diff --git a/apps/.i18n/native-source.json b/apps/.i18n/native-source.json
index 305a66d65dd..725948394ec 100644
--- a/apps/.i18n/native-source.json
+++ b/apps/.i18n/native-source.json
@@ -8475,7 +8475,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 162,
+      "line": 163,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "The session attaches once the gateway is ready.",
       "surface": "apple",
@@ -8483,7 +8483,7 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 261,
+      "line": 268,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Opens Settings / Gateway",
       "surface": "apple",
@@ -8491,7 +8491,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 286,
+      "line": 293,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "New Chat",
       "surface": "apple",
@@ -8499,7 +8499,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 299,
+      "line": 306,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "New Chat in Worktree",
       "surface": "apple",
@@ -8507,7 +8507,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 314,
+      "line": 321,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Export Transcript",
       "surface": "apple",
@@ -8515,7 +8515,7 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 324,
+      "line": 331,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Chat actions",
       "surface": "apple",
@@ -8523,7 +8523,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 377,
+      "line": 384,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Connected",
       "surface": "apple",
@@ -8531,7 +8531,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 377,
+      "line": 384,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Unavailable",
       "surface": "apple",
@@ -8539,7 +8539,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 389,
+      "line": 396,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Message \\(self.agentDisplayName)...",
       "surface": "apple",
@@ -8547,7 +8547,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 392,
+      "line": 399,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Message \\(self.agentDisplayName); sends when connected",
       "surface": "apple",
@@ -8555,7 +8555,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 394,
+      "line": 401,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Connect to a gateway",
       "surface": "apple",
@@ -8563,7 +8563,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 462,
+      "line": 469,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Check OpenClaw status",
       "surface": "apple",
@@ -8571,7 +8571,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 463,
+      "line": 470,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Summarize the current OpenClaw status and tell me what needs attention.",
       "surface": "apple",
@@ -8579,7 +8579,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 466,
+      "line": 473,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "What can I control here?",
       "surface": "apple",
@@ -8587,7 +8587,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 467,
+      "line": 474,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Show me which phone controls and device capabilities are available right now.",
       "surface": "apple",
@@ -8595,7 +8595,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 470,
+      "line": 477,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Help me start voice chat",
       "surface": "apple",
@@ -8603,7 +8603,7 @@
     },
     {
       "kind": "ui-localized-call",
-      "line": 471,
+      "line": 478,
       "path": "apps/ios/Sources/Design/ChatProTab.swift",
       "source": "Help me start a realtime voice session from this phone.",
       "surface": "apple",
@@ -12731,7 +12731,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 2855,
+      "line": 2878,
       "path": "apps/ios/Sources/Model/NodeAppModel.swift",
       "source": "Action required",
       "surface": "apple",
@@ -12739,7 +12739,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 2855,
+      "line": 2878,
       "path": "apps/ios/Sources/Model/NodeAppModel.swift",
       "source": "Approval needed",
       "surface": "apple",
@@ -12747,7 +12747,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 3417,
+      "line": 3440,
       "path": "apps/ios/Sources/Model/NodeAppModel.swift",
       "source": "Connecting…",
       "surface": "apple",
@@ -12755,7 +12755,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 3417,
+      "line": 3440,
       "path": "apps/ios/Sources/Model/NodeAppModel.swift",
       "source": "Reconnecting…",
       "surface": "apple",
@@ -12763,7 +12763,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 3421,
+      "line": 3444,
       "path": "apps/ios/Sources/Model/NodeAppModel.swift",
       "source": "Connecting...",
       "surface": "apple",
@@ -12771,7 +12771,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 3421,
+      "line": 3444,
       "path": "apps/ios/Sources/Model/NodeAppModel.swift",
       "source": "Reconnecting...",
       "surface": "apple",
@@ -12779,7 +12779,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 3713,
+      "line": 3736,
       "path": "apps/ios/Sources/Model/NodeAppModel.swift",
       "source": "Connected",
       "surface": "apple",
@@ -12787,7 +12787,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 3713,
+      "line": 3736,
       "path": "apps/ios/Sources/Model/NodeAppModel.swift",
       "source": "Offline",
       "surface": "apple",
@@ -14307,7 +14307,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 260,
+      "line": 280,
       "path": "apps/ios/Sources/Voice/VoiceWakeManager.swift",
       "source": "Off",
       "surface": "apple",
@@ -14315,7 +14315,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 260,
+      "line": 280,
       "path": "apps/ios/Sources/Voice/VoiceWakeManager.swift",
       "source": "Paused",
       "surface": "apple",
@@ -21811,7 +21811,7 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 214,
+      "line": 240,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Context \\(percentage)% used",
       "surface": "apple",
@@ -21819,7 +21819,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 218,
+      "line": 244,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Thinking",
       "surface": "apple",
@@ -21827,7 +21827,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 239,
+      "line": 265,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Model",
       "surface": "apple",
@@ -21835,7 +21835,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 256,
+      "line": 282,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Pinned",
       "surface": "apple",
@@ -21843,7 +21843,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 264,
+      "line": 290,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Recent",
       "surface": "apple",
