Generalist · Launch Video Breakdown: Hook, Pacing & Motion Design
GEN-1.5 learns new physical tasks in a few seconds and generalizes. Pretrained on physical data at scale toward general intelligence for the physical world.
Scene-by-scene timeline & spoken transcript
The Hook
Introducing GEN-1.5: Rapid Task Learning
“makes possible that you can just walk up. Our new model, GEN-1.5, is an to learn longer horizon tasks. And in some cases, it can just watch humans and it's showing physical generalization,”
- On screen
- makes possible that you can just walk up GEN-1.5 Our new model, GEN-1.5, is an to learn longer horizon tasks. A B And in some cases, it can just watch humans Human-to-Robot In-Context Learning And it's showing physical generalization, Novel Tool Use Improvisation
- Camera
- Dynamic cuts between various robot arm perspectives, including close-ups of grippers, wide shots of the robot interacting with objects, and split-screen views comparing different angles.
- Motion
- Rapid montage of robot demonstrations, split-screen comparisons, and text overlays that animate in with a subtle fade and scale.
Product Reveal
The Core Technology: Generalized In-Context Learning
“to do, and it generalizes it. we're seeing for the first time is a very never seen a model that can do this before. Beyond its context, it can also learn with”
- On screen
- to do, and it generalizes it. PROMPT FRAMES MODEL WEIGHTS ROBOT TRAJECTORIES INPUT FRAMES MODEL OUTPUTS Generalized In-Context Learning we're seeing for the first time is a very In-Context Learning 10 GRADIENT STEPS (5 MIN OF DATA) IN-CONTEXT LEARNING (3-12 SEC OF DEMOS) Retrieve Money From Wallet 60.7% Fold And Crease Paper 50% Twist Lid Off Glass Jar 60% Stack Two Small Cups 67% Sweep Trash With Brush 37.3% Open Book Cover 54.7% Brush Cube Into Bowl 60.8% Flip Phone Upside Down 78% Unzip Pencil Pouch 55.5% Remove Vacuum Pad 64% TASK SUCCESS RATE (%) never seen a model that can do this before. Held-Out Validation Task In-Context Learning Few-Shot Learning 100% TASK SUCCESS RATE (%) 67% 75% Beyond its context, it can also learn with 10 Steps
- Camera
- Static shots of data visualizations and bar charts, interspersed with close-ups of robot actions.
- Motion
- Animated data visualizations, text reveals, and seamless transitions between abstract data and concrete robot demonstrations.
Feature Teaser
Improvisational Intelligence & Tool Use
“In one example, we taught it to use a banana into an impromptu brush. block up and dumped it into the bowl. It's overall showing more improvisational we covered the bowl with a piece of paper, and it hand, and then it just did it with two hands.”
- On screen
- In one example, we taught it to use a banana into an impromptu brush. block up and dumped it into the bowl. It's overall showing more improvisational Improvisational Intelligence we covered the bowl with a piece of paper, and it Improvisational Intelligence hand, and then it just did it with two hands. Expected Improvisational
- Camera
- Dynamic close-ups and medium shots of the robot performing complex, improvised tasks, often from a first-person robot perspective.
- Motion
- Fast-paced cuts, split-screen comparisons highlighting 'Expected' vs. 'Improvisational' actions, and subtle camera movements to follow robot actions.
Call to Action
Human-to-Robot Learning & Future Vision
“in-context learning emerge, where a intelligence in robot foundation models.”
- On screen
- Human-to-Robot in-context learning emerge, where a In-Context Learning intelligence in robot foundation models.
- Camera
- Medium shots of a human interacting with the robot, demonstrating the learning process, followed by a final brand logo reveal.
- Motion
- Smooth transitions, a final brand logo animation with a subtle glow, and a sustained synth pad for a conclusive feel.
Motion design primitives
Synthesizable in Remotion · key components of the ai-robotics-launch-montage template.
MultiViewGridReveal
A grid of small video clips rapidly assembles and disassembles, revealing a central product or concept. Clips scale and fade in/out, creating a dynamic mosaic effect.
+ Code construction recipe
Use `map` to render an array of `Video` components within a `GridLayout`. Animate `scale` and `opacity` using `spring` functions with staggered delays for each grid item. Use `AbsoluteFill` for the background.
SplitScreenComparison
The screen splits into two or more distinct sections, each displaying a different video or image, often used for side-by-side comparisons or multi-angle views.
+ Code construction recipe
Employ `AbsoluteFill` with `flexDirection: 'row'` or `column` to create split sections. Render separate `Video` or `Img` components in each section. Use `interpolate` for smooth `width` or `height` transitions to reveal/hide sections.
DataVizReveal
Statistical charts (e.g., bar graphs, heatmaps) animate into view, with bars or data points growing/changing over time to illustrate performance metrics.
+ Code construction recipe
Render SVG `rect` elements for bar charts or `div` elements for heatmaps. Animate `height` or `backgroundColor` using `spring` or `interpolate` with `easeOutCubic` for a smooth, impactful reveal. Stagger animations for individual bars/cells.
Style DNA & aesthetic system
Visual aesthetic
Typography
Motion
Special effects
Pacing & rhythm
Conceptual story rhythm
- 1
The 'Wow' Moment
Purpose. Immediately showcase the product's core capability in a visually striking way to hook the audience.
Execution. Rapid montage of diverse robot tasks, demonstrating versatility and speed, accompanied by a strong audio riser and sub-bass drop.
- 2
The 'How It Works' Explanation
Purpose. Provide a high-level understanding of the underlying technology without getting bogged down in excessive detail.
Execution. Transition from robot demos to abstract data visualizations (e.g., neural network diagrams, bar charts) to illustrate the 'intelligence' aspect.
- 3
The 'What Else Can It Do?' Expansion
Purpose. Demonstrate advanced or unexpected capabilities, highlighting innovation and future potential.
Execution. Showcase improvisational tool use and complex, multi-step tasks, often with split-screen comparisons to emphasize the 'novelty'.
- 4
The 'Future Vision' Conclusion
Purpose. Reinforce the product's impact and broader implications, leaving the audience with a sense of excitement.
Execution. Human-robot interaction scenes, followed by a clean brand logo reveal, tying the technology back to real-world application and the company vision.
Remix in Animatiq Studio
Master remix prompt · agent-ready
Sample remix
Objective
Launch video for 'SkyMind', an AI-powered autonomous drone for environmental monitoring.How the recipe adapts
Remix blueprint · ai-robotics-launch-montage
Scene durations (10 scenes, 193s)
Key Remotion components
Related Hardware & Devices Product Launches
Explore all Hardware & Devices launches →
Bolt v2 pairs Claude Code and Codex agents with built-in hosting, database and storage so apps ship without setup.
Making life multiplanetary
Waves launches camera glasses built for creators, with stealth recording and all-day livestreaming, now on pre-order.