diff --git a/src/api/generated/data-contracts.ts b/src/api/generated/data-contracts.ts
index 157dc0f..985eb1b 100644
--- a/src/api/generated/data-contracts.ts
+++ b/src/api/generated/data-contracts.ts
@@ -964,6 +964,15 @@ export type ServerVadConfig = (
 ) & {
   /** Optional Smart Turn endpoint detection configuration. Runs after VAD silence detection to verify turn completion. */
   smartTurn?: SmartTurnConfig;
+  /**
+   * Duration in milliseconds to wait for the user to continue speaking after a barge-in interrupt. If silence is detected for this duration, ASR is stopped. Default: 3000.
+   * @min 500
+   * @max 10000
+   * @default 3000
+   */
+  bargeInSilenceTimeout?: number;
+  /** Optional placeholder text fed to the AI as user input when the user barge-ins but then stops speaking before the bargeInSilenceTimeout. The AI generates a response based on this prompt (e.g. "[you misheard something the user said]"). Default: [repeat after interruption]. */
+  bargeInSilencePlaceholder?: string;
 };
 
 export interface LegacyVadConfig {
diff --git a/src/api/openapi.json b/src/api/openapi.json
index 4eb4f7a..c550744 100644
--- a/src/api/openapi.json
+++ b/src/api/openapi.json
@@ -1557,6 +1557,17 @@
             "properties": {
               "smartTurn": {
                 "$ref": "#/components/schemas/SmartTurnConfig"
+              },
+              "bargeInSilenceTimeout": {
+                "type": "integer",
+                "minimum": 500,
+                "maximum": 10000,
+                "default": 3000,
+                "description": "Duration in milliseconds to wait for the user to continue speaking after a barge-in interrupt. If silence is detected for this duration, ASR is stopped. Default: 3000."
+              },
+              "bargeInSilencePlaceholder": {
+                "type": "string",
+                "description": "Optional placeholder text fed to the AI as user input when the user barge-ins but then stops speaking before the bargeInSilenceTimeout. The AI generates a response based on this prompt (e.g. \"[you misheard something the user said]\"). Default: [repeat after interruption]."
               }
             }
           }
diff --git a/src/api/websocket/websocket-contracts.json b/src/api/websocket/websocket-contracts.json
index 5060473..8d307f4 100644
--- a/src/api/websocket/websocket-contracts.json
+++ b/src/api/websocket/websocket-contracts.json
@@ -778,6 +778,17 @@
           "properties": {
             "smartTurn": {
               "$ref": "#/definitions/SmartTurnConfig"
+            },
+            "bargeInSilenceTimeout": {
+              "type": "integer",
+              "minimum": 500,
+              "maximum": 10000,
+              "default": 3000,
+              "description": "Duration in milliseconds to wait for the user to continue speaking after a barge-in interrupt. If silence is detected for this duration, ASR is stopped. Default: 3000."
+            },
+            "bargeInSilencePlaceholder": {
+              "type": "string",
+              "description": "Optional placeholder text fed to the AI as user input when the user barge-ins but then stops speaking before the bargeInSilenceTimeout. The AI generates a response based on this prompt (e.g. \"[you misheard something the user said]\"). Default: [repeat after interruption]."
             }
           }
         }
diff --git a/src/api/websocket/websocket-contracts.ts b/src/api/websocket/websocket-contracts.ts
index 0cf4998..87d2e04 100644
--- a/src/api/websocket/websocket-contracts.ts
+++ b/src/api/websocket/websocket-contracts.ts
@@ -407,6 +407,14 @@ export interface SmartTurnConfig {
  */
 export type ServerVadConfig = (LegacyVadConfig | SileroVadConfig | FireRedVadConfig) & {
   smartTurn?: SmartTurnConfig;
+  /**
+   * Duration in milliseconds to wait for the user to continue speaking after a barge-in interrupt. If silence is detected for this duration, ASR is stopped. Default: 3000.
+   */
+  bargeInSilenceTimeout?: number;
+  /**
+   * Optional placeholder text fed to the AI as user input when the user barge-ins but then stops speaking before the bargeInSilenceTimeout. The AI generates a response based on this prompt (e.g. "[you misheard something the user said]"). Default: [repeat after interruption].
+   */
+  bargeInSilencePlaceholder?: string;
 };
 