@@ -21851,7 +21851,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 272,
+      "line": 298,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Models",
       "surface": "apple",
@@ -21859,7 +21859,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 301,
+      "line": 327,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Pin model",
       "surface": "apple",
@@ -21867,7 +21867,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 301,
+      "line": 327,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Unpin model",
       "surface": "apple",
@@ -21875,7 +21875,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 305,
+      "line": 331,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Session",
       "surface": "apple",
@@ -21883,7 +21883,7 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 369,
+      "line": 395,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Add Image",
       "surface": "apple",
@@ -21891,15 +21891,23 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 370,
+      "line": 396,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Attachments",
       "surface": "apple",
       "id": "native.apple.1963dfa3285cb715"
     },
+    {
+      "kind": "ui-call",
+      "line": 431,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
+      "source": "Voice note",
+      "surface": "apple",
+      "id": "native.apple.f8dda832ed76441d"
+    },
     {
       "kind": "conditional-branch",
-      "line": 540,
+      "line": 589,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Talk",
       "surface": "apple",
@@ -21907,7 +21915,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 586,
+      "line": 635,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Start realtime chat",
       "surface": "apple",
@@ -21915,7 +21923,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 586,
+      "line": 635,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Stop realtime chat",
       "surface": "apple",
@@ -21923,7 +21931,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 619,
+      "line": 668,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Start",
       "surface": "apple",
@@ -21931,7 +21939,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 619,
+      "line": 668,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Stop",
       "surface": "apple",
@@ -21939,7 +21947,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 729,
+      "line": 778,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Loading commands",
       "surface": "apple",
@@ -21947,7 +21955,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 738,
+      "line": 787,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Commands unavailable",
       "surface": "apple",
@@ -21955,7 +21963,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 747,
+      "line": 796,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Retry",
       "surface": "apple",
@@ -21963,7 +21971,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 756,
+      "line": 805,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "No matching commands",
       "surface": "apple",
@@ -21971,7 +21979,7 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 888,
+      "line": 937,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Stop response",
       "surface": "apple",
@@ -21979,7 +21987,7 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 913,
+      "line": 962,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Send message",
       "surface": "apple",
@@ -21987,7 +21995,7 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 928,
+      "line": 977,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Refresh",
       "surface": "apple",
@@ -21995,7 +22003,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 1038,
+      "line": 1108,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Connecting...",
       "surface": "apple",
@@ -22003,7 +22011,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 1038,
+      "line": 1108,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift",
       "source": "Gateway connected",
       "surface": "apple",
@@ -22019,7 +22027,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 557,
+      "line": 569,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Show full output",
       "surface": "apple",
@@ -22027,7 +22035,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 557,
+      "line": 569,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Show less",
       "surface": "apple",
@@ -22035,7 +22043,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 616,
+      "line": 628,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Writing",
       "surface": "apple",
@@ -22043,7 +22051,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 642,
+      "line": 654,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Preparing audio…",
       "surface": "apple",
@@ -22051,7 +22059,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 645,
+      "line": 657,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Speaking…",
       "surface": "apple",
@@ -22059,7 +22067,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 653,
+      "line": 665,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Preparing audio, tap to cancel",
       "surface": "apple",
@@ -22067,7 +22075,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 654,
+      "line": 666,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Speaking, tap to stop",
       "surface": "apple",
@@ -22075,7 +22083,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 677,
+      "line": 689,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Queued",
       "surface": "apple",
@@ -22083,7 +22091,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 679,
+      "line": 691,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Sending…",
       "surface": "apple",
@@ -22091,7 +22099,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 681,
+      "line": 693,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Confirming…",
       "surface": "apple",
@@ -22099,7 +22107,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 683,
+      "line": 695,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Delivery unknown",
       "surface": "apple",
@@ -22107,7 +22115,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 685,
+      "line": 697,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Not sent",
       "surface": "apple",
@@ -22115,7 +22123,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 707,
+      "line": 719,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Queued, sends when reconnected",
       "surface": "apple",
@@ -22123,7 +22131,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 709,
+      "line": 721,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Sending",
       "surface": "apple",
@@ -22131,7 +22139,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 711,
