Skip to content

Releases: macedonga/visor

Visor 1.3.4

Choose a tag to compare

@github-actions github-actions released this 18 Oct 16:10

Fixed

  • Questions about what you see. On some glasses, the photo for "Hey Meta, what am I looking at?" came a moment before the question, and the model answered without it. Visor now keeps that photo for the question (#18).
  • Setup with glasses that belong to another account. Setup stopped with "status 0x1042". It now says that the glasses belong to another Meta account, and Use another account takes you back to the login (#20).

Full Changelog: v1.3.3...v1.3.4

Visor 1.3.3

Choose a tag to compare

@github-actions github-actions released this 18 Oct 14:27

New

  • Directions by voice. Say "Hey Meta, give me directions to..." and the route opens in your map app, ready to start (#17).
  • Music by voice. Ask for a song, an artist, a playlist or a podcast, and it plays in a music app on your phone. You can also pause, resume, skip, go back and ask what is playing (#17).
  • Companion device. The first time your glasses connect, Android asks whether they can be Visor's companion device. With it, Visor opens the map or music app while it runs in the background. Without it, a notification opens the app. You can also allow it in Settings > Phone access (#17).
  • Reply without "Hey Meta". When the assistant asks you something, the glasses listen for your answer without the wake word.
  • The assistant asks before it remembers. When you mention something worth keeping, such as where you work, it asks first and saves it only after you say yes.

Improved

  • Meta AI shows what the glasses did themselves. Requests such as "play something" or "stop" now appear in the history, marked "Done on the glasses".
  • Conversations stay on topic. A reply without the wake word continues the conversation. A new "Hey Meta" continues it only within two minutes of the last turn, and never after A conversation stays open runs out.

Fixed

  • "Sorry, I couldn't get an answer" with Claude. Models that think before they answer, such as Claude Haiku 5.5, sometimes spent the whole answer thinking. They now have room for the answer.
  • Voice requests the glasses understood. Visor now reads how the glasses understood each request, so replies to messages, conversation focus and group conversations by voice reach the right place.

Full Changelog: v1.3.2...v1.3.3

Visor 1.3.2

Choose a tag to compare

@github-actions github-actions released this 18 Oct 12:38

New

  • Pin features to the quick tiles. Hold a feature on its settings page, then tap Pin to the tiles. Conversation focus and Hearing correction start and stop from their tile. The other features turn on and off.

Fixed

  • Conversation focus on glasses set up without the Meta AI app. The glasses refused to start it and asked for early access. Allow conversation focus now gives the glasses that access too (#12).
  • Conversation focus mentions tap and hold only while Tap and hold starts it is on.

Full Changelog: v1.3.1...v1.3.2

Visor 1.3.1

Choose a tag to compare

@github-actions github-actions released this 18 Oct 10:17

Fixed

  • Conversation focus opens again. In 1.2.3 and 1.3.0, Visor closed every time you opened it (#12).

Full Changelog: v1.3.0...v1.3.1

Visor 1.3.0

Choose a tag to compare

@github-actions github-actions released this 17 Oct 20:04

New

  • Choose how long the glasses wait for an answer. Settings > Meta AI > The glasses wait for an answer offers 20, 30, 45 or 55 seconds. The glasses used to give up after 20 seconds and say "Open the Meta AI app", which cut off answers that needed a web search, a smart home action or a slow model. Visor now gives them 45 seconds unless you choose otherwise.
  • The assistant reads the text you look at. When you ask about something in front of you, the glasses read its text from the photo, and Visor now gives that text to your model. A model without vision can now answer questions like "what does this say?".

Fixed

  • A slow model no longer keeps the glasses hanging. When a model took too long, Visor kept waiting after the glasses had already stopped listening, so the answer never arrived. Visor now stops in time and the glasses say that it could not get an answer.

