Kung Fu BOT: The Smoothest Humanoid Robot

Imagine a world where humanoid robots master Kung Fu, AI helps design immersive game experiences, and video tools generate matching soundscapes in an instant. Well, that future is unfolding now.

Unitree’s G1 is here, and it’s not just any humanoid robot—it’s agile, powerful, and even capable of Kung Fu moves. With top-tier sensors and advanced motion control, it navigates rough terrain and handles objects with precision. Whether for research, industry, or entertainment, G1 is pushing the limits of what robots can do.

Microsoft is shaking up game development with Muse, an AI model that generates game visuals and controller actions. Built with Xbox Game Studios’ Ninja Theory, it turns gameplay data into interactive possibilities. The best part? It’s open-source, giving developers a new playground to experiment with AI-driven gaming.

Luma Labs just made AI video creation even better—now with sound! Dream Machine’s new Video to Audio feature lets users generate audio to match their AI-generated videos with a single click. From ambient nature sounds to sci-fi effects, it’s a game-changer for content creators looking for a seamless multimedia experience.

Let’s explore these innovations shaping the next frontier of technology!

Unitree G1: The Agile Humanoid Robot

Unitree Robotics unveiled its advanced humanoid robot, the G1, which is designed to push the boundaries of flexibility and manipulation capabilities. The G1 features a range of impressive movements, including navigating challenging terrain and performing Kung Fu moves, thanks to its "agile upgrades." It is equipped with advanced sensors like Intel RealSense D435 and LIVOX-MID360 3D LiDAR, allowing it to accurately perceive its surroundings. The robot can achieve impressive running speeds and is capable of precise object manipulation with its force-controlled dexterous hands. Unitree continues to upgrade the G1's algorithm, enabling it to learn and perform virtually any movement. Users are encouraged to share their desired movements in the comments, while maintaining a safe distance from the robot. The G1 is versatile, suitable for applications in research, industry, entertainment, and more.

Microsoft Introduces Muse: AI for Game Development

Microsoft Research has introduced Muse, the first World and Human Action Model (WHAM), a generative AI model for video games that can generate game visuals, controller actions, or both. Developed in collaboration with Xbox Games Studios' Ninja Theory, Muse was trained using gameplay data from Bleeding Edge, a 4-versus-4 online game. The model aims to support human creatives by enabling the exploration of new interaction paradigms and creative uses. Microsoft is open-sourcing the weights, sample data, and the WHAM Demonstrator, a concept prototype with a visual interface for interacting with WHAM models, on Azure AI Foundry. Muse demonstrates capabilities such as consistency, diversity, and persistency in generating gameplay sequences, showcasing its potential to support gameplay ideation and pave the way for future AI-based game experiences.

Dream Machine Now Offers Video to Audio

Luma Labs has introduced a significant update to its Dream Machine platform by adding a "Video to Audio" feature. This new capability allows users to generate audio for their AI-created videos with a single click or by providing customized prompts for more precise soundscapes. The audio generation is available in beta and is free for all users, making it a standout feature compared to other AI video tools that lack integrated audio capabilities. Users can now enhance their video creations with sounds that match the visual content, such as ambient nature sounds or sci-fi effects, further enriching the overall multimedia experience.

Hand Picked Video

In this video, we'll look at Elevenlabs Conversational AI Agents.

Top AI Products from this week

• Basalt - Basalt is the platform to build and operate AI features : Craft high-quality prompts with our AI-powered Copilot, test and evaluate LLM outputs, deploy seamlessly with our SDK, monitor and refine performance in real conditions—all in a collaborative workflow.

• OpenArt Consistent Characters - OpenArt Characters lets you create images of consistent characters from just one image or description. Pose, place, and combine them in any scene for infinite storytelling possibilities.

• Saywise - Saywise connects career advice seekers with experienced professionals through live video AMAs. After each AMA session, AI automatically generates bite-sized Q&As, building nuggets of wisdom. Get inspired, build new connections, and ask your questions!

• Pinch - Pinch is a virtual conferencing platform designed for cross-lingual communication. Real-time voice translation allows you to appear as a native speaker of over 30 languages.

• Forage Mail - Forage filters low-priority emails from your Gmail. It sends you a clean, daily summary of the emails it filtered out, including a TLDR for every newsletter. It learns your preferences and gives you full control with custom rules. Works with any email app.

• Steve by Wonder Family - We built the first AI that created a REAL eCommerce business and made $500K.

This week in AI

• QwQ-Max Preview Unveiled - Qwen's QwQ-Max-Preview, built on Qwen2.5-Max, excels in reasoning, math, & coding. Open-sourcing & APP release are planned

• Meta AI Launches in Middle East - Meta AI expands to the Middle East, offering Llama 3.2 on Facebook, Instagram, WhatsApp, and Messenger.

• DeepSeek's FlashMLA - DeepSeek launches FlashMLA, an open-source MLA decoding kernel optimized for NVIDIA Hopper GPUs, enhancing AI performance with reduced memory usage and faster inference.

• Google's Gemini Code Assist - Google offers a free Gemini Code Assist version for developers, supporting up to 180,000 code completions monthly. \

• Convergence's Proxy - Convergence's Proxy is an AI assistant that automates web tasks, learns repeatable actions, and offers global accessibility at a lower cost than competitors