+      "line": 723,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Sent, waiting for chat history confirmation",
       "surface": "apple",
@@ -22139,7 +22147,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 713,
+      "line": 725,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Delivery unconfirmed, touch and hold to retry or delete",
       "surface": "apple",
@@ -22147,7 +22155,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 715,
+      "line": 727,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Not sent, touch and hold to retry or delete",
       "surface": "apple",
@@ -22155,7 +22163,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 810,
+      "line": 822,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "Running tools…",
       "surface": "apple",
@@ -22163,7 +22171,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 818,
+      "line": 830,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift",
       "source": "\\(display.emoji) \\(display.label)",
       "surface": "apple",
@@ -22179,7 +22187,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 430,
+      "line": 434,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Retry Send",
       "surface": "apple",
@@ -22187,7 +22195,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 444,
+      "line": 448,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Delete",
       "surface": "apple",
@@ -22195,7 +22203,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 475,
+      "line": 479,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Stop Listening",
       "surface": "apple",
@@ -22203,7 +22211,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 478,
+      "line": 482,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Listen",
       "surface": "apple",
@@ -22211,7 +22219,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 551,
+      "line": 555,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Jump to latest",
       "surface": "apple",
@@ -22219,7 +22227,7 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 562,
+      "line": 566,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Jump to latest reply",
       "surface": "apple",
@@ -22227,7 +22235,7 @@
     },
     {
       "kind": "ui-named-argument",
-      "line": 582,
+      "line": 586,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Refresh",
       "surface": "apple",
@@ -22235,7 +22243,7 @@
     },
     {
       "kind": "ui-call",
-      "line": 1007,
+      "line": 1011,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Loading chat",
       "surface": "apple",
@@ -22243,15 +22251,31 @@
     },
     {
       "kind": "ui-modifier",
-      "line": 1087,
+      "line": 1091,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift",
       "source": "Dismiss",
       "surface": "apple",
       "id": "native.apple.5c481fea4eb3dc17"
     },
+    {
+      "kind": "ui-localized-call",
+      "line": 22,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+Attachments.swift",
+      "source": "Could not attach voice note: \\(error.localizedDescription)",
+      "surface": "apple",
+      "id": "native.apple.82befa424fe7bace"
+    },
+    {
+      "kind": "ui-localized-call",
+      "line": 27,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+Attachments.swift",
+      "source": "Voice note exceeds the 5 MB attachment limit",
+      "surface": "apple",
+      "id": "native.apple.ec5daba7e38184dc"
+    },
     {
       "kind": "conditional-branch",
-      "line": 62,
+      "line": 91,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+Attachments.swift",
       "source": "\\(baseName).jpg",
       "surface": "apple",
@@ -22259,7 +22283,7 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 1360,
+      "line": 1361,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel.swift",
       "source": "delivery unconfirmed",
       "surface": "apple",
@@ -22267,12 +22291,68 @@
     },
     {
       "kind": "conditional-branch",
-      "line": 1360,
+      "line": 1361,
       "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel.swift",
       "source": "queued after route change",
       "surface": "apple",
       "id": "native.apple.722f1f90b97e8e45"
     },
+    {
+      "kind": "ui-modifier",
+      "line": 39,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteComposerViews.swift",
+      "source": "Record Voice Note",
+      "surface": "apple",
+      "id": "native.apple.71e0219d1501de06"
+    },
+    {
+      "kind": "ui-modifier",
+      "line": 40,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteComposerViews.swift",
+      "source": "Record voice note",
+      "surface": "apple",
+      "id": "native.apple.e6770d7cc616021f"
+    },
+    {
+      "kind": "ui-call",
+      "line": 59,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteComposerViews.swift",
+      "source": "Recording",
+      "surface": "apple",
+      "id": "native.apple.c971dc7dc243d0d1"
+    },
+    {
+      "kind": "ui-modifier",
+      "line": 75,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteComposerViews.swift",
+      "source": "Cancel voice note",
+      "surface": "apple",
+      "id": "native.apple.0fb64b1b7ea0aa5e"
+    },
+    {
+      "kind": "ui-modifier",
+      "line": 84,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteComposerViews.swift",
+      "source": "Finish voice note",
+      "surface": "apple",
+      "id": "native.apple.b7a9de9b3c8e4bad"
+    },
+    {
+      "kind": "ui-localized-call",
+      "line": 108,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteRecorder.swift",
+      "source": "Microphone access is required. Enable it in Settings.",
+      "surface": "apple",
+      "id": "native.apple.1fc7530ccdfc937e"
+    },
+    {
+      "kind": "ui-localized-call",
+      "line": 121,
+      "path": "apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteRecorder.swift",
+      "source": "Could not start recording: \\(error.localizedDescription)",
+      "surface": "apple",
+      "id": "native.apple.c963f6dfaa0610b5"
+    },
     {
       "kind": "conditional-branch",
       "line": 235,
diff --git a/apps/ios/Sources/Design/ChatProTab.swift b/apps/ios/Sources/Design/ChatProTab.swift
index b08f7a64230..354a074fc65 100644
--- a/apps/ios/Sources/Design/ChatProTab.swift
+++ b/apps/ios/Sources/Design/ChatProTab.swift
@@ -150,6 +150,7 @@ struct ChatProTab: View {
                 emptyAssistantIntro: String(localized: "What would you like to work on?"),
                 emptyAssistantPrompts: Self.emptyAssistantPrompts,
                 talkControl: self.talkControl,
+                voiceNoteControl: self.voiceNoteControl,
                 speech: self.speech)
                 // iMessage-style grey bubbles for agent replies in the clean chrome.
                     .environment(\.openClawAssistantBubblesInCleanChrome, true)
@@ -245,6 +246,12 @@ struct ChatProTab: View {
             })
     }
 