 
diff --git a/src/components/modals/ServerVadSettingsModal.vue b/src/components/modals/ServerVadSettingsModal.vue
index 965ffc3..86e90e7 100644
--- a/src/components/modals/ServerVadSettingsModal.vue
+++ b/src/components/modals/ServerVadSettingsModal.vue
@@ -3,76 +3,32 @@
     <div class="modal-content" @click.stop>
       <h2 class="modal-header">Server-side VAD Settings</h2>
 
-      <form @submit.prevent="handleSubmit">
-        <p class="form-help-text mb-4">
-          Server autonomously detects speech boundaries — clients stream audio continuously without calling start/end voice input. This feature is experimental and may behave unexpectedly.
-        </p>
-
-        <FormField label="Algorithm" help="VAD algorithm to use for speech detection">
-          <select
-            v-model="form.algorithm"
-            class="form-select-auto min-w-48"
-            @change="onAlgorithmChange"
-          >
-            <option value="legacy">Silero (basic)</option>
-            <option value="silero">Silero (advanced)</option>
-            <option value="firered">FireRedVAD</option>
-          </select>
-        </FormField>
-
-        <template v-if="form.algorithm === 'legacy'">
-          <FormField label="Aggressiveness Mode" help="Higher values filter non-speech more aggressively (0–3, default: 2)">
-            <select
-              v-model.number="legacyForm.mode"
-              class="form-select-auto min-w-48"
-            >
-              <option :value="undefined">Default (2)</option>
-              <option :value="0">0 — Least aggressive</option>
-              <option :value="1">1</option>
-              <option :value="2">2 (default)</option>
-              <option :value="3">3 — Most aggressive</option>
-            </select>
-          </FormField>
+      <p class="form-help-text mb-4">
+        Server autonomously detects speech boundaries — clients stream audio continuously without calling start/end voice input. This feature is experimental and may behave unexpectedly.
+      </p>
 
-          <FormField label="Frame Duration" help="Duration of each VAD processing frame — must be 10, 20, or 30 ms (default: 20)">
+      <div class="border-b border-gray-200 dark:border-gray-700 mb-4">
+        <TabNavigator v-model="activeTab" :tabs="tabs" />
+      </div>
+
+      <form @submit.prevent="handleSubmit">
+        <!-- General tab: algorithm + grace period -->
+        <TabContent v-model="activeTab" tab="general">
+          <FormField label="Algorithm" help="VAD algorithm to use for speech detection">
             <select
-              v-model.number="legacyForm.frameDurationMs"
+              v-model="form.algorithm"
               class="form-select-auto min-w-48"
+              @change="onAlgorithmChange"
             >
-              <option :value="undefined">Default (20 ms)</option>
-              <option :value="10">10 ms</option>
-              <option :value="20">20 ms (default)</option>
-              <option :value="30">30 ms</option>
+              <option value="legacy">Silero (basic)</option>
+              <option value="silero">Silero (advanced)</option>
+              <option value="firered">FireRedVAD</option>
             </select>
           </FormField>
 
