diff --git a/src/App.vue b/src/App.vue
index cce05c1..abbb023 100644
--- a/src/App.vue
+++ b/src/App.vue
@@ -1014,8 +1014,8 @@ const sendWithEnter = ref(loadBoolPref(SEND_WITH_ENTER_KEY, true))
 const inProgressSendMode = ref<'steer' | 'queue'>(loadInProgressSendModePref())
 const darkMode = ref<'system' | 'light' | 'dark'>(loadDarkModePref())
 const chatWidth = ref<ChatWidthMode>(loadChatWidthPref())
-const dictationClickToToggle = ref(loadBoolPref(DICTATION_CLICK_TO_TOGGLE_KEY, false))
-const dictationAutoSend = ref(loadBoolPref(DICTATION_AUTO_SEND_KEY, true))
+const dictationClickToToggle = ref(loadBoolPref(DICTATION_CLICK_TO_TOGGLE_KEY, true))
+const dictationAutoSend = ref(loadBoolPref(DICTATION_AUTO_SEND_KEY, false))
 const dictationLanguage = ref(loadDictationLanguagePref())
 const dictationLanguageOptions = computed(() => buildDictationLanguageOptions())
 
diff --git a/src/components/content/ThreadComposer.vue b/src/components/content/ThreadComposer.vue
index a98311d..a58ae78 100644
--- a/src/components/content/ThreadComposer.vue
+++ b/src/components/content/ThreadComposer.vue
@@ -292,11 +292,12 @@
             class="thread-composer-mic"
             :class="{
               'thread-composer-mic--active': dictationState === 'recording',
+              'thread-composer-mic--transcribing': dictationState === 'transcribing',
             }"
             type="button"
             :aria-label="dictationButtonLabel"
             :title="dictationButtonLabel"
-            :disabled="isInteractionDisabled"
+            :disabled="isInteractionDisabled || dictationState === 'transcribing'"
             @click="onDictationToggle"
             @pointerdown="onDictationPressStart"
             @pointerup="onDictationPressEnd"
@@ -306,6 +307,7 @@
               v-if="dictationState === 'recording'"
               class="thread-composer-mic-icon thread-composer-mic-icon--stop"
             />
+            <span v-else-if="dictationState === 'transcribing'" class="thread-composer-mic-spinner" aria-hidden="true" />
             <IconTablerMicrophone v-else class="thread-composer-mic-icon" />
           </button>
 