+    private var voiceNoteControl: OpenClawChatVoiceNoteControl {
+        OpenClawChatVoiceNoteControl(
+            recorder: self.appModel.voiceNoteRecorder,
+            isTalkActive: self.appModel.talkMode.isEnabled || self.appModel.talkMode.isPushToTalkActive)
+    }
+
     private var activeAgentID: String {
         self.normalized(self.appModel.chatAgentId)
             ?? "main"
diff --git a/apps/ios/Sources/Model/NodeAppModel.swift b/apps/ios/Sources/Model/NodeAppModel.swift
index b571ad40721..e281396b348 100644
--- a/apps/ios/Sources/Model/NodeAppModel.swift
+++ b/apps/ios/Sources/Model/NodeAppModel.swift
@@ -255,6 +255,7 @@ final class NodeAppModel {
     private var gatewayHealthMonitorDisabled = false
     private let notificationCenter: NotificationCentering
     let voiceWake = VoiceWakeManager()
+    let voiceNoteRecorder: OpenClawVoiceNoteRecorder
     let talkMode: TalkModeManager
     private let locationService: any LocationServicing
     private let deviceStatusService: any DeviceStatusServicing
@@ -500,7 +501,8 @@ final class NodeAppModel {
         remindersService: any RemindersServicing = RemindersService(),
         motionService: any MotionServicing = MotionService(),
         watchMessagingService: any WatchMessagingServicing = WatchMessagingService(),
-        talkMode: TalkModeManager = TalkModeManager())
+        talkMode: TalkModeManager = TalkModeManager(),
+        voiceNoteRecorder: OpenClawVoiceNoteRecorder = OpenClawVoiceNoteRecorder())
     {
         self.screen = screen
         self.camera = camera
@@ -515,6 +517,7 @@ final class NodeAppModel {
         self.motionService = motionService
         self.watchMessagingService = watchMessagingService
         self.talkMode = talkMode
+        self.voiceNoteRecorder = voiceNoteRecorder
         self.apnsDeviceTokenHex = UserDefaults.standard.string(forKey: Self.apnsDeviceTokenUserDefaultsKey)
         restorePersistedWatchExecApprovalBridgeState()
         GatewayDiagnostics.bootstrap()
@@ -575,6 +578,9 @@ final class NodeAppModel {
                 // Best-effort only.
             }
         }
+        self.voiceNoteRecorder.onRecordingActiveChanged = { [weak self] isActive in
+            self?.voiceWake.setSuppressedByVoiceNote(isActive)
+        }
 