-          <FormField label="Silence Pre-roll Padding (ms)" help="Silence prepended before detected speech as a pre-roll buffer (0–1000 ms, default: 300)">
-            <input
-              v-model.number="legacyForm.silencePaddingMs"
-              type="number"
-              min="0"
-              max="1000"
-              step="10"
-              placeholder="300"
-              class="form-input max-w-xs"
-            />
-          </FormField>
-
-          <FormField label="Auto-End Silence Duration (ms)" help="Silence after speech that triggers end-of-utterance detection (100–5000 ms, default: 800)">
-            <input
-              v-model.number="legacyForm.autoEndSilenceDurationMs"
-              type="number"
-              min="100"
-              max="5000"
-              step="50"
-              placeholder="800"
-              class="form-input max-w-xs"
-            />
-          </FormField>
-
           <FormField label="Grace Period (ms)" help="Duration after VAD initialization during which speech_start is suppressed. Prevents false positives from phone connection noise (0–5000 ms, default: 1000)">
             <input
-              v-model.number="legacyForm.gracePeriodMs"
+              v-model.number="generalForm.gracePeriodMs"
               type="number"
               min="0"
               max="5000"
@@ -81,194 +37,225 @@
               class="form-input max-w-xs"
             />
           </FormField>
-        </template>
+        </TabContent>
+
+        <!-- Settings tab: algorithm-specific settings -->
+        <TabContent v-model="activeTab" tab="settings">
+          <template v-if="form.algorithm === 'legacy'">
+            <FormField label="Aggressiveness Mode" help="Higher values filter non-speech more aggressively (0–3, default: 2)">
+              <select
+                v-model.number="legacyForm.mode"
+                class="form-select-auto min-w-48"
+              >
+                <option :value="undefined">Default (2)</option>
+                <option :value="0">0 — Least aggressive</option>
+                <option :value="1">1</option>
+                <option :value="2">2 (default)</option>
+                <option :value="3">3 — Most aggressive</option>
+              </select>
+            </FormField>
 
-        <template v-if="form.algorithm === 'silero'">
-          <FormField label="Model Version" help="Silero VAD model version. v5 is the latest; legacy is the older model (default: v5)">
-            <select
-              v-model="sileroForm.model"
-              class="form-select-auto min-w-48"
-            >
-              <option :value="undefined">Default (v5)</option>
-              <option value="v5">v5 (default)</option>
-              <option value="legacy">legacy</option>
-            </select>
-          </FormField>
+            <FormField label="Frame Duration" help="Duration of each VAD processing frame — must be 10, 20, or 30 ms (default: 20)">
+              <select
+                v-model.number="legacyForm.frameDurationMs"
+                class="form-select-auto min-w-48"
+              >
+                <option :value="undefined">Default (20 ms)</option>
+                <option :value="10">10 ms</option>
+                <option :value="20">20 ms (default)</option>
+                <option :value="30">30 ms</option>
+              </select>
+            </FormField>
 
-          <FormField label="Positive Speech Threshold" help="Probability threshold above which a frame is considered speech (0–1, default: 0.5)">
-            <input
-              v-model.number="sileroForm.positiveSpeechThreshold"
-              type="number"
-              min="0"
-              max="1"
-              step="0.01"
-              placeholder="0.5"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Silence Pre-roll Padding (ms)" help="Silence prepended before detected speech as a pre-roll buffer (0–1000 ms, default: 300)">
+              <input
+                v-model.number="legacyForm.silencePaddingMs"
+                type="number"
+                min="0"
+                max="1000"
+                step="10"
+                placeholder="300"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Negative Speech Threshold" help="Probability threshold below which a frame is considered silence (0–1, default: 0.35)">
-            <input
-              v-model.number="sileroForm.negativeSpeechThreshold"
-              type="number"
-              min="0"
-              max="1"
-              step="0.01"
-              placeholder="0.35"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Auto-End Silence Duration (ms)" help="Silence after speech that triggers end-of-utterance detection (100–5000 ms, default: 800)">
+              <input
+                v-model.number="legacyForm.autoEndSilenceDurationMs"
+                type="number"
+                min="100"
+                max="5000"
+                step="50"
+                placeholder="800"
+                class="form-input max-w-xs"
+              />
+            </FormField>
+          </template>
 
-          <FormField label="Frame Samples" help="Number of audio samples per VAD frame. Silero was trained on 512, 1024, 1536 samples at 16kHz (default: 512)">
-            <input
-              v-model.number="sileroForm.frameSamples"
-              type="number"
-              min="1"
-              step="1"
-              placeholder="512"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+          <template v-if="form.algorithm === 'silero'">
+            <FormField label="Model Version" help="Silero VAD model version. v5 is the latest; legacy is the older model (default: v5)">
+              <select
+                v-model="sileroForm.model"
+                class="form-select-auto min-w-48"
+              >
+                <option :value="undefined">Default (v5)</option>
+                <option value="v5">v5 (default)</option>
+                <option value="legacy">legacy</option>
+              </select>
+            </FormField>
 
