diff --git a/android/app/src/main/java/com/soniox/remotemic/MainActivity.kt b/android/app/src/main/java/com/soniox/remotemic/MainActivity.kt
index 8bbdcb4..98742d5 100644
--- a/android/app/src/main/java/com/soniox/remotemic/MainActivity.kt
+++ b/android/app/src/main/java/com/soniox/remotemic/MainActivity.kt
@@ -354,12 +354,16 @@ class MainActivity : AppCompatActivity() {
         return when (event.action) {
             MotionEvent.ACTION_DOWN -> {
                 if (checkAudioPermission()) {
+                    // Fast sync: ensure socket is connected immediately
+                    streamDictationClient?.connectWebSocket()
+                    
                     val selStart = binding.etTranscript.selectionStart
                     val selEnd = binding.etTranscript.selectionEnd
-                    val totalLen = binding.etTranscript.text?.length ?: 0
+                    val currentText = binding.etTranscript.text?.toString() ?: ""
+                    val totalLen = currentText.length
                     
-                    voiceInsertionCursorStart = if (selStart >= 0) selStart else totalLen
-                    voiceInsertionCursorEnd = if (selEnd >= 0) selEnd else totalLen
+                    voiceInsertionCursorStart = if (selStart in 0..totalLen) selStart else totalLen
+                    voiceInsertionCursorEnd = if (selEnd in 0..totalLen) selEnd else totalLen
                     
                     startRecording()
                 }
@@ -367,6 +371,16 @@ class MainActivity : AppCompatActivity() {
             }
             MotionEvent.ACTION_UP, MotionEvent.ACTION_CANCEL -> {
                 if (isCurrentlyRecording) {
+                    // Update insertion point to where cursor was right before processing
+                    val selStart = binding.etTranscript.selectionStart
+                    val selEnd = binding.etTranscript.selectionEnd
+                    val currentText = binding.etTranscript.text?.toString() ?: ""
+                    val totalLen = currentText.length
+                    
+                    if (selStart in 0..totalLen && selEnd in 0..totalLen) {
+                        voiceInsertionCursorStart = selStart
+                        voiceInsertionCursorEnd = selEnd
+                    }
                     stopRecordingAndProcess()
                 }
                 true