Your data¶
Every Trinet recording is a synchronized set of three streams, all timestamped on the same camera clock:
| Stream | What it is | Rate |
|---|---|---|
| Video | H.264 or H.265, 1920×1080 per camera (per eye on stereo) | 30 fps |
| Motion (IMU) | Accelerometer, gyroscope, magnetometer, sensor temperature | 400 Hz (about 560–570 Hz on V1/V2) |
| Frame timestamps | One timestamp per video frame, at the middle of its exposure | One per frame |
Plus audio on cameras with microphones, and the camera's identity and calibration inside the video file.
One clock¶
The camera keeps one steadily increasing clock and stamps everything against it: each motion sample at the moment it was measured, and each video frame at the middle of its exposure. Lining up motion with video is therefore just comparing timestamps — no separate sync signal to decode. See Timing and sync.
Don't align with video playback times
Use the frame timestamps the camera records, not the video container's playback timestamps or your computer's clock — those carry buffering and transport jitter of milliseconds.
Where recordings come from¶
| How you recorded | What you get |
|---|---|
| Card recording (button) | .mp4 + .imu + .vts in the Trinet folder — see Files on the card |
| Trinet app / SDK over USB | A folder with video.mp4, imu.bin, frames.bin, meta.json — see Record and review |
| Computer (toolkit USB recorder) | The same set of files, reconstructed from the live stream |
Over USB, the motion data travels inside the video stream; the app, SDK and toolkit separate it back into the same set of files, so every tool reads every recording the same way.
Self-describing¶
- Every file states its format version in its header, and readers handle every version — so recordings from older cameras and firmware keep working.
- The video file carries the camera's device ID and calibration, so a clip is usable even if it gets separated from its folder.
- The motion data file records its own sample rate — always read it from there rather than assuming a value.
Working with your data¶
- Python toolkit — inspect, repair, visualize, export to MCAP / ROS 2.
- Calibration — intrinsics, camera–IMU extrinsics and time offset.
- File formats — layouts, versions and the full byte-level specifications.