Full Changelog: v1.2.3...v1.3.0

Visor 1.2.3

Choose a tag to compare

@github-actions github-actions released this 17 Oct 18:33

New

  • Save the event log. A new Save button next to Share opens the save window, so you can keep the log in Downloads or any other folder (#9).
  • Adaptive volume in Settings > Sound. The glasses use their microphones to hear how loud it is around you and adjust the volume to match. It works while the glasses are open (#11).

Fixed

  • GPT-6 models on OpenAI's API can use your integrations and memory. OpenAI refused every question that came with tools ("Function tools with reasoning_effort are not supported"), so Home Assistant and memory failed. Visor now asks these models to answer without extended reasoning, which they accept together with tools. Visor finds this out once per model and remembers it (#4).
  • Features now switch on for glasses that were never set up with the Meta AI app. Turning on spatial audio and the other features on those glasses ended in "refused", because the glasses rejected what Visor sent. They now take it (#10).
  • Conversation focus starts on glasses that were never set up with the Meta AI app. The glasses refused to start it because their model had no supported version on record. Visor now sends that along with the switch, and leaves alone a version Meta has already set (#12).

Full Changelog: v1.2.2...v1.2.3

Visor 1.2.2

Choose a tag to compare

@github-actions github-actions released this 17 Oct 17:51

Fixed

  • Integrations no longer go missing without a word. When your model's server cannot take tools (an Ollama model without tool support, for example), the answer now fails and shows the server's reason. Before, Visor asked again without the tools and said nothing, so the assistant answered as if your integrations were not there (#4).
  • The event log says when an integration sits out an answer. If a server is still connecting or cannot be reached when you ask something, the log names it and says why (#4).

Full Changelog: v1.2.1...v1.2.2

Visor 1.2.1

Choose a tag to compare

@github-actions github-actions released this 17 Oct 16:34

Fixed

  • Imported photos and videos now show up where you expect them. Visor sets the glasses' clock and time zone each time it connects, as the Meta AI app does. Before, nothing set the clock. On glasses whose clock had fallen behind, new captures were saved weeks in the past, below everything else in your phone's gallery and in Visor's Gallery, so they looked lost (#8).

    Captures you imported before this update keep the old date. Look for them around the date your glasses' clock showed.

Full Changelog: v1.2.0...v1.2.1

Visor 1.2.0

Choose a tag to compare

@github-actions github-actions released this 17 Oct 15:45

New

  • Features Meta keeps off, now on their own settings pages. You can switch each one on from Visor:

    • Hearing: Conversation focus, Hearing correction, Group conversations
    • Meta AI: Alarms, More follow-ups
    • Sound: Spatial audio

    Their raw switches are in Settings > Advanced.

  • Hearing test. Hearing correction sets itself to your hearing with the same test the Meta AI app uses: one tone at a time in each ear, shown on the 3D glasses. You can take it as often as you like. If the glasses save your profile but won't start it, Visor tells you.

  • Custom wake word. Say your own phrase three times into the glasses' microphone and the glasses learn it. For now the glasses accept it only while the assistant speaks English (US).

  • Group conversations run without Meta's key, so they are not encrypted. Anyone nearby with the right equipment can listen.

Improved

  • Setup asks for location and for location all the time as two separate steps, so allowing the first no longer looks like a failure.
  • Setup's buttons react to your finger and animate their labels.
  • Visor reads the glasses' configuration in seconds instead of about two minutes.
  • More space between the glasses and the list in Settings.

Fixed

  • Imports: a file that arrives damaged, or that the phone won't save, is now reported instead of disappearing. The event log names every saved file with its folder, size and date (#8).
  • Holding a feature no longer crashes the app.
  • Gesture settings are written the way the glasses expect them.
  • The update notification opens Visor, which then opens the download, instead of handing a link to any app.

Full Changelog: v1.1.2...v1.2.0

Visor 1.1.2

Choose a tag to compare

@github-actions github-actions released this 17 Oct 10:33

Full Changelog: v1.1.1...v1.1.2