a mixture of motion-language data and fine-tune it on prompt-based question-and-answer tasks. Extensive experiments demonstrate that MotionGPT achieves state-of-the-art performances on multiple motion tasks including text-driven motion generation, motion captioning, motion prediction, and motion in-between.
0:48

MotionGPT: Human Motion as a Foreign Language paper page: Though the advancement of pre-trained large language models unfolds, the exploration of building a unified model for language and other multi-modal data, such as motion, remains challenging and untouched so far. Fortunately, human motion displays a semantic coupling akin to human language, often perceived as a form of body language. By fusing language data with large-scale motion models, motion-language pre-training that can enhance the performance of motion-related tasks becomes feasible. Driven by this insight, we propose MotionGPT, a unified, versatile, and user-friendly motion-language model to handle multiple motion-relevant tasks. Specifically, we employ the discrete vector quantization for human motion and transfer 3D motion into motion tokens, similar to the generation process of word tokens. Building upon this "motion vocabulary", we perform language modeling on both motion and text in a unified manner, treating human motion as a specific language. Moreover, inspired by prompt learning, we pre-train MotionGPT with a mixture of motion-language data and fine-tune it on prompt-based question-and-answer tasks. Extensive experiments demonstrate that MotionGPT achieves state-of-the-art performances on multiple motion tasks including text-driven motion generation, motion captioning, motion prediction, and motion in-between.

AK

125,353 views • 3 years ago

NVIDIA MotionBricks brings humanoid robots closer to natural, fluid movement The system combines a modular latent generative model with Smart Primitives, packing more than 350,000 motion skills into a single backbone. It runs in real time at up to 15,000 FPS with just 2ms latency. Navigation, object interaction, sitting, jumping, obstacle crossing, and even different movement styles can be composed zero-shot, like reusable motion modules. It is already being used for whole-body control on the Unitree G1 and serves as the motion layer for NVIDIA GR00T Whole-Body Control. What makes this interesting is the move from training individual behaviors toward building a shared library of motion capabilities. If that approach scales, robot movement could become more modular, transferable, and easier to deploy across different tasks and environments.
1:43

NVIDIA MotionBricks brings humanoid robots closer to natural, fluid movement The system combines a modular latent generative model with Smart Primitives, packing more than 350,000 motion skills into a single backbone. It runs in real time at up to 15,000 FPS with just 2ms latency. Navigation, object interaction, sitting, jumping, obstacle crossing, and even different movement styles can be composed zero-shot, like reusable motion modules. It is already being used for whole-body control on the Unitree G1 and serves as the motion layer for NVIDIA GR00T Whole-Body Control. What makes this interesting is the move from training individual behaviors toward building a shared library of motion capabilities. If that approach scales, robot movement could become more modular, transferable, and easier to deploy across different tasks and environments.

CyberRobo

11,089 views • 2 months ago

Super clean and efficient meshes by an AI? YES! The typical 3D Generative AI solutions produce lots of artifacts and usually way to many polygons due to volumetric approaches. In comparison “MeshGPT creates triangle meshes by autoregressively sampling from a transformer model that has been trained to produce tokens from a learned geometric vocabulary. These tokens can then be decoded into the faces of a triangle mesh. This method generates clean, coherent, and compact meshes, characterized by sharp edges and high fidelity.” Surely it is limited by the trained vocabulary but various versions can be trained for specific sets to create generative model libraries for certain object groups. Very promising approach with the high quality.
0:25

Super clean and efficient meshes by an AI? YES! The typical 3D Generative AI solutions produce lots of artifacts and usually way to many polygons due to volumetric approaches. In comparison “MeshGPT creates triangle meshes by autoregressively sampling from a transformer model that has been trained to produce tokens from a learned geometric vocabulary. These tokens can then be decoded into the faces of a triangle mesh. This method generates clean, coherent, and compact meshes, characterized by sharp edges and high fidelity.” Surely it is limited by the trained vocabulary but various versions can be trained for specific sets to create generative model libraries for certain object groups. Very promising approach with the high quality.

