Labelled robot video, back before your next training run.

Send your recordings. Talo returns boxes, tracks, action segments and task descriptions in the schema your pipeline already reads.

Before
Before. A raw frame, as you would send it.
After right hand bottle cup action: pour
After. Object tracks and an action label, from the sample below. right handbottlecup

01 Problem

Robot experiments stall while recordings wait for labels.

Every frame needs a box.

One day of teleoperation produces thousands of frames to label.

Labels arrive without a trail.

Labels often arrive without the rules used or a record of what was checked.

Every export needs converting.

Someone on the team rewrites the vendor's format before the pipeline can read it.

02 Sample

Check our work before you talk to us.

A demonstration clip labelled by Talo for this site; not customer work.

First frame of the sample clip: a robot hand holding a bottle beside a cup.

Labelled Self-checked

Pour water, head camera

Task description

Labels by Talo, checked in a second pass by the same annotator and not independently reviewed.

Send a short brief and we label a small batch of your own recordings this way, to criteria agreed before work starts.

03 Collection

We also collect the demonstrations, labelled to the same criteria.

Partners record to your tasks and environments; Talo labels and delivers each episode to the same criteria as your annotation work.

Egocentric

Hands at work, filmed from the worker's eye line.

A head-mounted camera records both hands and the objects they handle, from the start of the task to the end. Each episode is labelled with object tracks, action segments and a task description.

Handheld gripper (UMI)

A handheld gripper records demonstrations without a robot.

Wrist-camera video, with the gripper's position, orientation and opening width recorded in step with each frame. The fingers match the robot's, so a robot fitted with the same gripper can learn from the demonstrations directly.

Teleoperation

An operator drives your robot through your tasks.

Video from the robot's own cameras, with its joint state and the operator's commands recorded in sync. The data comes from your hardware, so there is no gap between the body that demonstrates and the body that learns.

Ask about collection

Annotation pilot

Tell us what you need labelled, and by when.

Start with a small batch of your own footage, labelled to agreed criteria, before any larger order. Don't paste raw data here; we arrange transfer after scoping.

All fields are required unless marked optional.

Robot, task, and the labels you need. A rough description is enough.

Hours or clips of source video, not annotation labour.