Watch a clip
A recorded clip, replayed with what watching it wrote down: the film advances with the video, each frame printed with its time and the model's score, and the frame where it spoke is circled.
No recorded clips here yet.
How it learned
No taught skills here yet.
Teach it yourself
Say what to watch for and pick a frame. Muse Wonder writes the skill, Muse Image makes 12 variations of your frame (6 where it holds, 6 near-misses) and Muse Spark says yes or no to each.
Made when each skill was taught, not from your request:
Results
Every number here comes from the evaluation scripts.
All the numbers
No results yet: the evaluation has not been run for this copy of the playground.
Privacy
While watching, frames go from the glasses to your phone, then to your Mac and the model that scores them; before an alert is spoken, Muse Spark looks at that one frame to confirm it. Only the scores are written down. Frames are not stored unless you save a clip on purpose, and the clips here were saved that way.
Faces are blurred in everything published, this page included.
Images made for you go to your browser. The server keeps them only in its call cache, so the same request is not paid for twice, and they keep Muse Image's provenance, exactly as made.
Skills watch objects, states, text, displays, signs and vehicles. They never watch for a particular person, for bodies or for health, and the glasses' capture light is never hidden.
This page loads nothing from other sites. A cookie with a random id, and your address stored only as a salted hash, count what you spend live.