René Schulte

20,772 views • 2 years ago

We are excited to share a research preview of our generative agent. The agent is being trained to solve the hardest tasks in 3D and beyond, using only keyboard and mouse actions. Join the waitlist: Our agent app runs on Windows or Mac, either locally or with one-click setup for a Windows VM. It’s still early days, but this paves the way for production-level workflows for the first time ever. Blog:
1:37

We are excited to share a research preview of our generative agent. The agent is being trained to solve the hardest tasks in 3D and beyond, using only keyboard and mouse actions. Join the waitlist: Our agent app runs on Windows or Mac, either locally or with one-click setup for a Windows VM. It’s still early days, but this paves the way for production-level workflows for the first time ever. Blog:

Common Sense Machines

155,199 views • 1 year ago

How do you give a humanoid the general motion capability? Not just single motions, but all motion? Introducing SONIC, our new work on supersizing motion tracking for natural humanoid control. We argue that motion tracking is the scalable foundation task for humanoids. So we "supersized" it: 9k+ GPU hours and 100M+ motion frames. But tracking alone is not enough; we show how to make a useful control system out of it: - Universal Kinematic Planner: Enables game-like gamepad control and high-level teleoperation, just like controlling a character in a game. - VR Full-Body Teleop: Direct, real-time whole-body control by a human wearing a VR headset. - VR Keypoint Teleop: Control the upper body (hands/head) while our planner handles robust locomotion automatically. - VLA Integration: We connect this motion tracker to autonomous Visual-Language-Action (VLA) models for autonomous task execution! We use a Universal Token Space to UNIFY this command space, turning our robust tracker into a general-purpose, programmable humanoid brain. This is the generalist "System 1" for humanoids. 🚀 Project: #Humanoids #Robotics #AI #FoundationModels #NVIDIAResearch 🧠🔥
1:32

How do you give a humanoid the general motion capability? Not just single motions, but all motion? Introducing SONIC, our new work on supersizing motion tracking for natural humanoid control. We argue that motion tracking is the scalable foundation task for humanoids. So we "supersized" it: 9k+ GPU hours and 100M+ motion frames. But tracking alone is not enough; we show how to make a useful control system out of it: - Universal Kinematic Planner: Enables game-like gamepad control and high-level teleoperation, just like controlling a character in a game. - VR Full-Body Teleop: Direct, real-time whole-body control by a human wearing a VR headset. - VR Keypoint Teleop: Control the upper body (hands/head) while our planner handles robust locomotion automatically. - VLA Integration: We connect this motion tracker to autonomous Visual-Language-Action (VLA) models for autonomous task execution! We use a Universal Token Space to UNIFY this command space, turning our robust tracker into a general-purpose, programmable humanoid brain. This is the generalist "System 1" for humanoids. 🚀 Project: #Humanoids #Robotics #AI #FoundationModels #NVIDIAResearch 🧠🔥

Zhengyi “Zen” Luo

62,326 views • 9 months ago

Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation paper page: Recent advances in generative modeling have led to promising progress on synthesizing 3D human motion from text, with methods that can generate character animations from short prompts and specified durations. However, using a single text prompt as input lacks the fine-grained control needed by animators, such as composing multiple actions and defining precise durations for parts of the motion. To address this, we introduce the new problem of timeline control for text-driven motion synthesis, which provides an intuitive, yet fine-grained, input interface for users. Instead of a single prompt, users can specify a multi-track timeline of multiple prompts organized in temporal intervals that may overlap. This enables specifying the exact timings of each action and composing multiple actions in sequence or at overlapping intervals. To generate composite animations from a multi-track timeline, we propose a new test-time denoising method. This method can be integrated with any pre-trained motion diffusion model to synthesize realistic motions that accurately reflect the timeline. At every step of denoising, our method processes each timeline interval (text prompt) individually, subsequently aggregating the predictions with consideration for the specific body parts engaged in each action. Experimental comparisons and ablations validate that our method produces realistic motions that respect the semantics and timing of given text prompts.
0:21

Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation paper page: Recent advances in generative modeling have led to promising progress on synthesizing 3D human motion from text, with methods that can generate character animations from short prompts and specified durations. However, using a single text prompt as input lacks the fine-grained control needed by animators, such as composing multiple actions and defining precise durations for parts of the motion. To address this, we introduce the new problem of timeline control for text-driven motion synthesis, which provides an intuitive, yet fine-grained, input interface for users. Instead of a single prompt, users can specify a multi-track timeline of multiple prompts organized in temporal intervals that may overlap. This enables specifying the exact timings of each action and composing multiple actions in sequence or at overlapping intervals. To generate composite animations from a multi-track timeline, we propose a new test-time denoising method. This method can be integrated with any pre-trained motion diffusion model to synthesize realistic motions that accurately reflect the timeline. At every step of denoising, our method processes each timeline interval (text prompt) individually, subsequently aggregating the predictions with consideration for the specific body parts engaged in each action. Experimental comparisons and ablations validate that our method produces realistic motions that respect the semantics and timing of given text prompts.

