[BUG] Korean text is decomposed into separate Jamo characters on Even Realities G1 in MentraOS 2.12 beta 25 and later including 3.0

[heroson-lab]

Thank you for submitting a bug report to MentraOS! Please provide the following info as accurately as possible so that we can best assist you.

Bug Report

Glasses Hardware Model: Even Realities G1
Platform: Android
Phone OS Version: [Please specify Android version]
Phone Model: [Please specify phone model]
App Version: MentraOS 3.0 (and 2.12 beta 25 and later)
Network Conditions: WiFi / Cellular
Battery Level: Not relevant / Issue occurs regardless of battery level
Other Apps Running: Not relevant
Frequency: Always
User Impact: Major

Issue Details

Steps to Reproduce:

1. Connect Even Realities G1 glasses to MentraOS 3.0 (and 2.12 beta 25 and later)
2. Open the Captions app and use Korean speech-to-text transcription without translation.
3. Speak a Korean sentence, for example: 이것은 한국어입니다.
4. Check the transcription displayed in the Mentra smartphone app. The Korean text is displayed correctly.
5. Check the same transcription on the Even Realities G1 display. The Korean characters are displayed as separated Hangul Jamo instead of composed Hangul syllables.
6. The same issue can also be reproduced when Korean text is displayed using the Live Translation app.

Expected Behavior:

Korean text should be displayed on the Even Realities G1 as properly composed Hangul syllables.

For example:

이것은 한국어입니다.

This worked correctly on the same Even Realities G1 hardware when using MentraOS 2.12 beta 24.

Actual Behavior:

After upgrading to MentraOS 3.0 (and 2.12 beta 25 and later), Korean Hangul syllables displayed on the G1 are separated into individual consonants and vowels (Jamo).

For example:

Expected:

이것은 한국어입니다.

Displayed on the G1 approximately as:

ㅇㅣㄱㅓㅅㅇㅡㄴ ㅎㅏㄴㄱㅜㄱㅇㅓㅇㅣㅂㄴㅣㄷㅏ.

The important observation is that the Korean text is displayed correctly in the Mentra smartphone app, but incorrectly on the G1 display.

The problem also occurs in Captions without using any translation, so this does not appear to be caused by the translation engine.

Other Relevant Details:
This appears to be a regression introduced in MentraOS 2.12 beta 25 and later.

The comparison is:

• MentraOS 2.12 beta 24 + Even Realities G1 → Korean displays correctly.
• MentraOS 3.0 (and 2.12 beta 25 and later)+ Even Realities G1 → Korean Hangul syllables are separated into Jamo.
• MentraOS 3.0 (and 2.12 beta 25 and later) + Captions → Same issue.
• MentraOS 3.0 (and 2.12 beta 25 and later)+ Live Translation → Same issue.
• Mentra smartphone app → Korean displays correctly.

Because the issue occurs in both Captions and Live Translation, while the corresponding Korean text is displayed correctly on the smartphone, the speech recognition and translation results themselves appear to be correct.

The issue therefore seems likely to occur in the common MentraOS 2.12 beta 25 and later including 3.0 text display/rendering path for the Even Realities G1.

It may be worth checking for changes between MentraOS 2.12 beta 24 and MentraOS 2.12 beta 25 involving:

• Unicode normalization (NFC/NFD)
• Korean Hangul composition
• G1 text encoding
• font/text shaping
• the G1 display adapter
• text processing before transmission to the glasses

For diagnostic purposes, it may also be useful to check the Unicode code points immediately before the text is sent to the G1.

For example, the Korean syllable:

한

should normally be represented in NFC form as:

U+D55C

If it is being converted somewhere in the MentraOS 3.0 (and 2.12 beta 25 and later) display pipeline into decomposed Hangul Jamo such as:

ᄒ + ᅡ + ᆫ

(U+1112 U+1161 U+11AB)

this could explain the behavior.

However, if the text is still correctly composed (for example, U+D55C) immediately before being sent to the G1, the issue may instead be occurring in the G1-specific encoding or rendering layer.

Since the same G1 hardware displays Korean correctly with MentraOS 2.12 beta 24, this strongly suggests a software regression introduced in MentraOS 2.12 beta 25.