Alibaba Qwen · Launch Video Breakdown: Hook, Pacing & Motion Design

Scene-by-scene timeline & spoken transcript

1. 0:00 – 0:10

   The Hook

   Introducing Qwen-RobotNav: Embodied Agent Mobility

   “(No spoken dialogue — Energetic electronic score with driving 808s and synth percussion. High-energy audio transitions with risers and sub-bass drops.)”

   On screen

   Qwen-Robot Suite 01 Qwen-RobotNav The Gateway to Physical Mobility for Embodied Agents Target Tracking and Object-Goal Navigation Target Tracking I'm taking a break here. Could you help me find a restroom nearby? Qwen-RobotNav Turn 1 Tokens: 6.635 Steps: 0 Tools: navigate_by_instruction Thought complete I'll help you find an open restroom. Let me start by loading the task reference and observing my surroundings. Qwen-RobotNav Turn 2 Tokens: 7.802 Steps: 0 Tools: navigate_by_instruction Thought complete Let me navigate toward the exit entrance area to find building signage. navigate_by_instruction - 77 steps Instruction: Turn around and go around the gray chair. Find the wall, then turn left and stop at the exit entrance. Steps_taken: 77 | mps_steps: 112 [Step 0] The agent is in a modern, open-lounge area with polished floors. A large digital display on a stand is to the left, and a round table with chairs is to the right. In the background, there are glass partitions and an escalator. [Step 4] The view remains in the same lounge area. The camera angle is slightly to the left, with the left side of the room, with the large gray chairs, and the digital display are still prominent landmarks. [Step 8] The agent appears to be moving slightly forward or adjusting its position under the lounge. The perspective shifts slightly, bringing the round table and the glass partitions underneath it more into the center of the frame. [Step 12] The agent turns significantly to the right, facing away from the lounge area and toward a long corridor. The digital display is no longer in the background; a person in a black shirt stands near a counter, and a black umbrella-like object sits on the floor. [Step 16] The agent continues to navigate around the furniture. A white wooden wall is visible to the left, and the digital display is now out of view. To the right, a person in shorts and a black t-shirt stands near blue chairs. The black umbrella object is visible to the left. [Step 24] The agent has moved past the partition and is now close to a wooden wall with a door. The camera angle is wider, and the agent is looking at the sign of the door and table. In the background, there is a wooden wall with vertical strips and a door with Chinese characters. [Step 28] The agent moves away from the table and chairs, heading towards the wooden wall with a door. The camera angle now encompasses a polished floor reflecting the ceiling lights. The wood. Instruction following & Object-Goal Navigation Found it, the restroom is right next to the corridor.

   Camera

   Initial static text, then follows a robot dog from behind, tracking its movement. Later, a static shot of a person with a chat bubble, then a wider shot showing the robot and a detailed UI overlay.

   Motion

   Text reveals with subtle fades and scale, followed by real-world footage of robot interaction. UI elements animate in with a clean, modern aesthetic. Chat bubbles appear with a quick pop-in effect.

2. 0:11 – 0:14

   Feature Teaser

   Qwen-RobotManip: Physical Interaction

   “(No spoken dialogue — The electronic score maintains its energetic pace, with a slight build-up in intensity as new product features are introduced.)”

   On screen

   02 Qwen-RobotManip The Foundation of Physical Interaction for Embodied Agents Sequential Execution under Task Interference

   Camera

   Static shot of a robotic arm performing a task, with text overlay. The camera remains fixed, focusing on the robot's precise movements.

   Motion

   Text appears with a clean fade-in. The focus is on the real-world demonstration of the robotic arm, with minimal additional motion graphics, emphasizing clarity and directness.

3. 0:15 – 0:20

   Product Reveal

   Qwen-RobotWorld: Natural Language-Driven World Model

   “(No spoken dialogue — The electronic score continues, providing a consistent high-energy backdrop. Percussive elements are prominent, driving the narrative forward.)”

   On screen

   03 Qwen-RobotWorld Infinite Worlds for Embodied Agents Natural Language-Driven World Model Prompt: Pick up the green plate and place the carrot on it Model-Generated Prediction Video of robotic arm manipulation Prompt: Follow the marked path Model-Generated Prediction Video of robot dog movement Prompt: Put the green gift box into the shopping cart Model-Generated Prediction Video of robotic arm manipulation

   Camera

   Static shots showcasing side-by-side video comparisons. The camera is fixed on the screen content, highlighting the model's predictions.

   Motion

   Split-screen comparison with clear labels. Text overlays for prompts and descriptions animate in with simple, direct fades. Blue bounding boxes highlight the video content, drawing attention to the visual demonstration.