AK

126,595 views • 2 years ago

General Intuition CEO Pim de Witte, who's building foundation models trained on video game controller input data ("action-labeled gameplay clips"), says general intelligence won't "taste like an LLM": "We have a scale of data that's going to allow us to jump to the frontier in one capability — which is any system that can be controlled with a game controller (which is most robots) — and then, you can use that to create a sufficiently general intelligence." "As humans, the decision to talk or type is a very, very small subset of the actions that we can actually take." "So in order to create a sufficiently general intelligence to play 10,000+ video games, the model has to be able to predict across the entire action space of human cognition when they're interacting with these environments. Which are 2D and 3D environments, interfaces, long-horizon tasks, short-horizon tasks, [etc.]." "It has to be a sufficiently general intelligence in order to predict actions. Therefore, the type of model you get out is not going to taste like an LLM. This model is going to be incredibly good at navigating unforeseen environments. It's going to be incredibly good at zero-shotting any task that can be done with a game controller."
1:53

General Intuition CEO Pim de Witte, who's building foundation models trained on video game controller input data ("action-labeled gameplay clips"), says general intelligence won't "taste like an LLM": "We have a scale of data that's going to allow us to jump to the frontier in one capability — which is any system that can be controlled with a game controller (which is most robots) — and then, you can use that to create a sufficiently general intelligence." "As humans, the decision to talk or type is a very, very small subset of the actions that we can actually take." "So in order to create a sufficiently general intelligence to play 10,000+ video games, the model has to be able to predict across the entire action space of human cognition when they're interacting with these environments. Which are 2D and 3D environments, interfaces, long-horizon tasks, short-horizon tasks, [etc.]." "It has to be a sufficiently general intelligence in order to predict actions. Therefore, the type of model you get out is not going to taste like an LLM. This model is going to be incredibly good at navigating unforeseen environments. It's going to be incredibly good at zero-shotting any task that can be done with a game controller."

TBPN

80,482 views • 1 month ago

Humanoids need a single, generalist control policy for all of their physical tasks, not a new one for every new chore or demo. A policy for walking can't dance. A policy for dancing can't support mowing the lawn. We need to scale up humanoid control for diverse behaviors, just like we did with language and video models. Watch this humanoid, supported by a single control policy, to start its day, mow the lawn, jog, dance, and even hold the elevator for a colleague. This is the "System 1" for humanoids—a robust, natural humanoid controller that will support general-purpose autonomy. Project Page: #Robotics #AI #NVIDIA #SONIC #FoundationModel 🤖🔥
1:33

Humanoids need a single, generalist control policy for all of their physical tasks, not a new one for every new chore or demo. A policy for walking can't dance. A policy for dancing can't support mowing the lawn. We need to scale up humanoid control for diverse behaviors, just like we did with language and video models. Watch this humanoid, supported by a single control policy, to start its day, mow the lawn, jog, dance, and even hold the elevator for a colleague. This is the "System 1" for humanoids—a robust, natural humanoid controller that will support general-purpose autonomy. Project Page: #Robotics #AI #NVIDIA #SONIC #FoundationModel 🤖🔥

Zhengyi “Zen” Luo

15,921 views • 9 months ago

I don’t know if we live in a Matrix, but I know for sure that robots will spend most of their lives in simulation. Let machines train machines. I’m excited to introduce DexMimicGen, a massive-scale synthetic data generator that enables a humanoid robot to learn complex skills from only a handful of human demonstrations. Yes, as few as 5! DexMimicGen addresses the biggest pain point in robotics: where do we get data? Unlike with LLMs, where vast amounts of texts are readily available, you cannot simply download motor control signals from the internet. So researchers teleoperate the robots to collect motion data via XR headsets. They have to repeat the same skill over and over and over again, because neural nets are data hungry. This is a very slow and uncomfortable process. At NVIDIA, we believe the majority of high-quality tokens for robot foundation models will come from simulation. What DexMimicGen does is to trade GPU compute time for human time. It takes one motion trajectory from human, and multiplies into 1000s of new trajectories. A robot brain trained on this augmented dataset will generalize far better in the real world. Think of DexMimicGen as a learning signal amplifier. It maps a small dataset to a large (de facto infinite) dataset, using physics simulation in the loop. In this way, we free humans from babysitting the bots all day. The future of robot data is generative. The future of the entire robot learning pipeline will also be generative. 🧵
0:58