@@ -326,10 +328,10 @@
             class="thread-composer-submit"
             :class="{ 'thread-composer-submit--queue': isTurnInProgress && activeInProgressMode === 'queue' }"
             type="button"
-            :aria-label="isTurnInProgress && activeInProgressMode === 'queue' ? 'Queue message' : 'Send message'"
-            :title="isTurnInProgress ? `Send as ${activeInProgressMode}` : 'Send'"
-            :disabled="!canSubmit"
-            @click="onSubmit(isTurnInProgress ? activeInProgressMode : 'steer')"
+            :aria-label="submitButtonAriaLabel"
+            :title="submitButtonTitle"
+            :disabled="isSubmitButtonDisabled"
+            @click="onPrimarySubmit(isTurnInProgress ? activeInProgressMode : 'steer')"
           >
             <IconTablerArrowUp class="thread-composer-submit-icon" />
           </button>
@@ -496,12 +498,12 @@ const {
   cancel: cancelDictation,
 } = useDictation({
   getLanguage: () => props.dictationLanguage ?? 'auto',
-  onTranscript: (text) => {
+  onTranscript: (text, mode) => {
     draft.value = draft.value ? `${draft.value}\n${text}` : text
     dictationFeedback.value = ''
-    if (props.dictationAutoSend !== false) {
-      const mode = props.isTurnInProgress ? activeInProgressMode.value : 'steer'
-      onSubmit(mode)
+    if (mode === 'send' || props.dictationAutoSend !== false) {
+      const submitMode = props.isTurnInProgress ? activeInProgressMode.value : 'steer'
+      onSubmit(submitMode)
       return
     }
     nextTick(() => inputRef.value?.focus())
@@ -581,6 +583,14 @@ const canSubmit = computed(() => {
   if (pendingAttachmentCount.value > 0) return false
   return draft.value.trim().length > 0 || selectedImages.value.length > 0 || fileAttachments.value.length > 0
 })
+const canRequestDictationSend = computed(() =>
+  dictationState.value === 'recording'
+  && !props.disabled
+  && !props.isUpdatingSpeedMode
+  && !!props.activeThreadId
+  && !isPlanModeWaitingForModel.value
+  && pendingAttachmentCount.value <= 0,
+)
 const hasUnsavedDraft = computed(() =>
   draft.value.trim().length > 0
   || selectedImages.value.length > 0
@@ -618,6 +628,7 @@ const activeInProgressMode = ref<'steer' | 'queue'>(inProgressMode.value)
 const isDictationRecording = computed(() => dictationState.value === 'recording')
 const dictationButtonLabel = computed(() => {
   if (dictationState.value === 'recording') return 'Stop dictation'
+  if (dictationState.value === 'transcribing') return 'Transcribing dictation'
   return props.dictationClickToToggle ? 'Click to dictate' : 'Hold to dictate'
 })
 const dictationErrorText = computed(() =>
@@ -652,6 +663,21 @@ const dictationDurationLabel = computed(() => {
   const seconds = totalSeconds % 60
   return `${minutes}:${String(seconds).padStart(2, '0')}`
 })
+const submitButtonAriaLabel = computed(() => {
+  if (dictationState.value === 'recording') return 'Transcribe and send'
+  if (dictationState.value === 'transcribing') return 'Transcribing dictation'
+  return props.isTurnInProgress && activeInProgressMode.value === 'queue' ? 'Queue message' : 'Send message'
+})
+const submitButtonTitle = computed(() => {
+  if (dictationState.value === 'recording') return 'Transcribe and send'
+  if (dictationState.value === 'transcribing') return 'Transcribing dictation'
+  return props.isTurnInProgress ? `Send as ${activeInProgressMode.value}` : 'Send'
+})
+const isSubmitButtonDisabled = computed(() => {
+  if (dictationState.value === 'transcribing') return true
+  if (canRequestDictationSend.value) return false
+  return !canSubmit.value
+})
 
 const placeholderText = computed(() =>
   !props.activeThreadId
@@ -1033,6 +1059,16 @@ function onInterrupt(): void {
   emit('interrupt')
 }
 
+function onPrimarySubmit(mode: 'steer' | 'queue' = 'steer'): void {
+  if (dictationState.value === 'recording') {
+    if (!canRequestDictationSend.value) return
+    stopRecording('send')
+    return
+  }
+  if (dictationState.value === 'transcribing') return
+  onSubmit(mode)
+}
+
 function onModelSelect(value: string): void {
   emit('update:selected-model', value)
 }
@@ -2107,10 +2143,18 @@ watch(
   @apply bg-red-100 text-red-600 hover:bg-red-200 hover:text-red-700;
 }
 
+.thread-composer-mic--transcribing {
+  @apply bg-zinc-200 text-zinc-600;
+}
+
 .thread-composer-mic-icon {
   @apply h-5 w-5;
 }
 
+.thread-composer-mic-spinner {
+  @apply block h-4 w-4 rounded-full border-2 border-current border-t-transparent animate-spin;
+}
+
 .thread-composer-dictation-waveform-wrap {
   @apply min-w-0 flex-1;
 }
diff --git a/src/composables/useDictation.ts b/src/composables/useDictation.ts
index 47f8388..f8047f6 100644
--- a/src/composables/useDictation.ts
+++ b/src/composables/useDictation.ts
@@ -1,13 +1,14 @@
 import { onBeforeUnmount, ref } from 'vue'
 
 export type DictationState = 'idle' | 'recording' | 'transcribing'
+export type DictationStopMode = 'insert' | 'send'
 const DICTATION_SILENCE_THRESHOLD = 0.0025
 const DICTATION_BAR_WIDTH = 3
 const DICTATION_BAR_GAP = 2
 const MAX_WAVEFORM_SAMPLES = 256
 