         let enabled = UserDefaults.standard.bool(forKey: "voiceWake.enabled")
         self.voiceWake.setEnabled(enabled)
@@ -881,6 +887,9 @@ final class NodeAppModel {
         }
         UserDefaults.standard.set(enabled, forKey: "talk.enabled")
         if enabled {
+            if self.voiceNoteRecorder.isRecording || self.voiceNoteRecorder.isRequestingPermission {
+                self.voiceNoteRecorder.cancel()
+            }
             // Voice wake holds the microphone continuously; talk mode needs exclusive access for STT.
             // When talk is enabled from the UI, prioritize talk and pause voice wake.
             self.voiceWake.setSuppressedByTalk(true)
@@ -1973,6 +1982,20 @@ final class NodeAppModel {
     }
 
     private func handleTalkInvoke(_ req: BridgeInvokeRequest) async throws -> BridgeInvokeResponse {
+        if req.command == OpenClawTalkCommand.pttOnce.rawValue ||
+            req.command == OpenClawTalkCommand.pttStart.rawValue
+        {
+            // Composer recording and PTT use separate AVAudio stacks. Refuse the
+            // second owner before it can reconfigure the shared audio session.
+            guard !self.voiceNoteRecorder.isRecording,
+                  !self.voiceNoteRecorder.isRequestingPermission
+            else {
+                throw NSError(domain: "TalkMode", code: 8, userInfo: [
+                    NSLocalizedDescriptionKey: "Voice note recording is active",
+                ])
+            }
+        }
+
         if req.command == OpenClawTalkCommand.pttOnce.rawValue {
             self.acquirePttVoiceWakeLease()
             defer { self.releasePttVoiceWakeLease() }
diff --git a/apps/ios/Sources/Voice/VoiceWakeManager.swift b/apps/ios/Sources/Voice/VoiceWakeManager.swift
index c2adb9fe859..8a069021f47 100644
--- a/apps/ios/Sources/Voice/VoiceWakeManager.swift
+++ b/apps/ios/Sources/Voice/VoiceWakeManager.swift
@@ -49,6 +49,11 @@ private enum VoiceWakeAudioError: LocalizedError {
     }
 }
 
