Exploring robotics｜Open to collaborations — DMs open｜ Building autonomous AI co-scientists｜ 👩💻 SDE 💬 CN, EN, KR 🌏 UTC+8｜ opinions are yours
Shenzhen, Guangdong, China
Joined October 2022
- Tweets5.9K
- Following2.2K
- Followers1.1K
- Likes9.7K
Pinned Tweet
Wow, my first post to hit 2k likes! 🥹
Thank you all so much!!🥹
Really happy to see Motion2MixamoRig pass 200 stars on GitHub too! ⭐️
I've noticed a lot of new people following me after that post, so I thought it'd be nice to reintroduce myself:
I'm a software engineer based in Shenzhen, China. During the day, I build software for medical devices. Outside of work, I explore Autonomous AI4S and robotics.
I'll mostly post about what I'm building, what I wanna find out, and sometimes life in Shenzhen.
Anyway, feel free to hit me up!
Always down to chat or build something together :)
github.com/MaxLiu22/Motion2M…
Made an open-source one that does the same thing, and there's a Blender extension available for it now too.
MaxLiu09 retweeted
cooking up some more sim ready articulated assets
fully functional rack in Isaac, xArm7 wiring in deformable wires, with feedback from actual state machines, task success a function of the asset itself.
1/1000000
wow
We’re excited to announce our new CapCut plugin is coming soon to @ChatGPT at CapCutWorldLA.
The plugin brings CapCut's video editing and generation tools alongside ChatGPT. You can start with an idea in ChatGPT, bring it to life with the CapCut plugin right inside ChatGPT, then take your draft video to CapCut's web editor to refine, share and publish.
By bringing the CapCut product you already love and use in ChatGPT, we're making content creation more seamless and efficient.
MaxLiu09 retweeted
That's genuinely insane! 🤯
An AI agent built a physics twin of a robot task from a phone video in under an hour.
This project RoboSimStudio was built on Newton by researchers from UC Berkeley and NVIDIA Robotics.
Open toolkit, every physics parameter exposed as something a human can slide or an agent can edit.
The demo is a bimanual setup, one arm holding a bar, the other wiping a marker stripe off it with a foam sponge. Input was a smartphone video.
What's more important no CAD, and no calibrated cameras. Claude and GPT-6 Astra each built a simulation of it independently. One finished in 56 minutes.
Then they ran it on the real robot and the bar support moved. In simulation it had been modelled as fixed, so the sponge kept contact. In reality it shifted and contact was lost.
The agent went back and replaced the fixed constraint with a free rigid body. The revised simulation then reproduced the tipping the original could not.
That is the loop actually closing. The real world found the modelling error and the fix went back into the twin.
🔗 Check the project here: lnkd.in/dgXJxAhw
~~
♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com
MaxLiu09 retweeted
• 论文标题：Why is AI-generated language so hard to understand? (NYU & UT Austin)
• arXiv 链接：arxiv.org/abs/2610.29162
• PDF 原文：arxiv.org/pdf/2610.29162
MaxLiu09 retweeted
This navigation paper visualised: "Navigating the Crowd: Non-linear MPC with Social Forces Dynamics for Human-Aware Robot Navigation" (Trepella et al., 2026) puts a model of the crowd inside the robot's planner.
- Each person is pulled toward their goal and pushed away from anyone too close
- Predict everyone 2 s ahead, including how they react to the robot
- Take the path that stays out of everyone's personal space
- Replan 20 times a second
MaxLiu09 retweeted
Just 4 days after SuperMimic, MorphMimic from Caltech extends it to whole-body motion tracking across several humanoids! What a pace of progress!
One world model and one policy are shared by Unitree G1, AgiBot X2 Ultra and Booster T2, conditioned on each robot's actuation, geometry and mass. In simulation, the shared model tracks about as well as per-robot specialists. On a real G1, it dances, walks and follows a VR teleoperator.