 export function useDictation(options: {
-  onTranscript: (text: string) => void
+  onTranscript: (text: string, mode: DictationStopMode) => void
   getLanguage?: () => string
   onEmpty?: () => void
   onError?: (error: unknown) => void
@@ -28,6 +29,7 @@ export function useDictation(options: {
   let isStartingRecording = false
   let stopRequestedBeforeStart = false
   let transcribeAbortController: AbortController | null = null
+  let pendingStopMode: DictationStopMode = 'insert'
 
   function cancelTranscription(): void {
     if (transcribeAbortController) {
@@ -160,8 +162,10 @@ export function useDictation(options: {
       mediaRecorder.onstop = () => {
         const recordedChunks = chunks
         const recordedMimeType = mediaRecorder?.mimeType || recordedChunks[0]?.type || 'audio/webm'
+        const stopMode = pendingStopMode
+        pendingStopMode = 'insert'
         cleanup()
-        void transcribe(recordedChunks, recordedMimeType)
+        void transcribe(recordedChunks, recordedMimeType, stopMode)
       }
       startWaveformCapture(mediaStream)
       mediaRecorder.start(250)
@@ -178,13 +182,15 @@ export function useDictation(options: {
     }
   }
 
-  function stopRecording() {
+  function stopRecording(mode: DictationStopMode = 'insert') {
     if (isStartingRecording && state.value === 'idle') {
       stopRequestedBeforeStart = true
+      pendingStopMode = mode
       return
     }
     if (state.value !== 'recording' || !mediaRecorder) return
     if (mediaRecorder.state !== 'inactive') {
+      pendingStopMode = mode
       state.value = 'transcribing'
       try {
         mediaRecorder.requestData()
@@ -202,7 +208,7 @@ export function useDictation(options: {
     state.value = 'idle'
   }
 
-  async function transcribe(recordedChunks: Blob[], mimeType: string) {
+  async function transcribe(recordedChunks: Blob[], mimeType: string, mode: DictationStopMode) {
     if (recordedChunks.length === 0) {
       options.onEmpty?.()
       state.value = 'idle'
@@ -246,7 +252,7 @@ export function useDictation(options: {
 
       const text = (data?.text ?? '').trim()
       if (text.length > 0) {
-        options.onTranscript(text)
+        options.onTranscript(text, mode)
       } else {
         options.onEmpty?.()
       }
diff --git a/tests.md b/tests.md
index 6228857..d195a3a 100644
--- a/tests.md
+++ b/tests.md
@@ -2232,3 +2232,27 @@ Toggle "Free mode" in settings to use free OpenRouter models without an OpenAI A
 
 #### Rollback/Cleanup
 - Run `codexui login` to restore Codex authentication if needed.
+
+### Feature: Tap-to-toggle dictation with optional send-on-stop flow
+
+#### Prerequisites
+- App is running from this repository.
+- The browser has microphone permission.
+- A thread is selected so the composer is enabled.
+
+#### Steps
+1. Open Settings and confirm `Click to toggle dictation` is enabled by default and `Auto send dictation` is disabled by default.
+2. In the composer, click the microphone once and verify recording starts without holding the button.
+3. Speak a short phrase, then click the microphone again to stop recording.
+4. While transcription is processing, confirm the microphone button shows a loading spinner.
+5. Confirm the transcript is inserted into the composer without being auto-sent.
+6. Start dictation again, speak another short phrase, and click the send arrow instead of the microphone stop button.
+
+#### Expected Results
+- Dictation starts on first click and keeps recording hands-free until the next click.
+- Stopping with the microphone inserts the transcript into the composer and leaves it editable.
+- The microphone shows a transcribing spinner while `/codex-api/transcribe` is still in progress.
+- Clicking the send arrow during recording transcribes the audio and submits the message automatically.
+
+#### Rollback/Cleanup
+- Clear the draft or delete the test thread message if needed.