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.

I put motion-bricks.cpp up on GitHub! It generates animations from key-frames in realtime using a tiny model that easily runs on a desktop CPU and Vulkan. There are still bugs to iron out and I'm thinking this can be combined with physical simulation and robotic control policies...

Richard Palethorpe

1,697 subscribers

59,244 views • 9 days ago •via X (Twitter)

Science & Technology Education

Anya Rossi• Live Now

Private livecam show

26 Comments

Richard Palethorpe9 days ago

Also checkout my livestreams on this:

Milo9 days ago

They look very similar; is there a connection between them?

Richard Palethorpe9 days ago

Yes that's the original in PyTorch

PCI Express9 days ago

I'm curious how this will be used in games, like I still don't understand how people chain different animations and make them fluid like we can see in Arc Raiders. Weighted and grounded movements.

Richard Palethorpe9 days ago

I have no idea how current games do it, but this is designed to be used with robots that are constrained by physics and have a control policy (another neural net that is trained with RL to keep the robot upright) that will attempt to stabalise the robot while MotionBricks provides movement targets. So in a game you could physically simulate a character with a control policy to balance them, then use motion bricks to generate target movements. At the end you only get physically possible movements

PCI Express9 days ago

That’s really interesting because motion bricks and kimodo are pretty similar, except one is text to motion. OwO

Broadstreet8 days ago

Awesome for animators and game creators

Fabio Angela8 days ago

Cool thanks, I'll take a look, I'm working on a pipeline where I want to generate poses with openai Imagine2 and I extract keypoints with mediapipe or similar and map to a skeleton to have the pose I can place on a 3d canvas, this may be the way

Jonathan Steinberg9 days ago

Nice!

Milo9 days ago

Please attach the head to it.😱

Richard Palethorpe9 days ago

It's the skeleton for the Unitree G1 which has a head but I guess it's static

Milo9 days ago

Got it, it's great.

xenstudio9 days ago

Pretty badass dude. Cool beans!

𝗘𝗜𝗞𝗢8 days ago

Seems like Cascadeur!

AINZX9 days ago

Bro how are u achiving it its going beyond my head

AI Mastery Guide9 days ago

Runs on CPU, that's wild

Simeon Nedelchev9 days ago

Looks impressive! BTW if you are intetested in neural character animation check out the "Control Operators for Interactive Character Animation" and "Learned Motion Matching"

PCI Express9 days ago

@taksasDESUYO い

Kevin Rajan9 days ago

wtf that's so cool, how close is this to the pareto-frontier? or is it just an experiment?

Starkium9 days ago

might play with this in my game engine / project

Leon S.Ch9 days ago

两个人互动可以做吗？

Richard Palethorpe9 days ago

I'm not sure that translates correctly, but probably the answer is no

namaissur8 days ago

@NickChapmn when in @SubstrataVr ?

₭URO𓃵9 days ago

Hey dude:)) this is amazing, can u send me dm? i can´t

namaissur8 days ago

looks cool! thanx for sharing!

Christian Jäger8 days ago

Nice work! I had the same idea what are your future plans? I'm still trying to learn more about motion bricks. uncharted territory :D

Related Videos

This model lets you put in up to 100 images and generates a physically consistent simulation of a world that you can explore from any angle with no drifting. We might not be in a simulation, but we're about to put a lot of AIs into one.

This model lets you put in up to 100 images and generates a physically consistent simulation of a world that you can explore from any angle with no drifting. We might not be in a simulation, but we're about to put a lot of AIs into one.

Alix Ollivier

11,063 views • 10 days ago

In Figma motion, you can easily nest your animation in a frame, and use that parent frame as a camera with position, scale, and rotation. Also J and K will move the play head to the next and previous visible key frames. #figmatip Who wants a full tutorial on this concept?

In Figma motion, you can easily nest your animation in a frame, and use that parent frame as a camera with position, scale, and rotation. Also J and K will move the play head to the next and previous visible key frames. #figmatip Who wants a full tutorial on this concept?

miggi from figgi

32,621 views • 1 month ago

hi all, excited to join! i'm building an expressive mini "shoggoth" robot which will eventually be hooked up to gpt4o realtime voice. i'm currently working on the low-level policies, which are trained in a mujoco simulation with RL. to delay working on raw-pixels for now, i trained a pose-estimation model using deeplabcut and triangulate the position in 3d space using the stereo cameras. eventually, i'll use gpt4o's tool calling capabilities to activate several of these policies (closed and open loop) based on the dialog flow! captions: manual actuation of the tentacle / 3d pose estimation / target design

hi all, excited to join! i'm building an expressive mini "shoggoth" robot which will eventually be hooked up to gpt4o realtime voice. i'm currently working on the low-level policies, which are trained in a mujoco simulation with RL. to delay working on raw-pixels for now, i trained a pose-estimation model using deeplabcut and triangulate the position in 3d space using the stereo cameras. eventually, i'll use gpt4o's tool calling capabilities to activate several of these policies (closed and open loop) based on the dialog flow! captions: manual actuation of the tentacle / 3d pose estimation / target design

