diff --git a/lib/features/chat/services/voice_call_service.dart b/lib/features/chat/services/voice_call_service.dart
index 388634e..bb786d4 100644
--- a/lib/features/chat/services/voice_call_service.dart
+++ b/lib/features/chat/services/voice_call_service.dart
@@ -343,20 +343,17 @@ class VoiceCallService {
 
       // Set up periodic keep-alive to refresh wake lock (every 5 minutes)
       _keepAliveTimer?.cancel();
-      _keepAliveTimer = Timer.periodic(
-        const Duration(minutes: 5),
-        (_) async {
-          final success = await BackgroundStreamingHandler.instance.keepAlive();
-          if (!success) {
-            // Keep-alive failed but don't stop the call - service may still work
-            developer.log(
-              'Voice call keep-alive failed',
-              name: 'VoiceCallService',
-              level: 900, // WARNING
-            );
-          }
-        },
-      );
+      _keepAliveTimer = Timer.periodic(const Duration(minutes: 5), (_) async {
+        final success = await BackgroundStreamingHandler.instance.keepAlive();
+        if (!success) {
+          // Keep-alive failed but don't stop the call - service may still work
+          developer.log(
+            'Voice call keep-alive failed',
+            name: 'VoiceCallService',
+            level: 900, // WARNING
+          );
+        }
+      });
 
       // Set up socket event listener for assistant responses
       _socketSubscription = _socketService.addChatEventHandler(
diff --git a/lib/shared/widgets/middle_ellipsis_text.dart b/lib/shared/widgets/middle_ellipsis_text.dart
index 9d353fd..58f8fc4 100644
--- a/lib/shared/widgets/middle_ellipsis_text.dart
+++ b/lib/shared/widgets/middle_ellipsis_text.dart
@@ -2,6 +2,9 @@ import 'package:flutter/widgets.dart';
 
 /// A single-line text widget that truncates the middle of long strings
 /// with an ellipsis (e.g., "prefix…suffix") so both ends remain visible.
+///
+/// This widget handles Unicode text safely, including emojis and other
+/// characters that span multiple UTF-16 code units (surrogate pairs).
 class MiddleEllipsisText extends StatelessWidget {
   final String text;
   final TextStyle? style;
@@ -20,6 +23,9 @@ class MiddleEllipsisText extends StatelessWidget {
 
   @override
   Widget build(BuildContext context) {
+    // Sanitize text to remove any unpaired surrogates that could cause crashes.
+    final String safeText = _sanitizeUtf16(text);
+
     return LayoutBuilder(
       builder: (context, constraints) {
         final TextStyle effectiveStyle = DefaultTextStyle.of(
@@ -29,7 +35,7 @@ class MiddleEllipsisText extends StatelessWidget {
         final double maxWidth = constraints.maxWidth;
 
         // Measure full text width first.
-        final fullSpan = TextSpan(text: text, style: effectiveStyle);
+        final fullSpan = TextSpan(text: safeText, style: effectiveStyle);
         final fullPainter = TextPainter(
           text: fullSpan,
           textDirection: direction,
@@ -38,7 +44,7 @@ class MiddleEllipsisText extends StatelessWidget {
 
         if (fullPainter.width <= maxWidth) {
           return Text(
-            text,
+            safeText,
             style: effectiveStyle,
             maxLines: 1,
             overflow: TextOverflow.clip,
@@ -47,6 +53,11 @@ class MiddleEllipsisText extends StatelessWidget {
           );
         }
 
+        // Use grapheme clusters (Characters) to safely split text without
+        // breaking surrogate pairs or emoji sequences.
+        final characters = safeText.characters;
+        final int totalGraphemes = characters.length;
+
         // Pre-measure ellipsis width (used implicitly during search).
         final ellipsisSpan = TextSpan(text: ellipsis, style: effectiveStyle);
         final ellipsisPainter = TextPainter(
@@ -56,24 +67,25 @@ class MiddleEllipsisText extends StatelessWidget {
         )..layout(minWidth: 0, maxWidth: double.infinity);
         final double _ = ellipsisPainter.width; // hint width; not used directly
 
-        // Binary search the maximum number of visible characters (k), split
+        // Binary search the maximum number of visible graphemes (k), split
         // between start and end. For a given k, we use ceil(k/2) from start
         // and floor(k/2) from end.
         int low = 0;
-        int high = text.length; // exclusive upper bound in practice
+        int high = totalGraphemes;
         int bestK = 0;
         String bestStart = '';
         String bestEnd = '';
 
         while (low <= high) {
-          final int k = (low + high) >> 1; // candidate visible char count
+          final int k = (low + high) >> 1; // candidate visible grapheme count
           final int leftCount = (k + 1) >> 1; // ceil(k/2)
           final int rightCount = k - leftCount; // floor(k/2)
 
-          final String start = text.substring(0, leftCount);
+          // Use Characters.take/takeLast to safely extract grapheme clusters.
+          final String start = characters.take(leftCount).toString();
           final String end = rightCount == 0
               ? ''
-              : text.substring(text.length - rightCount);
+              : characters.takeLast(rightCount).toString();
 
           final trialSpan = TextSpan(
             text: '$start$ellipsis$end',
@@ -102,7 +114,7 @@ class MiddleEllipsisText extends StatelessWidget {
             maxLines: 1,
             overflow: TextOverflow.clip,
             textAlign: textAlign,
-            semanticsLabel: semanticsLabel ?? text,
+            semanticsLabel: semanticsLabel ?? safeText,
           );
         }
 
@@ -113,9 +125,48 @@ class MiddleEllipsisText extends StatelessWidget {
           maxLines: 1,
           overflow: TextOverflow.clip,
           textAlign: textAlign,
-          semanticsLabel: semanticsLabel ?? text,
+          semanticsLabel: semanticsLabel ?? safeText,
         );
       },
     );
   }
+
+  /// Removes unpaired UTF-16 surrogates that would cause "not well-formed
+  /// UTF-16" errors during text layout.
+  ///
+  /// A valid UTF-16 string requires:
+  /// - High surrogates (0xD800-0xDBFF) must be followed by low surrogates
+  /// - Low surrogates (0xDC00-0xDFFF) must be preceded by high surrogates
+  static String _sanitizeUtf16(String input) {
+    if (input.isEmpty) return input;
+
+    final buffer = StringBuffer();
+    for (int i = 0; i < input.length; i++) {
+      final int codeUnit = input.codeUnitAt(i);
+
+      // Check if this is a high surrogate (0xD800-0xDBFF)
+      if (codeUnit >= 0xD800 && codeUnit <= 0xDBFF) {
+        // Check if next character is a valid low surrogate
+        if (i + 1 < input.length) {
+          final int nextCodeUnit = input.codeUnitAt(i + 1);
+          if (nextCodeUnit >= 0xDC00 && nextCodeUnit <= 0xDFFF) {
+            // Valid surrogate pair - include both
+            buffer.writeCharCode(codeUnit);
+            buffer.writeCharCode(nextCodeUnit);
+            i++; // Skip the low surrogate in next iteration
+            continue;
+          }
+        }
+        // Unpaired high surrogate - replace with replacement character
+        buffer.writeCharCode(0xFFFD);
+      } else if (codeUnit >= 0xDC00 && codeUnit <= 0xDFFF) {
+        // Unpaired low surrogate - replace with replacement character
+        buffer.writeCharCode(0xFFFD);
+      } else {
+        // Regular character - include as-is
+        buffer.writeCharCode(codeUnit);
+      }
+    }
+    return buffer.toString();
+  }
 }