+private enum VoiceWakeSuppressionReason: Hashable {
+    case talk
+    case voiceNote
+}
+
 extension AVAudioPCMBuffer {
     fileprivate func deepCopy() -> AVAudioPCMBuffer? {
         let format = self.format
@@ -111,7 +116,7 @@ final class VoiceWakeManager: NSObject {
     private var lastDispatched: String?
     private var onCommand: (@Sendable (String) async -> Void)?
     private var userDefaultsObserver: NSObjectProtocol?
-    private var suppressedByTalk: Bool = false
+    private var suppressionReasons: Set<VoiceWakeSuppressionReason> = []
 
     private let externalAudioResumeDelayNs: UInt64
     private let recognitionErrorRestartDelayNs: UInt64
@@ -167,8 +172,23 @@ final class VoiceWakeManager: NSObject {
     }
 
     func setSuppressedByTalk(_ suppressed: Bool) {
-        self.suppressedByTalk = suppressed
+        self.setSuppressed(suppressed, reason: .talk)
+    }
+
+    func setSuppressedByVoiceNote(_ suppressed: Bool) {
+        self.setSuppressed(suppressed, reason: .voiceNote)
+    }
+
+    private func setSuppressed(_ suppressed: Bool, reason: VoiceWakeSuppressionReason) {
         if suppressed {
+            self.suppressionReasons.insert(reason)
+        } else {
+            self.suppressionReasons.remove(reason)
+        }
+
+        // Each microphone owner clears only its own reason, so Talk ending
+        // cannot restart Voice Wake over an active voice-note recording.
+        if !self.suppressionReasons.isEmpty {
             self.cancelScheduledStart()
             if self.isListening {
                 self.isListening = false
@@ -214,7 +234,7 @@ final class VoiceWakeManager: NSObject {
         self.isStarting = true
         defer { self.isStarting = false }
 
-        guard !self.suppressedByTalk else {
+        guard self.suppressionReasons.isEmpty else {
             self.isListening = false
             self.statusText = "Paused"
             return
@@ -255,7 +275,7 @@ final class VoiceWakeManager: NSObject {
             return
         }
 
-        guard self.isEnabled, !self.suppressedByTalk, !self.isSuspendedForExternalAudio else {
+        guard self.isEnabled, self.suppressionReasons.isEmpty, !self.isSuspendedForExternalAudio else {
             self.isListening = false
             self.statusText = self.isEnabled ? "Paused" : "Off"
             return
@@ -303,7 +323,7 @@ final class VoiceWakeManager: NSObject {
     }
 
     private func startRecognition() throws {
-        guard self.isEnabled, !self.suppressedByTalk, !self.isSuspendedForExternalAudio else { return }
+        guard self.isEnabled, self.suppressionReasons.isEmpty, !self.isSuspendedForExternalAudio else { return }
 
         self.recognitionTask?.cancel()
         self.recognitionTask = nil
diff --git a/apps/ios/Tests/NodeAppModelInvokeTests.swift b/apps/ios/Tests/NodeAppModelInvokeTests.swift
index cde3afefc90..524217cca4f 100644
--- a/apps/ios/Tests/NodeAppModelInvokeTests.swift
+++ b/apps/ios/Tests/NodeAppModelInvokeTests.swift
@@ -79,6 +79,21 @@ private func makeProjectedWatchChatRawMessage(
     return try JSONDecoder().decode(AnyCodable.self, from: data)
 }
 
+@MainActor
+private final class VoiceNoteCaptureStub: VoiceNoteAudioCapture {
+    func requestPermission() async -> Bool {
+        true
+    }
+
+    func start(url _: URL) throws {}
+
+    func stop() -> TimeInterval {
+        1
+    }
+
+    func cancel() {}
+}
+
 @MainActor
 private func waitForMainActorWork(_ condition: () -> Bool) async {
     for _ in 0..<100 {
@@ -724,6 +739,23 @@ private actor WatchSnapshotSendGate {
         appModel.voiceWake.stop()
     }
 