Matthieu LC

45,587 views • 1 year ago

skin-tokens.cpp is up on GitHub! This let's you generate a skeleton and skin weights for humanoid and non-humanoid meshes. I tried combining this with skeletal animations from kimodo.cpp, but still having issues there as can be seen at the end of the video

skin-tokens.cpp is up on GitHub! This let's you generate a skeleton and skin weights for humanoid and non-humanoid meshes. I tried combining this with skeletal animations from kimodo.cpp, but still having issues there as can be seen at the end of the video

Richard Palethorpe

27,308 views • 12 days ago

Mighty Camera runs VIO on-device in a tiny package. But for it to be useful, you need things like mapping (and later occupancy, loop closure etc). Here is a demo of lightweight mapping which uses VIO pose from Mighty and generates a semi-dense map on host-side in realtime. It’s early but this will be part of the SDK along with other goodies.

Mighty Camera runs VIO on-device in a tiny package. But for it to be useful, you need things like mapping (and later occupancy, loop closure etc). Here is a demo of lightweight mapping which uses VIO pose from Mighty and generates a semi-dense map on host-side in realtime. It’s early but this will be part of the SDK along with other goodies.

Asad Memon

21,285 views • 3 months ago

AI for robotics has some monumental challenges, like high dimensionality and reasoning with the physical world. Ashok Elluswamy, AI lead at Tesla, thinks the hardest problem is Evaluation. - Loss isn't a perfect indicator of Policy Neural Network's quality. - Testing in the real world for all the long-tail failure cases is not possible. To solve the evaluation challenge, Tesla has developed a World Simulator Model. It is trained on state and action pairs, so for a given current camera feed and current action, the model generates the next frame in the cameras. Once the world simulation is good enough, it can be hooked up to the policy network for a closed-loop simulation. In this setup, the model generates the next video frames, and based on those, the policy network comes up with the next action, which feeds back into the world simulation; this continuous loop creates a coherent simulation. The world simulation neural network is not just applicable to self-driving but also to humanoids, because Tesla trains it on common data. The examples show Optimus navigating or manipulating inside the world simulation, with all the pixels being generated by the world simulation model. The robotic agents can be tested in an accurately reproduced edge case and in other potential variations of the same edge case.

AI for robotics has some monumental challenges, like high dimensionality and reasoning with the physical world. Ashok Elluswamy, AI lead at Tesla, thinks the hardest problem is Evaluation. - Loss isn't a perfect indicator of Policy Neural Network's quality. - Testing in the real world for all the long-tail failure cases is not possible. To solve the evaluation challenge, Tesla has developed a World Simulator Model. It is trained on state and action pairs, so for a given current camera feed and current action, the model generates the next frame in the cameras. Once the world simulation is good enough, it can be hooked up to the policy network for a closed-loop simulation. In this setup, the model generates the next video frames, and based on those, the policy network comes up with the next action, which feeds back into the world simulation; this continuous loop creates a coherent simulation. The world simulation neural network is not just applicable to self-driving but also to humanoids, because Tesla trains it on common data. The examples show Optimus navigating or manipulating inside the world simulation, with all the pixels being generated by the world simulation model. The robotic agents can be tested in an accurately reproduced edge case and in other potential variations of the same edge case.

The Humanoid Hub

17,065 views • 7 months ago

Wow. This is crazy. A developer trained an AI agent in simulation and deployed it onto a real robotic air hockey table using reinforcement learning. This robot can track the puck with millimeter-level accuracy and react in roughly 20 milliseconds, fast enough to challenge even skilled human players. We’re moving from robots that follow programmed rules to machines that learn strategies in simulation and execute them in the physical world.

Wow. This is crazy. A developer trained an AI agent in simulation and deployed it onto a real robotic air hockey table using reinforcement learning. This robot can track the puck with millimeter-level accuracy and react in roughly 20 milliseconds, fast enough to challenge even skilled human players. We’re moving from robots that follow programmed rules to machines that learn strategies in simulation and execute them in the physical world.