I don’t know if we live in a Matrix, but I know for sure that robots will spend most of their lives in simulation. Let machines train machines. I’m excited to introduce DexMimicGen, a massive-scale synthetic data generator that enables a humanoid robot to learn complex skills from only a handful of human demonstrations. Yes, as few as 5! DexMimicGen addresses the biggest pain point in robotics: where do we get data? Unlike with LLMs, where vast amounts of texts are readily available, you cannot simply download motor control signals from the internet. So researchers teleoperate the robots to collect motion data via XR headsets. They have to repeat the same skill over and over and over again, because neural nets are data hungry. This is a very slow and uncomfortable process. At NVIDIA, we believe the majority of high-quality tokens for robot foundation models will come from simulation. What DexMimicGen does is to trade GPU compute time for human time. It takes one motion trajectory from human, and multiplies into 1000s of new trajectories. A robot brain trained on this augmented dataset will generalize far better in the real world. Think of DexMimicGen as a learning signal amplifier. It maps a small dataset to a large (de facto infinite) dataset, using physics simulation in the loop. In this way, we free humans from babysitting the bots all day. The future of robot data is generative. The future of the entire robot learning pipeline will also be generative. 🧵

Jim Fan

165,246 views • 1 year ago

Glad that our work “Inference-Time Enhancement of Generative Robot Policies via Predictive World Modeling”, led by Han Qi, has been accepted to IEEE Robotics and Automation Letters! 🎉 We propose Generative Predictive Control (GPC): sample action proposals from a pretrained diffusion policy (“look back”), roll them out with a diffusion-based action-conditioned video world model (“look forward”), then rank or optimize the actions using either a learned reward model or VLM preferences. Conceptually, this is trajectory optimization / MPC with hybrid sampling + gradient optimization, interpreted through modern diffusion priors and video world models. Interestingly, we first posted the paper on arXiv in Feb 2025, when action-conditioned video world models for planning were still rare—now this direction is rapidly gaining traction. Still many open questions, e.g., • how to avoid local minima in planning • what representations work best for world models • how to balance physics priors vs. data-driven learning Paper:
1:00

Glad that our work “Inference-Time Enhancement of Generative Robot Policies via Predictive World Modeling”, led by Han Qi, has been accepted to IEEE Robotics and Automation Letters! 🎉 We propose Generative Predictive Control (GPC): sample action proposals from a pretrained diffusion policy (“look back”), roll them out with a diffusion-based action-conditioned video world model (“look forward”), then rank or optimize the actions using either a learned reward model or VLM preferences. Conceptually, this is trajectory optimization / MPC with hybrid sampling + gradient optimization, interpreted through modern diffusion priors and video world models. Interestingly, we first posted the paper on arXiv in Feb 2025, when action-conditioned video world models for planning were still rare—now this direction is rapidly gaining traction. Still many open questions, e.g., • how to avoid local minima in planning • what representations work best for world models • how to balance physics priors vs. data-driven learning Paper:

Heng Yang

18,994 views • 5 months ago

Voicebox: Text-Guided Multilingual Universal Speech Generation at Scale blog: Large-scale generative models such as GPT and DALL-E have revolutionized natural language processing and computer vision research. These models not only generate high fidelity text or image outputs, but are also generalists which can solve tasks not explicitly taught. In contrast, speech generative models are still primitive in terms of scale and task generalization. In this paper, we present Voicebox, the most versatile text-guided generative model for speech at scale. Voicebox is a non-autoregressive flow-matching model trained to infill speech, given audio context and text, trained on over 50K hours of speech that are neither filtered nor enhanced. Similar to GPT, Voicebox can perform many different tasks through in-context learning, but is more flexible as it can also condition on future context. Voicebox can be used for mono or cross-lingual zero-shot text-to-speech synthesis, noise removal, content editing, style conversion, and diverse sample generation. In particular, Voicebox outperforms the state-of-the-art zero-shot TTS model VALL-E on both intelligibility (5.9% vs 1.9% word error rates) and audio similarity (0.580 vs 0.681) while being up to 20 times faster.
2:43

Voicebox: Text-Guided Multilingual Universal Speech Generation at Scale blog: Large-scale generative models such as GPT and DALL-E have revolutionized natural language processing and computer vision research. These models not only generate high fidelity text or image outputs, but are also generalists which can solve tasks not explicitly taught. In contrast, speech generative models are still primitive in terms of scale and task generalization. In this paper, we present Voicebox, the most versatile text-guided generative model for speech at scale. Voicebox is a non-autoregressive flow-matching model trained to infill speech, given audio context and text, trained on over 50K hours of speech that are neither filtered nor enhanced. Similar to GPT, Voicebox can perform many different tasks through in-context learning, but is more flexible as it can also condition on future context. Voicebox can be used for mono or cross-lingual zero-shot text-to-speech synthesis, noise removal, content editing, style conversion, and diverse sample generation. In particular, Voicebox outperforms the state-of-the-art zero-shot TTS model VALL-E on both intelligibility (5.9% vs 1.9% word error rates) and audio similarity (0.580 vs 0.681) while being up to 20 times faster.

