> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cyberwave.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Replay timeline & recordings

> stub — Reading the replay timeline, the color legend, and slicing recordings into episodes.

The **Replay** timeline shows everything an [environment](/feature-reference/environment-editor/index) recorded over a time window, laid out left-to-right. Dots mark captured frames; filled bars show where each track has data.

## Open replay over a recording

1. Switch the environment to **Replay**.
2. Click the **calendar** button.
3. Pick the date (the day you recorded, or **May 8, 2026** if you're using the suggested demo environment).
4. **Load sessions**, no filters needed and the timeline fills in.

## Color legend

Open the **Legend** popover in the toolbar at any time. Colors map to the kind of data, not severity:

| Color               | Track                                                  |
| ------------------- | ------------------------------------------------------ |
| 🔵 Blue             | Robot / point cloud (joint states, depth/LiDAR frames) |
| 🟣 Purple           | Camera / video                                         |
| 🟢 Green (dots/bar) | Audio                                                  |
| 🟢 Green (span)     | An **episode** — a labeled segment you created         |

Tracks that overlap in time — for example a camera and a microphone recorded together —
are drawn in **separate lanes** so you can see both at once. As the playhead moves, the
panel on the right shows only the recordings active at that moment (a camera's video, a
microphone's audio, or both), and clears when nothing is recorded there.

## What you can do here

* **Scrub & play back** at 1× or 2×; click any frame to jump to it.
* **Zoom** into a stretch of the timeline for frame-level detail.
* **Filter markers** by twin with the **Markers** dropdown.
* **Browse sessions** by date with the calendar picker.
* **Create an episode** — select a window and label it as a [training/dataset](/feature-reference/datasets/import) segment.

## Why it matters

* **Research:** turn real runs into labeled [episodes and datasets](/overview/features/replay-and-historical-data) for training [VLAs](/feature-reference/ml-models/smolvla-training).
* **Security & review:** re-watch exactly what a robot did, find the moment something went wrong, and inspect it frame-by-frame.

See [Recording replay (signed URLs)](/feature-reference/environment-editor/replay) for how playback data loads, and [Replay & Historical Data](/overview/features/replay-and-historical-data) for the full pipeline.