SciTech Era

1,585,398 views • 3 months ago

I built a Claude skill that makes motion-style animation. > Paste a reference image > It generates frames with Nano Banana 2 > and animates the video with Seedance 2.0 Everything runs in Claude through the Arcads MCP Comment "Motion", and I'll send you the skill

I built a Claude skill that makes motion-style animation. > Paste a reference image > It generates frames with Nano Banana 2 > and animates the video with Seedance 2.0 Everything runs in Claude through the Arcads MCP Comment "Motion", and I'll send you the skill

Romain Torres

418,317 views • 2 months ago

Claude Code on desktop just added an embedded loop that can run your dev server, view the app, review local changes, and watch pull requests until they merge. It can now preview a running app inside the desktop UI while it iterates on fixes. It can also monitor CI for an open PR and either fix failures or merge when checks go green. Earlier “agentic coding” workflows usually bounced between a terminal, a browser, and GitHub, so the model could not directly see what the app rendered or stay attached to CI results without manual babysitting. The new server preview flow starts and stops dev servers from a Preview control, shows the app in an embedded browser, reads logs, and can interact with the page to verify changes. That closes a common gap where a model edits code but cannot reliably confirm the UI and runtime behavior match the intent. The local “Review code” step runs before pushing, and it leaves inline comments on the diff so basic bugs and risky changes get caught earlier. For PR monitoring, the desktop app will track status and CI checks, and it can use auto-fix to attempt repairs and auto-merge to land the PR once checks pass, using GitHub tooling under the hood. Session mobility adds a /desktop command to move a CLI session into the desktop app, and sessions can be pushed to the cloud to resume from web or phone.

Claude Code on desktop just added an embedded loop that can run your dev server, view the app, review local changes, and watch pull requests until they merge. It can now preview a running app inside the desktop UI while it iterates on fixes. It can also monitor CI for an open PR and either fix failures or merge when checks go green. Earlier “agentic coding” workflows usually bounced between a terminal, a browser, and GitHub, so the model could not directly see what the app rendered or stay attached to CI results without manual babysitting. The new server preview flow starts and stops dev servers from a Preview control, shows the app in an embedded browser, reads logs, and can interact with the page to verify changes. That closes a common gap where a model edits code but cannot reliably confirm the UI and runtime behavior match the intent. The local “Review code” step runs before pushing, and it leaves inline comments on the diff so basic bugs and risky changes get caught earlier. For PR monitoring, the desktop app will track status and CI checks, and it can use auto-fix to attempt repairs and auto-merge to land the PR once checks pass, using GitHub tooling under the hood. Session mobility adds a /desktop command to move a CLI session into the desktop app, and sessions can be pushed to the cloud to resume from web or phone.

Rohan Paul

33,431 views • 6 months ago

I wrote some code! It renders frames to a virtual canvas and then distributes packets over WiFi to ESP32 modules that receive them to make up a larger display. So you could make a jumbotron out of a bunch of ESP32s and one PC. Code can be found on Github, and would love to get your suggestions on it!

I wrote some code! It renders frames to a virtual canvas and then distributes packets over WiFi to ESP32 modules that receive them to make up a larger display. So you could make a jumbotron out of a bunch of ESP32s and one PC. Code can be found on Github, and would love to get your suggestions on it!

Dave W Plummer

34,156 views • 1 year ago

I DON’T THINK PEOPLE REALIZE HOW MASSIVE THIS IS YET Hugging Face and Pollen Robotics just released Microduck. It is a new legged robot for $399. It passed $1 million in sales in under SEVEN hours 🤯 The hardware packs 15 actuators, a camera, and LiDAR into a 25 cm frame. It weighs under 800 grams and runs for about an hour on a single charge. Out of the box, it can > walk > roller-skate > pick up objects > recover from falls .. and so much more! You can drive the 7 shipped policies from the first boot using a gamepad. But the real hook is the training stack. The software is entirely open: → MuJoCo simulation and PPO training → 50 Hz onboard policy loop → Browser simulator included alongside hardware → Full control stack available on GitHub The mechanical and electronic design files are still proprietary. But don’t let the cute robotic duck fool you. Under the closed chassis is a FULLY OPEN-SOURCE, hackable brain you can experiment with and build on. Repo in 🧵↓

I DON’T THINK PEOPLE REALIZE HOW MASSIVE THIS IS YET Hugging Face and Pollen Robotics just released Microduck. It is a new legged robot for $399. It passed $1 million in sales in under SEVEN hours 🤯 The hardware packs 15 actuators, a camera, and LiDAR into a 25 cm frame. It weighs under 800 grams and runs for about an hour on a single charge. Out of the box, it can > walk > roller-skate > pick up objects > recover from falls .. and so much more! You can drive the 7 shipped policies from the first boot using a gamepad. But the real hook is the training stack. The software is entirely open: → MuJoCo simulation and PPO training → 50 Hz onboard policy loop → Browser simulator included alongside hardware → Full control stack available on GitHub The mechanical and electronic design files are still proprietary. But don’t let the cute robotic duck fool you. Under the closed chassis is a FULLY OPEN-SOURCE, hackable brain you can experiment with and build on. Repo in 🧵↓