AK

429,355 views • 3 years ago

The power of generative models — now embodied in humanoids. Announcing DreamControl –– After a year-long research effort at General Robotics — we present a scalable framework for whole-body humanoid control that fuses diffusion priors with reinforcement learning to unlock real-world scene interaction. Diffusion + RL → natural whole-body skills on real robots. DreamControl enables humanoids to move beyond locomotion demos → performing natural, human-like skills such as –– Picking & lifting objects, Opening drawers & doors, Precise punching, kicking, and jumping, Bimanual manipulation tasks Our key innovation: a diffusion prior over human motion that guides RL, eliminating the need for massive teleoperation datasets, and producing motions that look human while transferring to real hardware. Trained purely in simulation, deployed on the Unitree G1 humanoid, DreamControl policies run in real time, bridging sim-to-real with unprecedented naturalness. We leverage a novel hybrid edge + cloud infrastructure that runs RL-trained policies on the edge backed by powerful AI models running in the cloud This is the next step in General Robotics’ journey toward general-purpose humanoid assistants that interact, adapt, and assist autonomously. Paper: Blog: 1/n
1:54

The power of generative models — now embodied in humanoids. Announcing DreamControl –– After a year-long research effort at General Robotics — we present a scalable framework for whole-body humanoid control that fuses diffusion priors with reinforcement learning to unlock real-world scene interaction. Diffusion + RL → natural whole-body skills on real robots. DreamControl enables humanoids to move beyond locomotion demos → performing natural, human-like skills such as –– Picking & lifting objects, Opening drawers & doors, Precise punching, kicking, and jumping, Bimanual manipulation tasks Our key innovation: a diffusion prior over human motion that guides RL, eliminating the need for massive teleoperation datasets, and producing motions that look human while transferring to real hardware. Trained purely in simulation, deployed on the Unitree G1 humanoid, DreamControl policies run in real time, bridging sim-to-real with unprecedented naturalness. We leverage a novel hybrid edge + cloud infrastructure that runs RL-trained policies on the edge backed by powerful AI models running in the cloud This is the next step in General Robotics’ journey toward general-purpose humanoid assistants that interact, adapt, and assist autonomously. Paper: Blog: 1/n

Ashish Kapoor

118,133 views • 11 months ago

Everything you love about generative models — now powered by real physics! Announcing the Genesis project — after a 24-month large-scale research collaboration involving over 20 research labs — a generative physics engine able to generate 4D dynamical worlds powered by a physics simulation platform designed for general-purpose robotics and physical AI applications. Genesis's physics engine is developed in pure Python, while being 10-80x faster than existing GPU-accelerated stacks like Isaac Gym and MJX. It delivers a simulation speed ~430,000 faster than in real-time, and takes only 26 seconds to train a robotic locomotion policy transferrable to the real world on a single RTX4090 (see tutorial: The Genesis physics engine and simulation platform is fully open source at We'll gradually roll out access to our generative framework in the near future. Genesis implements a unified simulation framework all from scratch, integrating a wide spectrum of state-of-the-art physics solvers, allowing simulation of the whole physical world in a virtual realm with the highest realism. We aim to build a universal data engine that leverages an upper-level generative framework to autonomously create physical worlds, together with various modes of data, including environments, camera motions, robotic task proposals, reward functions, robot policies, character motions, fully interactive 3D scenes, open-world articulated assets, and more, aiming towards fully automated data generation for robotics, physical AI and other applications. Open Source Code: Project webpage: Documentation: 1/n
3:26

Everything you love about generative models — now powered by real physics! Announcing the Genesis project — after a 24-month large-scale research collaboration involving over 20 research labs — a generative physics engine able to generate 4D dynamical worlds powered by a physics simulation platform designed for general-purpose robotics and physical AI applications. Genesis's physics engine is developed in pure Python, while being 10-80x faster than existing GPU-accelerated stacks like Isaac Gym and MJX. It delivers a simulation speed ~430,000 faster than in real-time, and takes only 26 seconds to train a robotic locomotion policy transferrable to the real world on a single RTX4090 (see tutorial: The Genesis physics engine and simulation platform is fully open source at We'll gradually roll out access to our generative framework in the near future. Genesis implements a unified simulation framework all from scratch, integrating a wide spectrum of state-of-the-art physics solvers, allowing simulation of the whole physical world in a virtual realm with the highest realism. We aim to build a universal data engine that leverages an upper-level generative framework to autonomously create physical worlds, together with various modes of data, including environments, camera motions, robotic task proposals, reward functions, robot policies, character motions, fully interactive 3D scenes, open-world articulated assets, and more, aiming towards fully automated data generation for robotics, physical AI and other applications. Open Source Code: Project webpage: Documentation: 1/n