+    @Test @MainActor func `PTT start is refused while a voice note owns the microphone`() async {
+        let voiceNoteRecorder = OpenClawVoiceNoteRecorder(capture: VoiceNoteCaptureStub())
+        let talkMode = TalkModeManager(allowSimulatorCapture: true)
+        let appModel = NodeAppModel(talkMode: talkMode, voiceNoteRecorder: voiceNoteRecorder)
+        #expect(await voiceNoteRecorder.start())
+        defer { voiceNoteRecorder.cancel() }
+
+        let request = BridgeInvokeRequest(
+            id: "ptt-start-during-voice-note",
+            command: OpenClawTalkCommand.pttStart.rawValue)
+        let response = await appModel._test_handleInvoke(request)
+
+        #expect(response.ok == false)
+        #expect(response.error?.message.contains("Voice note recording is active") == true)
+        #expect(talkMode.isPushToTalkActive == false)
+    }
+
     @Test @MainActor func `overlapping PTT owners keep voice wake suspended until final release`() {
         let appModel = NodeAppModel(talkMode: TalkModeManager(allowSimulatorCapture: true))
         appModel.voiceWake.isEnabled = true
diff --git a/apps/ios/Tests/VoiceWakeManagerSuppressionTests.swift b/apps/ios/Tests/VoiceWakeManagerSuppressionTests.swift
index 4aece2c82d8..d568c861607 100644
--- a/apps/ios/Tests/VoiceWakeManagerSuppressionTests.swift
+++ b/apps/ios/Tests/VoiceWakeManagerSuppressionTests.swift
@@ -69,4 +69,23 @@ struct VoiceWakeManagerSuppressionTests {
         #expect(manager.statusText == "Voice Wake isn’t supported on Simulator")
         #expect(manager.isListening == false)
     }
+
+    @Test
+    @MainActor func `clearing Talk suppression does not clobber voice note suppression`() async {
+        let manager = VoiceWakeManager._test_withoutRestartDelays()
+        manager.isEnabled = true
+        manager.statusText = "Listening"
+
+        manager.setSuppressedByVoiceNote(true)
+        manager.setSuppressedByTalk(true)
+        manager.setSuppressedByTalk(false)
+
+        await manager._test_waitForScheduledStart()
+        #expect(manager.statusText == "Paused")
+        #expect(manager.isListening == false)
+
+        manager.setSuppressedByVoiceNote(false)
+        await manager._test_waitForScheduledStart()
+        #expect(manager.statusText == "Voice Wake isn’t supported on Simulator")
+    }
 }
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift
index c5fb8928648..e8313e97585 100644
--- a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatComposer.swift
@@ -87,6 +87,8 @@ struct OpenClawChatComposer: View {
     let isAttachmentInputEnabled: Bool
     let messagePlaceholder: String?
     let talkControl: OpenClawChatTalkControl?
+    let voiceNoteControl: OpenClawChatVoiceNoteControl?
+    @State private var stagingVoiceNoteURL: URL?
 
     #if !os(macOS)
     @State private var pickerItems: [PhotosPickerItem] = []
@@ -101,7 +103,7 @@ struct OpenClawChatComposer: View {
 
     var body: some View {
         VStack(alignment: .leading, spacing: 4) {
-            if self.showsToolbar {
+            if self.showsToolbar, self.voiceNoteControl?.recorder.isRecording != true {
                 self.composerToolbar
             }
 
@@ -109,7 +111,12 @@ struct OpenClawChatComposer: View {
                 self.attachmentsStrip
             }
 
-            self.editor
+            if let voiceNoteControl, voiceNoteControl.recorder.isRecording {
+                OpenClawVoiceNoteRecordingRow(recorder: voiceNoteControl.recorder)
+                    .padding(self.editorPadding)
+            } else {
+                self.editor
+            }
         }
         .padding(self.composerPadding)
         .background {
@@ -163,6 +170,18 @@ struct OpenClawChatComposer: View {
                 self.viewModel.loadSlashCommandsIfNeeded()
             }
         #endif
+            .onChange(of: self.voiceNoteControl?.recorder.completedRecording) { _, recording in
+                    guard recording != nil else { return }
+                    self.stageCompletedVoiceNoteIfNeeded()
+                }
+                .onChange(of: self.voiceNoteControl?.recorder.errorMessage) { _, message in
+                    if let message {
+                        self.viewModel.errorText = message
+                    }
+                }
+                .onAppear {
+                    self.stageCompletedVoiceNoteIfNeeded()
+                }
     }
 
