SCENE-07: Recognize faces only when I ask
Last updated:
Part of the EchoVision Wish List, under Scene Description and Questions.
- Status: Under review.
- What owners asked: I would like to suggest making facial recognition a more intentional, user-controlled feature. One option would be to require a specific voice command, such as "Who is this person?" or "Identify this person," before initiating facial recognition. Another option would be to add an on/off toggle within the companion app, allowing users to enable the feature only when they want to use it.
- Related: GEN-05: Recognize people I know
- Discussed in: Face recognition prompt, on the beta-feedback list (members only); more observations of Facial recognition, on the beta-feedback list (members only); Face recognition, on the beta-feedback list (members only)
The need
Scene description is for learning about your surroundings. In the face recognition beta, when a person fills most of the view, the glasses try to recognize them and then offer to save them, even when you only wanted to know about the room. Testers describing a presentation, or the room they were in, heard an offer to name the presenter or a family member at the end of every answer.
Today's workaround
None. Recognition can't be turned off, and the offer to save someone comes at the end of the answer.
What it would look like
Testers suggested three ways, and they don't all agree:
- Recognize only when asked, with a question such as "Who is this?" or "Identify this person".
- An on or off switch for face recognition in the EchoVision app.
- Keep recognizing by itself, but make the offer to save someone quick to turn down, with a button tap or "No thanks".
Pros and cons
- Pro: scene description answers stay about the scene.
- Pro: the glasses don't analyze the faces of bystanders unless you ask, which respects their privacy (guiding principle 3).
- Pro: a tester wondered whether checking every face against a large set of saved people slows scene description down.
- Con: you may miss being told that someone you know is in front of you. One tester prefers that to happen by itself.
- Con: an extra switch or command to learn (guiding principle 2).
Open questions
- Should recognizing people you saved and offering to save new people be separate choices? A tester might want to hear known names by themselves, but only save people when they ask.
- Which should be the default when face recognition is released?
History
- November 7, 2026: Under review. From the beta-feedback list.