Zhou Xian

3,819,325 views • 1 year ago

We RL'ed humanoid robots to Cristiano Ronaldo, LeBron James, and Kobe Byrant! These are neural nets running on real hardware at our GEAR lab. Most robot demos you see online speed videos up. We actually *slow them down* so you can enjoy the fluid motions. I'm excited to announce "ASAP", a "real2sim2real" model that masters extremely smooth and dynamic motions for humanoid whole body control. We pretrain the robot in simulation first, but there is a notorious "sim2real" gap: it's very difficult for hand-engineered physics equations to match real world dynamics. Our fix is simple: just deploy a pretrained policy on real hardware, collect data, and replay the motion in sim. The replay will obviously have many errors, but that gives a rich signal to compensate for the physics discrepancy. Use another neural net to learn the delta. Basically, we "patch up" a traditional physics engine, so that the robot can experience almost the real world at scale in GPUs. The future is hybrid simulation: combine the power of classical sim engines refined over decades and the uncanny ability of modern NNs to capture a messy world.
0:50

We RL'ed humanoid robots to Cristiano Ronaldo, LeBron James, and Kobe Byrant! These are neural nets running on real hardware at our GEAR lab. Most robot demos you see online speed videos up. We actually *slow them down* so you can enjoy the fluid motions. I'm excited to announce "ASAP", a "real2sim2real" model that masters extremely smooth and dynamic motions for humanoid whole body control. We pretrain the robot in simulation first, but there is a notorious "sim2real" gap: it's very difficult for hand-engineered physics equations to match real world dynamics. Our fix is simple: just deploy a pretrained policy on real hardware, collect data, and replay the motion in sim. The replay will obviously have many errors, but that gives a rich signal to compensate for the physics discrepancy. Use another neural net to learn the delta. Basically, we "patch up" a traditional physics engine, so that the robot can experience almost the real world at scale in GPUs. The future is hybrid simulation: combine the power of classical sim engines refined over decades and the uncanny ability of modern NNs to capture a messy world.

Jim Fan

566,544 views • 1 year ago

This is how ALOHA's "teleoperation" system works - a fancy word for "remote control". Training robots will be more and more like playing games in the physical world. A human operates a "joystick++" to perform tasks and collect data, or intervene if there's any safety concern. There's actually a learning curve to master the controller, much like practicing gaming skills. Teleoperation can be done in many different ways. ALOHA is an impressive custom-built system with very low cost. Here're a few alternatives: (1) Motion Capture (MoCap): apply the MoCap systems used for Hollywood movies to capture the fine-grained motions of hand joints. There would be no "embodiment gap" if the robot hand has 5 fingers. For instance, a demonstrator can wear a CyberGlove ( and manipulate the objects. CyberGlove will capture the motion signals & haptic feedback in real-time, which can be re-targeted onto the humanoid. (2) Wearing gloves & markers can be clumsy. An alternative way to do MoCap is through computer vision. DexPilot from NVIDIA enables marker-less and glove-free data collection. The human operator simply uses their bare hands to perform the tasks. 4 Intel RealSense depth cameras and 2 NVIDIA Titan XP GPUs (yeah, 2019 work) translate the pixels to precise motion signals for robot learning. (3) VR Headset: turn the training room into a VR game and "role play" the robot. This has the advantage of scalable remote data collection - annotators from around the world can contribute without coming onsite. VR demonstration technique appeared in research projects like the iGibson home robot simulator, an initiative that I participated in at Stanford: Behind-the-scene video by Litian Liang
0:52