Closest to MorphMimic, X-WBC (arxiv.org/abs/2609.15213) trains one whole-body controller across nine simulated and four real humanoids, but gives each robot its own input and output modules. MorphMimic shares every weight, world model included.
For contrast, @NicoBohlinger et al. trained one locomotion policy on 50 legged robots with different joint layouts, which transfers zero-shot to unseen real robots (now scaled up in γ₀: gamma-zero.com). However, MorphMimic trains a whole-body tracker, which does far more than walking, though all robots must adhere to the same 29-joint layout, and the method has not been tested on unseen robots yet.
The next step is inviting itself: MorphMimic across different joint layouts 🚀
One world, one policy.
MorphMimic shares a morphology-conditioned world model and policy across #Unitree G1, #AgiBot X2 Ultra, and #Booster T2—from the same dance in MuJoCo to a live browser demo and Unitree G1 hardware, with more to come. Code and report at lzyang2000.github.io/morphmi…
Huge thanks to the SuperMimic team for the inspiration!
MaxLiu09 retweeted
同僚が週末に試してみたという Claude Code（Opus 5.5）とSimulink Agentic Toolkit による Tesla Roadster シミュレーション
Simscape Fluids でモデル化したスラスター、エアタンク、吸引ファン。Stateflow によるモード切り替えロジックと、スラスターを制御するモデル予測制御を実装したそうな🧐
Wow, thank you so much for using Motion2MixamoRig in your 3D character creation workflow and sharing the entire process on YouTube!
Btw, about that "One thing about Motion2MixamoRig is made by a guy who definitely uses MacOS" comment in the yt video at 3:52...
You got the MacOS part right. 😂
Plot twist: I'm actually a gal. 🙋♀️
Really appreciate you taking the time to test the tool and share both what worked well and what could be improved. I've taken note of the issues you pointed out, and I'll keep working on making Motion2MixamoRig better.
I've also noticed more open-source tools emerging in the community, some building on or taking inspiration from Motion2MixamoRig. It's really exciting to see!
At the end of the day, I hope this project can help make motion retargeting more accessible and affordable for everyone.
Thanks again for sharing your experience! 🙌
Guys, you can now turn any video into Mixamo animation, free and local 🔥
It's called Motion2MixamoRig. I tested it on a full workflow, from one anime image to an animated 3D character, and it's really good.
You still need some cleanup on top, but overall it works great.
youtu.be/58UpDAnbdkM
MaxLiu09 retweeted
お家作ってみた
NvidiaがHuggingfaceに公開したSimreadyデータを早速シミュレーションに使ってみた。
huggingface.co/spaces/nvidia…
見た目リアルでいい感じ。
Motion2MixamoRig v0.1.8 is out! 🚀
github.com/MaxLiu22/Motion2M…
A few updates this time:
🔍 New Diagnostics tool: auto-detect motion issues, manually mark bad segments, and export JSON notes for AI-assisted fixes.
🦶 Improved turning stability and added foot locking to reduce sliding.
🖐️ Experimental finger tracking with HaMeR.
🧩 Added a pluggable motion extractor interface (--set-extractor). Only GVHMR is supported for now, but I'd love to see people try different models and settings to compare real-world results!
👩💻 P.S.
I read as much feedback about Motion2MixamoRig as I can find, in any language.
So please try it out and share your bugs, ideas, or feature requests!
I'll keep improving it at my own pace, based on your feedback. And who knows? Your suggestion might make it into the next release!
Ofc, I'll credit you in the changelog. :)
🍻 Enjoy!
The floor markers are now included in the new Diagnostics feature in Motion2MixamoRig v0.1.8 as well!
🔋You'll see them in both Blender and the exported HTML report.
👣Foot sliding is still something I'm working on. In the meantime, I want to give animators and game developers better tools to spot issues, review the results, and make their own adjustments.
🙏 Hopefully this makes Motion2MixamoRig a bit more useful in real production workflows!