Skip to main content

LeRobot Studio: a browser for embodied datasets

About

LeRobot Studio (https://io-ai.tech/lerobot/) is a local LeRobot inspection tool built by IO-AI.
Drag a local LeRobot folder or archive into the browser to play it back—data stays on your machine, balancing speed and privacy.

Why the LeRobot format matters​

High-quality datasets drive embodied AI, yet robotics data has long suffered from fragmented formats, inefficient storage, and poor reuse.

The LeRobot dataset spec (v2.1 / v3.0) is becoming a de-facto interchange format. Major stacks—including OpenPI and NVIDIA GR00T (Isaac Lab)—support or interoperate with it. It addresses the “data Tower of Babel” with:

  1. Unified structure — Parquet + MP4 bundles vision, motor state, and actions for cross-platform use.
  2. Cloud-native workflows — streaming lets you load from the Hugging Face Hub or local storage without downloading everything upfront.
  3. Community interoperability — standardized layouts improve discoverability, indexing, and sharing.

LeRobot Studio is a general-purpose viewer built on that standard so you can understand and validate datasets quickly.

Core features​

LeRobot Studio runs entirely in the browser—no local Python setup required for inspection.

1. Multimodal synchronized playback​

Robot logs combine high-rate sensors and variable-FPS video. Studio aligns them on a single timeline so you can see:

  • Multi-camera video
  • Joint state curves
  • Commanded actions

Scrub the timeline to inspect perception and control at any instant.

Multi-view sync

2. Command vs state comparison​

During training or debugging, comparing commanded actions to measured state is essential.

  • Action: commanded control (dashed in plots).
  • State: measured feedback (solid).

The chart view highlights delay, jitter, or bias—useful for spotting hardware issues vs logging bugs.

Chart panel

3. Quick validation and dataset health​

Studio loads local folders or .tar.gz archives in-browser (no upload). You can also open remote Hub datasets by URL for quick previews.

After loading, open the Dataset health sidebar to see structural checks (errors / warnings / pass), including:

  • Filesystem layout — meta/info.json, meta/episodes, data/, videos/ for v2.1 or v3.0
  • Metadata — required fields such as fps, codebase_version
  • Features — dtypes and shapes
  • Episodes — counts, lengths, zero-length segments

Issues are listed explicitly so you can confirm exports before training.

How to access it​

Option A — IO-AI website​

https://io-ai.tech/lerobot/

  • No login required.
  • Best for quick local inspection and demo datasets.

Welcome page

Option B — IO-AI data platform​

IO-AI data platform

  • Integrated with dataset management workflows.
  • Best after capture, labeling, or conversion when you want QA inside the platform.

UI tour​

Episode sidebar​

Browse and search episodes with metadata to jump to interesting segments.

Episode sidebar

Raw data panel​

Inspect the per-frame JSON payload—useful for copying timestamp, action, or state values into debugging scripts.

Raw JSON panel

Playback controls​

Frame-accurate controls including frame stepping for analyzing grasp or insertion phases.

Playback controls

Dataset health panel​

Open Dataset health to review automatic validation results (pass / warning / error) covering v2.1/v3.0 structure, meta/info, features, and episodes.

About IO-AI​

IO-AI (io-ai.tech) builds infrastructure for embodied intelligence. LeRobot Studio is part of our data toolchain to tighten the data loop.