This is how ALOHA's "teleoperation" system works - a fancy word for "remote control". Training robots will be more and more like playing games in the physical world. A human operates a "joystick++" to perform tasks and collect data, or intervene if there's any safety concern. There's actually a learning curve to master the controller, much like practicing gaming skills. Teleoperation can be done in many different ways. ALOHA is an impressive custom-built system with very low cost. Here're a few alternatives: (1) Motion Capture (MoCap): apply the MoCap systems used for Hollywood movies to capture the fine-grained motions of hand joints. There would be no "embodiment gap" if the robot hand has 5 fingers. For instance, a demonstrator can wear a CyberGlove ( and manipulate the objects. CyberGlove will capture the motion signals & haptic feedback in real-time, which can be re-targeted onto the humanoid. (2) Wearing gloves & markers can be clumsy. An alternative way to do MoCap is through computer vision. DexPilot from NVIDIA enables marker-less and glove-free data collection. The human operator simply uses their bare hands to perform the tasks. 4 Intel RealSense depth cameras and 2 NVIDIA Titan XP GPUs (yeah, 2019 work) translate the pixels to precise motion signals for robot learning. (3) VR Headset: turn the training room into a VR game and "role play" the robot. This has the advantage of scalable remote data collection - annotators from around the world can contribute without coming onsite. VR demonstration technique appeared in research projects like the iGibson home robot simulator, an initiative that I participated in at Stanford: Behind-the-scene video by Litian Liang

Jim Fan

124,588 views • 2 years ago

We’re excited to finally introduce Kled Special Tasks, the final major feature included in the V2 app update. Users will now have access to a fully interactive terminal where they can view and complete domain specific upload tasks directly from enterprise buyers. These tasks can be region locked and person specific. For example, PhD students at Stanford might be prompted to upload their coursework or research materials and get paid for it. Our first domain specific task will focus on homework collection from high school and college students across Europe and the United States. Students will verify their emails and academic credentials directly within the app. We’ve built labeling workflows to ensure all uploaded content meets our criteria, and participants will receive weighted payouts based on the value of their submissions. We’ve already built a network of over 3,800 students from Stanford, MIT, UIUC, Rutgers, and Duke who will be actively onboarded to contribute content. Kled will work hand in hand with our research division, HADES, to justify the large scale purchase of this homework content. Several enterprise buyers have already expressed interest, each confirming that academic data from students represents a growing multi year industry requiring a continuous flow of fresh material. Kled Special Tasks also gives us the ability to internally identify valuable content types, issue calls for specific datasets, and collect 1,000-2,000 unique samples per task. We can then package these datasets into specialized data packs that our sales team will use to pitch directly to AI labs and enterprise clients with matching data needs. This will be one of our most powerful tools for expanding Kled’s buyer network. All of this will be fully available in the V2 update. We’re excited to show just how advanced our segmentation and data validation software has become as we bring this release to market.
0:44

We’re excited to finally introduce Kled Special Tasks, the final major feature included in the V2 app update. Users will now have access to a fully interactive terminal where they can view and complete domain specific upload tasks directly from enterprise buyers. These tasks can be region locked and person specific. For example, PhD students at Stanford might be prompted to upload their coursework or research materials and get paid for it. Our first domain specific task will focus on homework collection from high school and college students across Europe and the United States. Students will verify their emails and academic credentials directly within the app. We’ve built labeling workflows to ensure all uploaded content meets our criteria, and participants will receive weighted payouts based on the value of their submissions. We’ve already built a network of over 3,800 students from Stanford, MIT, UIUC, Rutgers, and Duke who will be actively onboarded to contribute content. Kled will work hand in hand with our research division, HADES, to justify the large scale purchase of this homework content. Several enterprise buyers have already expressed interest, each confirming that academic data from students represents a growing multi year industry requiring a continuous flow of fresh material. Kled Special Tasks also gives us the ability to internally identify valuable content types, issue calls for specific datasets, and collect 1,000-2,000 unique samples per task. We can then package these datasets into specialized data packs that our sales team will use to pitch directly to AI labs and enterprise clients with matching data needs. This will be one of our most powerful tools for expanding Kled’s buyer network. All of this will be fully available in the V2 update. We’re excited to show just how advanced our segmentation and data validation software has become as we bring this release to market.

Kled AI

74,618 views • 9 months ago

Let's reverse engineer this demo. You need 3 things: (1) robust hardware and motor designs that treat simulation as first-class citizen; (2) a human motion capture ("mocap") dataset, such as those for film and gaming characters; (3) massively parallel RL training in GPU-accelerated simulation. Last October, our team trained a 1.5M parameter foundation model called HOVER for such agile motor control. It follows this recipe, roughly speaking (details in thread): (1) Simulation used to be an after-thought. Now, it has to be part of the hardware design process. If your robot doesn't simulate well, you can kiss RL goodbye. Hardware-simulation co-design is a very interesting emergent topic that only becomes meaningful with today's compute capability. (2) Human mocap dataset to produce natural-looking walking and running gaits. That's one huge advantage of using humanoid robot - you get to imitate from tons of human motions that were originally captured for movies or AAA games. At least 3 ways to use the data: - For initialization: pre-train the neural net to imitate human, and then finetune it into the robot form factor with physics turned on; - For reward function: penalize any deviations from the target pose; - For representation learning: treat the human poses as a "motion prior" to constrain the space of robot behaviors. (3) Shove the above into Isaac sim, add a lot of randomization, pump it through PPO, throw in a bunch of GPUs, and then watch Netflix till loss converges. If you have an urge to comment this is CGI, let me save you a few keystrokes — many academic labs now own the G1 robot in the flesh. See our team's HOVER work in the thread: 🧵
1:04