-          <FormField label="Redemption Frames" help="Number of silent frames after speech before end-of-utterance is triggered. If speech resumes during this window, the utterance is not ended (default: 8)">
-            <input
-              v-model.number="sileroForm.redemptionFrames"
-              type="number"
-              min="1"
-              step="1"
-              placeholder="8"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Positive Speech Threshold" help="Probability threshold above which a frame is considered speech (0–1, default: 0.5)">
+              <input
+                v-model.number="sileroForm.positiveSpeechThreshold"
+                type="number"
+                min="0"
+                max="1"
+                step="0.01"
+                placeholder="0.5"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Pre-Speech Pad Frames" help="Number of frames of pre-roll silence prepended to the audio segment on speech start (default: 1)">
-            <input
-              v-model.number="sileroForm.preSpeechPadFrames"
-              type="number"
-              min="0"
-              step="1"
-              placeholder="1"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Negative Speech Threshold" help="Probability threshold below which a frame is considered silence (0–1, default: 0.35)">
+              <input
+                v-model.number="sileroForm.negativeSpeechThreshold"
+                type="number"
+                min="0"
+                max="1"
+                step="0.01"
+                placeholder="0.35"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Min Speech Frames" help="Minimum frames required to consider a segment as speech. Shorter segments trigger onVADMisfire instead (default: 3)">
-            <input
-              v-model.number="sileroForm.minSpeechFrames"
-              type="number"
-              min="1"
-              step="1"
-              placeholder="3"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Frame Samples" help="Number of audio samples per VAD frame. Silero was trained on 512, 1024, 1536 samples at 16kHz (default: 512)">
+              <input
+                v-model.number="sileroForm.frameSamples"
+                type="number"
+                min="1"
+                step="1"
+                placeholder="512"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Submit User Speech On Pause" help="Whether to submit partial speech when VAD is paused">
-            <select
-              v-model="sileroForm.submitUserSpeechOnPause"
-              class="form-select-auto min-w-48"
-            >
-              <option :value="undefined">Default</option>
-              <option :value="true">Enabled</option>
-              <option :value="false">Disabled</option>
-            </select>
-          </FormField>
+            <FormField label="Redemption Frames" help="Number of silent frames after speech before end-of-utterance is triggered. If speech resumes during this window, the utterance is not ended (default: 8)">
+              <input
+                v-model.number="sileroForm.redemptionFrames"
+                type="number"
+                min="1"
+                step="1"
+                placeholder="8"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Grace Period (ms)" help="Duration after VAD initialization during which speech_start is suppressed. Prevents false positives from phone connection noise (0–5000 ms, default: 1000)">
-            <input
-              v-model.number="sileroForm.gracePeriodMs"
-              type="number"
-              min="0"
-              max="5000"
-              step="100"
-              placeholder="1000"
-              class="form-input max-w-xs"
-            />
-          </FormField>
-        </template>
+            <FormField label="Pre-Speech Pad Frames" help="Number of frames of pre-roll silence prepended to the audio segment on speech start (default: 1)">
+              <input
+                v-model.number="sileroForm.preSpeechPadFrames"
+                type="number"
+                min="0"
+                step="1"
+                placeholder="1"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-        <template v-if="form.algorithm === 'firered'">
-          <FormField label="Speech Threshold" help="Probability threshold above which a smoothed frame is classified as speech (0–1, default: 0.5)">
-            <input
-              v-model.number="fireredForm.speechThreshold"
-              type="number"
-              min="0"
-              max="1"
-              step="0.01"
-              placeholder="0.5"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Min Speech Frames" help="Minimum frames required to consider a segment as speech. Shorter segments trigger onVADMisfire instead (default: 3)">
+              <input
+                v-model.number="sileroForm.minSpeechFrames"
+                type="number"
+                min="1"
+                step="1"
+                placeholder="3"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Smooth Window Size" help="Size of the moving-average smoothing window applied to raw frame probabilities (min: 1, default: 5)">
-            <input
-              v-model.number="fireredForm.smoothWindowSize"
-              type="number"
-              min="1"
-              step="1"
-              placeholder="5"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Submit User Speech On Pause" help="Whether to submit partial speech when VAD is paused">
+              <select
+                v-model="sileroForm.submitUserSpeechOnPause"
+                class="form-select-auto min-w-48"
+              >
+                <option :value="undefined">Default</option>
+                <option :value="true">Enabled</option>
+                <option :value="false">Disabled</option>
+              </select>
+            </FormField>
+          </template>
 
