Scale AI

Scale AI · Launch Video Breakdown: Hook, Pacing & Motion Design

Open robot dataset with 40K+ demonstrations and a step-by-step reasoning model

0:00 · The Hook · Warehouse Facility & Co-branding
0:00 / 0:00
AI & MLProductionOctober 29, 2026

High-tech facility showcase with co-branded partner logos

10

Cuts / min

3:22

Runtime

26K

Views

270

Likes

0

Reposts

0

Replies

Scene-by-scene timeline & spoken transcript

  1. The Hook

    Warehouse Facility & Co-branding

    “(No spoken dialogue — ambient electronic synth score)”

    On screen
    Scale AI Logo | Hugging Face Logo
    Camera
    Static wide establishing shot of warehouse and mobile gantry
    Motion
    Overlay fade-in of dual partner brand marks
  2. Product Reveal

    Introducing FineART & FineART-VLA

    “We're introducing FineART, Scale's first open-source robotics dataset. FineART is the world's largest subtask-annotated data. With Hugging Face's LeRobot team, we're releasing FineART-VLA, the first open-source System 1-System 2 policy.”

    On screen
    Jade Choghari / Scale, Physical AI Research | FineART Core Team
    Camera
    Medium close-up cutting to wide seated interview in lab
    Motion
    Clean lower-third title card transition
  3. Feature Teaser

    Scale and Subtask Annotations

    “Our dataset features over 40,000 episodes, 100% subtask-annotated. With more than half a million subtasks, we enable learning, step by step, at scale.”

    On screen
    Jackson Lee / Scale, Physical AI Research Ops / 100% subtask-annotated. / With more than half a million subtasks, we enable learning, step by step, at scale.
    Camera
    Medium shot in robotic testing bay pan to monitor screen and triptych cut
    Motion
    3-way vertical split screen showing synchronized robot manipulation tasks
  4. Feature Teaser

    Unified Architecture & LeRobot Integration

    “FineART-VLA combines a high-level planner with a low-level motor executor into a single unified architecture. FineART-VLA is the first dual-system VLA natively integrated in LeRobot.”

    On screen
    Pepijn Kooijmans / Hugging Face LeRobot | FineART Core Team
    Camera
    Medium close-up shot transition between research leads
    Motion
    Direct-to-camera presentation with branded neon background
  5. Feature Teaser

    System 1 / System 2 Architecture & Objectives

    “FineART-VLA is trained on three objectives: tokenized actions, tokenized text, which are the subtasks, and continuous actions. The model trained on our annotations has two modes of inference: hierarchical inference and flat mode. Hierarchical inference allows the model to predict its own subgoal before acting in the real world. The dataset, including the annotation and model weights and training code, are available on Hugging Face.”

    On screen
    FINEART-VLA END-TO-END SYSTEM 2 / SYSTEM 1 / System 2 plans the subtask. System 1 acts. Three objectives train both.
    Camera
    Full-screen technical diagram display to seated speaker
    Motion
    Animated UI flowchart with real-time waveform and action matrices
  6. Feature Teaser

    Real-World Benchmark Evaluation

    “We then tested our policy in the real world, benchmarking across embodiments against baselines with and without mid-training. Spanning 3,400 rollouts, it is one of the largest evaluations on record.”

    On screen
    Jackson Lee / Scale, Physical AI Research Ops
    Camera
    Medium close-up to 4x4 matrix grid view of simultaneous rollouts
    Motion
    16-pane simultaneous video grid playback
  7. Feature Teaser

    Empirical Gains & Zero-Shot Transfer

    “Mid-training on FineART's subtask annotation raised the success rate of our policy on spatial-instruction tasks by 68 percentage points. For instance, it can recognize the difference between left, right, or center. It also boosts long-horizon performance, raising success rate of our prepare-breakfast task by 60 percentage points. We also see zero-shot transfer from the model on a completely new task and on a new target embodiment, which has a completely different morphology.”

    On screen
    (No on-screen text)
    Camera
    Overhead and tabletop robotic execution demonstrations
    Motion
    Real-world manipulation showcase with bimanual arms
  8. Call to Action

    Open Source Release & Conclusion

    “Are you ready to build? Annotate your dataset with LeRobot Annotate, fine-tune FineART-VLA, and get started. Everything is open source. At Scale, we are pushing the frontier of robotics research with high-quality, high-diversity data. We work on developing new multimodal capabilities with leading labs and organizations like Hugging Face. FineART is our first open-source robotics dataset and model, built to help the community move faster. This is just a canvas, and we can't wait to see what you build with it. After all, robotics is a fine art.”

    On screen
    Pragna Mannam / Scale, Physical AI Research / Scale AI Logo | Hugging Face Logo
    Camera
    Medium shots alternating with robotic arm close-ups and closing logo lockup
    Motion
    Final static brand card overlay on out-of-focus robot hardware

Related AI & ML Product Launches

Explore all AI & ML launches →
Grok Bot 1.0 — AI teammates for your team
@bot
AI & ML
@bot
Grok Bot 1.0 — AI teammates for your team@bot
Tauvus
@tavus
AI & ML
@tavus
Tauvus@tavus
Fish Audio
@FishAudio
AI & ML
@FishAudio
Fish Audio@FishAudio