YouTube

[
  {
    "title": "World’s First General-Purpose Cerebellum GPT Foundation Model for Humanoid Robots: AstraBrain-WBC0.5",
    "id": "iNBvCLjOmVw",
    "url": "https://www.youtube.com/watch?v=iNBvCLjOmVw",
    "thumbnailUrl": "https://yt3.ggpht.com/ZxvDraFJbBvP9wJDmcVCuRqU9evkPCE3rIL9NzYPPQgzJugGySHi-_zQ9v8n9iyI1pSmalx3D7Y=s176-c-k-c0x00ffffff-no-rj",
    "viewCount": "2,778,116 views",
    "date": "Jun 22, 2026",
    "likes": "559",
    "location": "en_US",
    "channelName": "Galbot",
    "channelUrl": "https://www.youtube.com/@GalbotRobotics",
    "numberOfSubscribers": "529 thousand subscribers",
    "duration": 42,
    "commentsCount": "3",
    "text": "Trained on 20,000 hours of human motion data — the largest-scale dataset to date — AstraBrain-WBC 0.5 marks the first validation of Scaling Law in the field of humanoid robot motion control, demonstrating that a robot’s motion control capabilities can continue to improve with data scale, just like large AI models.\n\nMore importantly, the model enables real humanoid robots to execute a wide range of unseen motions through zero-shot generalization for the first time. Humanoid motion control is beginning to move beyond the traditional paradigm of “learning one motion, performing one motion,” and toward a foundation model approach capable of continuous generalization and evolution.",
    "descriptionLinks": [
      "Product information and related details."
    ],
    "subtitles": "Wireless Noise Cancelling Headphones",
    "video_supported_languages": "en-US",
    "video_supported_audio_languages": "en-US",
    "isMonetized": false,
    "commentsTurnedOff": false,
    "discovery_input": {
      "order_by": "Order by value"
    },
    "success": true,
    "input": "https://www.youtube.com/@GalbotRobotics/videos",
    "error": "Error value",
    "error_code": "ID-20260115-001"
  }
]