-          <FormField label="Min Speech Frames" help="Minimum consecutive speech frames required before speech_start is emitted (min: 1, default: 8)">
-            <input
-              v-model.number="fireredForm.minSpeechFrame"
-              type="number"
-              min="1"
-              step="1"
-              placeholder="8"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+          <template v-if="form.algorithm === 'firered'">
+            <FormField label="Speech Threshold" help="Probability threshold above which a smoothed frame is classified as speech (0–1, default: 0.5)">
+              <input
+                v-model.number="fireredForm.speechThreshold"
+                type="number"
+                min="0"
+                max="1"
+                step="0.01"
+                placeholder="0.5"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Max Speech Frames" help="Maximum consecutive speech frames before a forced speech_end (long-utterance cutoff) (min: 1, default: 6000)">
-            <input
-              v-model.number="fireredForm.maxSpeechFrame"
-              type="number"
-              min="1"
-              step="1"
-              placeholder="6000"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Smooth Window Size" help="Size of the moving-average smoothing window applied to raw frame probabilities (min: 1, default: 5)">
+              <input
+                v-model.number="fireredForm.smoothWindowSize"
+                type="number"
+                min="1"
+                step="1"
+                placeholder="5"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Min Silence Frames" help="Minimum consecutive silence frames after speech before speech_end is emitted (min: 1, default: 80)">
-            <input
-              v-model.number="fireredForm.minSilenceFrame"
-              type="number"
-              min="1"
-              step="1"
-              placeholder="80"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Min Speech Frames" help="Minimum consecutive speech frames required before speech_start is emitted (min: 1, default: 8)">
+              <input
+                v-model.number="fireredForm.minSpeechFrame"
+                type="number"
+                min="1"
+                step="1"
+                placeholder="8"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Pad Start Frames" help="Number of frames of pre-roll audio prepended to the detected speech start (min: 0, default: 5)">
-            <input
-              v-model.number="fireredForm.padStartFrame"
-              type="number"
-              min="0"
-              step="1"
-              placeholder="5"
-              class="form-input max-w-xs"
-            />
-          </FormField>
+            <FormField label="Max Speech Frames" help="Maximum consecutive speech frames before a forced speech_end (long-utterance cutoff) (min: 1, default: 6000)">
+              <input
+                v-model.number="fireredForm.maxSpeechFrame"
+                type="number"
+                min="1"
+                step="1"
+                placeholder="6000"
+                class="form-input max-w-xs"
+              />
+            </FormField>
 
-          <FormField label="Grace Period (ms)" help="Duration after VAD initialization during which speech_start is suppressed. Prevents false positives from phone connection noise (0–5000 ms, default: 1000)">
-            <input
-              v-model.number="fireredForm.gracePeriodMs"
-              type="number"
-              min="0"
-              max="5000"
-              step="100"
-              placeholder="1000"
-              class="form-input max-w-xs"
-            />
-          </FormField>
-        </template>
+            <FormField label="Min Silence Frames" help="Minimum consecutive silence frames after speech before speech_end is emitted (min: 1, default: 80)">
+              <input
+                v-model.number="fireredForm.minSilenceFrame"
+                type="number"
+                min="1"
+                step="1"
+                placeholder="80"
+                class="form-input max-w-xs"
+              />
+            </FormField>
+
+            <FormField label="Pad Start Frames" help="Number of frames of pre-roll audio prepended to the detected speech start (min: 0, default: 5)">
+              <input
+                v-model.number="fireredForm.padStartFrame"
+                type="number"
+                min="0"
+                step="1"
+                placeholder="5"
+                class="form-input max-w-xs"
+              />
+            </FormField>
+          </template>
+        </TabContent>
 