Let's reverse engineer this demo. You need 3 things: (1) robust hardware and motor designs that treat simulation as first-class citizen; (2) a human motion capture ("mocap") dataset, such as those for film and gaming characters; (3) massively parallel RL training in GPU-accelerated simulation. Last October, our team trained a 1.5M parameter foundation model called HOVER for such agile motor control. It follows this recipe, roughly speaking (details in thread): (1) Simulation used to be an after-thought. Now, it has to be part of the hardware design process. If your robot doesn't simulate well, you can kiss RL goodbye. Hardware-simulation co-design is a very interesting emergent topic that only becomes meaningful with today's compute capability. (2) Human mocap dataset to produce natural-looking walking and running gaits. That's one huge advantage of using humanoid robot - you get to imitate from tons of human motions that were originally captured for movies or AAA games. At least 3 ways to use the data: - For initialization: pre-train the neural net to imitate human, and then finetune it into the robot form factor with physics turned on; - For reward function: penalize any deviations from the target pose; - For representation learning: treat the human poses as a "motion prior" to constrain the space of robot behaviors. (3) Shove the above into Isaac sim, add a lot of randomization, pump it through PPO, throw in a bunch of GPUs, and then watch Netflix till loss converges. If you have an urge to comment this is CGI, let me save you a few keystrokes — many academic labs now own the G1 robot in the flesh. See our team's HOVER work in the thread: 🧵

Jim Fan

216,139 views • 1 year ago

Can GPT-4 teach a robot hand to do pen spinning tricks better than you do? I'm excited to announce Eureka, an open-ended agent that designs reward functions for robot dexterity at super-human level. It’s like Voyager in the space of a physics simulator API! Eureka bridges the gap between high-level reasoning (coding) and low-level motor control. It is a “hybrid-gradient architecture”: a black box, inference-only LLM instructs a white box, learnable neural network. The outer loop runs GPT-4 to refine the reward function (gradient-free), while the inner loop runs reinforcement learning to train a robot controller (gradient-based). We are able to scale up Eureka thanks to IsaacGym, a GPU-accelerated physics simulator that speeds up reality by 1000x. On a benchmark suite of 29 tasks across 10 robots, Eureka rewards outperform expert human-written ones on 83% of the tasks by 52% improvement margin on average. We are surprised that Eureka is able to learn pen spinning tricks, which are very difficult even for CGI artists to animate frame by frame! Eureka also enables a new form of in-context RLHF, which is able to incorporate a human operator’s feedback in natural language to steer and align the reward functions. It can serve as a powerful co-pilot for robot engineers to design sophisticated motor behaviors. As usual, we open-source everything! Welcome you all to check out our video gallery and try the codebase today: Paper: Code: Deep dive with me: 🧵
0:24

Can GPT-4 teach a robot hand to do pen spinning tricks better than you do? I'm excited to announce Eureka, an open-ended agent that designs reward functions for robot dexterity at super-human level. It’s like Voyager in the space of a physics simulator API! Eureka bridges the gap between high-level reasoning (coding) and low-level motor control. It is a “hybrid-gradient architecture”: a black box, inference-only LLM instructs a white box, learnable neural network. The outer loop runs GPT-4 to refine the reward function (gradient-free), while the inner loop runs reinforcement learning to train a robot controller (gradient-based). We are able to scale up Eureka thanks to IsaacGym, a GPU-accelerated physics simulator that speeds up reality by 1000x. On a benchmark suite of 29 tasks across 10 robots, Eureka rewards outperform expert human-written ones on 83% of the tasks by 52% improvement margin on average. We are surprised that Eureka is able to learn pen spinning tricks, which are very difficult even for CGI artists to animate frame by frame! Eureka also enables a new form of in-context RLHF, which is able to incorporate a human operator’s feedback in natural language to steer and align the reward functions. It can serve as a powerful co-pilot for robot engineers to design sophisticated motor behaviors. As usual, we open-source everything! Welcome you all to check out our video gallery and try the codebase today: Paper: Code: Deep dive with me: 🧵

Jim Fan

2,674,348 views • 2 years ago