Charly Wargnier ♨️

649,329 views • 12 days ago

I've been working on this KMP task tracking app that runs on Android, iOS, and Desktop, thanks to the power of Compose Multiplatform. It's a WIP, planning to add Web support. Check it out on GitHub: Playstore: #KMP #AndroidDev

I've been working on this KMP task tracking app that runs on Android, iOS, and Desktop, thanks to the power of Compose Multiplatform. It's a WIP, planning to add Web support. Check it out on GitHub: Playstore: #KMP #AndroidDev

Kanyi

14,904 views • 2 years ago

Image Georeferencing Using GeoLibre in the Browser Learn how to georeference images, scanned maps, and historical maps directly in your web browser using GeoLibre, with no desktop GIS software required. Video tutorial: In this tutorial, I demonstrate how to georeference a campus map by selecting ground control points, choosing a transformation method, evaluating residual errors, and aligning the image with a basemap. You'll also learn how to export the georeferenced result as a GeoTIFF, reload it into GeoLibre, compare it with the basemap, and share the finished map online. GeoLibre is a free and open-source, cloud-native GIS platform that runs in the browser, on desktop and mobile devices, and in Jupyter notebooks. Try it out: - Web App: - GitHub: #opensource #geospatial #GeoLibre

Image Georeferencing Using GeoLibre in the Browser Learn how to georeference images, scanned maps, and historical maps directly in your web browser using GeoLibre, with no desktop GIS software required. Video tutorial: In this tutorial, I demonstrate how to georeference a campus map by selecting ground control points, choosing a transformation method, evaluating residual errors, and aligning the image with a basemap. You'll also learn how to export the georeferenced result as a GeoTIFF, reload it into GeoLibre, compare it with the basemap, and share the finished map online. GeoLibre is a free and open-source, cloud-native GIS platform that runs in the browser, on desktop and mobile devices, and in Jupyter notebooks. Try it out: - Web App: - GitHub: #opensource #geospatial #GeoLibre

Qiusheng Wu

37,480 views • 5 days ago

Fortnite has improved the animations on ALL platforms! "Motion Matching and Procedural Layering are features that result in improved animations for things like transitioning from walking to running, changing directions, and using a weapon."

Fortnite has improved the animations on ALL platforms! "Motion Matching and Procedural Layering are features that result in improved animations for things like transitioning from walking to running, changing directions, and using a weapon."

Shiina

1,989,077 views • 2 years ago

Been using the new Claude 3.5 model as a tester and now that it is out, I can say it is very very impressive, and the "artifacts" that it generates are like a simpler version of Code Interpreter This is a real-time video of me creating a playable game and editing it with Claude

Been using the new Claude 3.5 model as a tester and now that it is out, I can say it is very very impressive, and the "artifacts" that it generates are like a simpler version of Code Interpreter This is a real-time video of me creating a playable game and editing it with Claude

Ethan Mollick

162,293 views • 2 years ago

You can run Google new Gemma 4 on mobile easily. I am using Gemma 4 version E2B on my Pixel 10 Pro. Here is all you need to do: - Go to the App Store and install Google AI Edge Gallery. If you already have it, just update it. - From there, you can install the model directly and start using it. Still exploring what this small but powerful model can do :)

You can run Google new Gemma 4 on mobile easily. I am using Gemma 4 version E2B on my Pixel 10 Pro. Here is all you need to do: - Go to the App Store and install Google AI Edge Gallery. If you already have it, just update it. - From there, you can install the model directly and start using it. Still exploring what this small but powerful model can do :)

AshutoshShrivastava

226,599 views • 5 months ago

This is a worlds first I believe.. Realtime sculpting and animation directly transformed to video. This is a Realtime ai "render", running on top of sculpting and animation I'm doing as I see the video playing. Done using the beta version of KREA AI's new Realtime video model ❤️

This is a worlds first I believe.. Realtime sculpting and animation directly transformed to video. This is a Realtime ai "render", running on top of sculpting and animation I'm doing as I see the video playing. Done using the beta version of KREA AI's new Realtime video model ❤️

Martin Nebelong

19,125 views • 1 year ago

Claude Code just got a lot easier to use 🙌 Just updated You can now install agents, commands, MCPs, and templates with a single command. Click on any card and it gives you the exact line to run. Everything is pulled from GitHub, the site runs on GitHub Pages, and downloads are logged with GitHub Actions.

Claude Code just got a lot easier to use 🙌 Just updated You can now install agents, commands, MCPs, and templates with a single command. Click on any card and it gives you the exact line to run. Everything is pulled from GitHub, the site runs on GitHub Pages, and downloads are logged with GitHub Actions.

Daniel San

135,292 views • 1 year ago