-        <div class="mt-6 pt-4 border-t border-gray-200 dark:border-gray-700">
+        <!-- Turn Detection tab -->
+        <TabContent v-model="activeTab" tab="turn-detection">
           <h3 class="text-sm font-semibold text-gray-700 dark:text-gray-300 mb-3">Smart Turn Detection</h3>
           <p class="form-help-text mb-4">
             Runs ONNX inference on the full utterance audio after VAD detects silence to verify the speaker has finished their turn. Helps reduce premature interruptions.
@@ -297,7 +284,31 @@
               />
             </FormField>
           </template>
-        </div>
+        </TabContent>
+
+        <!-- Silence tab -->
+        <TabContent v-model="activeTab" tab="silence">
+          <FormField label="Barge-in Silence Timeout (ms)" help="Duration to wait for the user to continue speaking after a barge-in interrupt. If silence is detected for this duration, ASR is stopped (500–10000 ms, default: 3000)">
+            <input
+              v-model.number="silenceForm.bargeInSilenceTimeout"
+              type="number"
+              min="500"
+              max="10000"
+              step="100"
+              placeholder="3000"
+              class="form-input max-w-xs"
+            />
+          </FormField>
+
+          <FormField label="Barge-in Silence Placeholder" help="Optional placeholder text fed to the AI as user input when the user barge-ins but then stops speaking before the bargeInSilenceTimeout. The AI generates a response based on this prompt (e.g. &quot;[you misheard something the user said]&quot;). Default: [repeat after interruption].">
+            <input
+              v-model="silenceForm.bargeInSilencePlaceholder"
+              type="text"
+              class="form-input"
+              placeholder="[repeat after interruption]"
+            />
+          </FormField>
+        </TabContent>
 
         <div class="modal-footer">
           <button type="button" @click="$emit('close')" class="btn-secondary">
@@ -316,6 +327,9 @@
 import { ref, watch } from 'vue'
 import type { FireRedVadConfig, LegacyVadConfig, ServerVadConfig, SileroVadConfig, SmartTurnConfig } from '@/api/types'
 import FormField from '@/components/FormField.vue'
+import TabNavigator from '@/components/TabNavigator.vue'
+import type { TabDefinition } from '@/components/TabNavigator.vue'
+import TabContent from '@/components/TabContent.vue'
 
 const props = defineProps<{
   config: ServerVadConfig
@@ -326,19 +340,31 @@ const emit = defineEmits<{
   save: [config: ServerVadConfig]
 }>()
 
+const activeTab = ref<'general' | 'settings' | 'turn-detection' | 'silence'>('general')
+
+const tabs: TabDefinition[] = [
+  { key: 'general', label: 'General' },
+  { key: 'settings', label: 'Settings' },
+  { key: 'turn-detection', label: 'Turn Detection' },
+  { key: 'silence', label: 'Silence' },
+]
+
 const form = ref<{ algorithm: 'legacy' | 'silero' | 'firered' }>({
   algorithm: 'legacy'
 })
 
-const legacyForm = ref<Omit<LegacyVadConfig, 'algorithm'>>({
+const generalForm = ref({
+  gracePeriodMs: undefined as number | undefined,
+})
+
+const legacyForm = ref<Omit<LegacyVadConfig, 'algorithm' | 'gracePeriodMs'>>({
   mode: undefined,
   frameDurationMs: undefined,
   silencePaddingMs: undefined,
   autoEndSilenceDurationMs: undefined,
-  gracePeriodMs: undefined,
 })
 
-const sileroForm = ref<Omit<SileroVadConfig, 'algorithm'>>({
+const sileroForm = ref<Omit<SileroVadConfig, 'algorithm' | 'gracePeriodMs'>>({
   model: undefined,
   positiveSpeechThreshold: undefined,
   negativeSpeechThreshold: undefined,
@@ -347,24 +373,27 @@ const sileroForm = ref<Omit<SileroVadConfig, 'algorithm'>>({
   preSpeechPadFrames: undefined,
   minSpeechFrames: undefined,
   submitUserSpeechOnPause: undefined,
-  gracePeriodMs: undefined,
 })
 
-const fireredForm = ref<Omit<FireRedVadConfig, 'algorithm'>>({
-    speechThreshold: undefined,
-    smoothWindowSize: undefined,
-    minSpeechFrame: undefined,
-    maxSpeechFrame: undefined,
-    minSilenceFrame: undefined,
-    padStartFrame: undefined,
-    gracePeriodMs: undefined,
-  })
+const fireredForm = ref<Omit<FireRedVadConfig, 'algorithm' | 'gracePeriodMs'>>({
+  speechThreshold: undefined,
+  smoothWindowSize: undefined,
+  minSpeechFrame: undefined,
+  maxSpeechFrame: undefined,
+  minSilenceFrame: undefined,
+  padStartFrame: undefined,
+})
 
 const smartTurnForm = ref<Omit<SmartTurnConfig, ''>>({
   enabled: false,
   threshold: undefined,
 })
 
+const silenceForm = ref({
+  bargeInSilenceTimeout: undefined as number | undefined,
+  bargeInSilencePlaceholder: undefined as string | undefined,
+})
+
 const hasValue = (v: any) => v !== undefined && v !== ''
 
 function initFromConfig() {
@@ -376,13 +405,14 @@ function initFromConfig() {
 
   form.value.algorithm = props.config.algorithm
 
+  generalForm.value.gracePeriodMs = props.config.gracePeriodMs
+
   if (props.config.algorithm === 'legacy') {
     legacyForm.value = {
       mode: props.config.mode,
       frameDurationMs: props.config.frameDurationMs,
       silencePaddingMs: props.config.silencePaddingMs,
       autoEndSilenceDurationMs: props.config.autoEndSilenceDurationMs,
-      gracePeriodMs: props.config.gracePeriodMs,
     }
   } else if (props.config.algorithm === 'silero') {
     sileroForm.value = {
@@ -394,7 +424,6 @@ function initFromConfig() {
       preSpeechPadFrames: props.config.preSpeechPadFrames,
       minSpeechFrames: props.config.minSpeechFrames,
       submitUserSpeechOnPause: props.config.submitUserSpeechOnPause,
-      gracePeriodMs: props.config.gracePeriodMs,
     }
   } else {
     fireredForm.value = {
@@ -404,7 +433,6 @@ function initFromConfig() {
       maxSpeechFrame: props.config.maxSpeechFrame,
       minSilenceFrame: props.config.minSilenceFrame,
       padStartFrame: props.config.padStartFrame,
-      gracePeriodMs: props.config.gracePeriodMs,
     }
   }
 
@@ -419,15 +447,20 @@ function initFromConfig() {
       threshold: undefined,
     }
   }
+
+  silenceForm.value = {
+    bargeInSilenceTimeout: props.config.bargeInSilenceTimeout,
+    bargeInSilencePlaceholder: props.config.bargeInSilencePlaceholder,
+  }
 }
 
 function resetForms() {
+  generalForm.value.gracePeriodMs = undefined
   legacyForm.value = {
     mode: undefined,
     frameDurationMs: undefined,
     silencePaddingMs: undefined,
     autoEndSilenceDurationMs: undefined,
-    gracePeriodMs: undefined,
   }
   sileroForm.value = {
     model: undefined,
@@ -438,7 +471,6 @@ function resetForms() {
     preSpeechPadFrames: undefined,
     minSpeechFrames: undefined,
     submitUserSpeechOnPause: undefined,
-    gracePeriodMs: undefined,
   }
   fireredForm.value = {
     speechThreshold: undefined,
@@ -447,16 +479,20 @@ function resetForms() {
     maxSpeechFrame: undefined,
     minSilenceFrame: undefined,
     padStartFrame: undefined,
-    gracePeriodMs: undefined,
   }
   smartTurnForm.value = {
     enabled: false,
     threshold: undefined,
   }
+  silenceForm.value = {
+    bargeInSilenceTimeout: undefined,
+    bargeInSilencePlaceholder: undefined,
+  }
 }
 
 function onAlgorithmChange() {
   resetForms()
+  generalForm.value.gracePeriodMs = props.config?.gracePeriodMs
 }
 
 watch(() => props.config, initFromConfig, { immediate: true })
@@ -469,6 +505,8 @@ function handleSubmit() {
       }
     : undefined
 
+  const gracePeriodMs = generalForm.value.gracePeriodMs
+
   if (form.value.algorithm === 'legacy') {
     const config: ServerVadConfig = {
       algorithm: 'legacy',
@@ -476,8 +514,10 @@ function handleSubmit() {
       ...(hasValue(legacyForm.value.frameDurationMs) && { frameDurationMs: legacyForm.value.frameDurationMs }),
       ...(hasValue(legacyForm.value.silencePaddingMs) && { silencePaddingMs: legacyForm.value.silencePaddingMs }),
       ...(hasValue(legacyForm.value.autoEndSilenceDurationMs) && { autoEndSilenceDurationMs: legacyForm.value.autoEndSilenceDurationMs }),
-      ...(hasValue(legacyForm.value.gracePeriodMs) && { gracePeriodMs: legacyForm.value.gracePeriodMs }),
+      ...(hasValue(gracePeriodMs) && { gracePeriodMs }),
       ...(smartTurn && { smartTurn }),
+      ...(hasValue(silenceForm.value.bargeInSilenceTimeout) && { bargeInSilenceTimeout: silenceForm.value.bargeInSilenceTimeout }),
+      ...(silenceForm.value.bargeInSilencePlaceholder !== undefined && { bargeInSilencePlaceholder: silenceForm.value.bargeInSilencePlaceholder }),
     }
     emit('save', config)
   } else if (form.value.algorithm === 'silero') {
@@ -491,8 +531,10 @@ function handleSubmit() {
       ...(hasValue(sileroForm.value.preSpeechPadFrames) && { preSpeechPadFrames: sileroForm.value.preSpeechPadFrames }),
       ...(hasValue(sileroForm.value.minSpeechFrames) && { minSpeechFrames: sileroForm.value.minSpeechFrames }),
       ...(hasValue(sileroForm.value.submitUserSpeechOnPause) && { submitUserSpeechOnPause: sileroForm.value.submitUserSpeechOnPause }),
-      ...(hasValue(sileroForm.value.gracePeriodMs) && { gracePeriodMs: sileroForm.value.gracePeriodMs }),
+      ...(hasValue(gracePeriodMs) && { gracePeriodMs }),
       ...(smartTurn && { smartTurn }),
+      ...(hasValue(silenceForm.value.bargeInSilenceTimeout) && { bargeInSilenceTimeout: silenceForm.value.bargeInSilenceTimeout }),
+      ...(silenceForm.value.bargeInSilencePlaceholder !== undefined && { bargeInSilencePlaceholder: silenceForm.value.bargeInSilencePlaceholder }),
     }
     emit('save', config)
   } else {
@@ -504,8 +546,10 @@ function handleSubmit() {
       ...(hasValue(fireredForm.value.maxSpeechFrame) && { maxSpeechFrame: fireredForm.value.maxSpeechFrame }),
       ...(hasValue(fireredForm.value.minSilenceFrame) && { minSilenceFrame: fireredForm.value.minSilenceFrame }),
       ...(hasValue(fireredForm.value.padStartFrame) && { padStartFrame: fireredForm.value.padStartFrame }),
-      ...(hasValue(fireredForm.value.gracePeriodMs) && { gracePeriodMs: fireredForm.value.gracePeriodMs }),
+      ...(hasValue(gracePeriodMs) && { gracePeriodMs }),
       ...(smartTurn && { smartTurn }),
+      ...(hasValue(silenceForm.value.bargeInSilenceTimeout) && { bargeInSilenceTimeout: silenceForm.value.bargeInSilenceTimeout }),
+      ...(silenceForm.value.bargeInSilencePlaceholder !== undefined && { bargeInSilencePlaceholder: silenceForm.value.bargeInSilencePlaceholder }),
     }
     emit('save', config)
   }