     private var composerToolbar: some View {
@@ -193,6 +212,13 @@ struct OpenClawChatComposer: View {
             if self.style == .standard {
                 self.refreshButton
                 self.attachmentPicker
+                if let voiceNoteControl, !voiceNoteControl.isTalkActive {
+                    OpenClawVoiceNoteButton(
+                        control: voiceNoteControl,
+                        compact: false,
+                        isComposerEnabled: self.isComposerEnabled,
+                        isAttachmentInputEnabled: self.isAttachmentInputEnabled)
+                }
             }
         }
         .padding(.horizontal, 10)
@@ -400,13 +426,27 @@ struct OpenClawChatComposer: View {
                                 .scaledToFill()
                                 .frame(width: 22, height: 22)
                                 .clipShape(RoundedRectangle(cornerRadius: 6, style: .continuous))
+                        } else if att.mimeType.hasPrefix("audio/") {
+                            Image(systemName: "waveform")
+                            Text("Voice note")
+                                .font(OpenClawChatTypography.caption)
+                            if let durationSeconds = att.durationSeconds {
+                                Text(openClawVoiceNoteDurationLabel(durationSeconds))
+                                    .font(OpenClawChatTypography.caption)
+                                    .foregroundStyle(.secondary)
+                            }
                         } else {
                             Image(systemName: "photo")
+                            Text(att.fileName)
+                                .font(OpenClawChatTypography.caption)
+                                .lineLimit(1)
                         }
 
-                        Text(att.fileName)
-                            .font(OpenClawChatTypography.caption)
-                            .lineLimit(1)
+                        if att.preview != nil {
+                            Text(att.fileName)
+                                .font(OpenClawChatTypography.caption)
+                                .lineLimit(1)
+                        }
 
                         Button {
                             self.viewModel.removeAttachment(att.id)
@@ -497,6 +537,15 @@ struct OpenClawChatComposer: View {
                 self.attachmentPicker
                     .frame(width: self.cleanIconControlSize, height: self.cleanEditorMinHeight)
 
+                if let voiceNoteControl, !voiceNoteControl.isTalkActive {
+                    OpenClawVoiceNoteButton(
+                        control: voiceNoteControl,
+                        compact: true,
+                        isComposerEnabled: self.isComposerEnabled,
+                        isAttachmentInputEnabled: self.isAttachmentInputEnabled)
+                        .frame(width: self.cleanIconControlSize, height: self.cleanEditorMinHeight)
+                }
+
                 self.editorOverlay
                     .padding(.vertical, self.cleanEditorTextPadding)
                     .padding(.horizontal, 14)
@@ -1030,10 +1079,31 @@ struct OpenClawChatComposer: View {
 
     private var canSendMessage: Bool {
         self.isComposerEnabled
+            && self.voiceNoteControl?.recorder.completedRecording == nil
             && self.viewModel.canSend
             && (self.isAttachmentInputEnabled || self.viewModel.attachments.isEmpty)
     }
 
+    private func stageCompletedVoiceNoteIfNeeded() {
+        guard let recorder = self.voiceNoteControl?.recorder,
+              let recording = recorder.completedRecording,
+              self.stagingVoiceNoteURL != recording.fileURL
+        else { return }
+
+        self.stagingVoiceNoteURL = recording.fileURL
+        Task {
+            await self.viewModel.addVoiceNoteAttachment(
+                fileURL: recording.fileURL,
+                durationSeconds: recording.durationSeconds)
+            if recorder.completedRecording == recording {
+                recorder.clearCompletedRecording()
+            }
+            if self.stagingVoiceNoteURL == recording.fileURL {
+                self.stagingVoiceNoteURL = nil
+            }
+        }
+    }
+
     private var connectionStatusText: String {
         self.connectionOK ? "Gateway connected" : "Connecting..."
     }
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift
index 9215c41bd66..2355c0c1379 100644
--- a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatMessageViews.swift
@@ -478,17 +478,29 @@ private struct AttachmentRow: View {
 
     var body: some View {
         HStack(spacing: 8) {
-            Image(systemName: "paperclip")
-            Text(self.att.fileName ?? "Attachment")
+            Image(systemName: self.isAudio ? "waveform" : "paperclip")
+            Text(self.isAudio ? "Voice note" : (self.att.fileName ?? "Attachment"))
                 .font(OpenClawChatTypography.footnote)
                 .lineLimit(1)
                 .foregroundStyle(self.isUser ? OpenClawChatTheme.userText : OpenClawChatTheme.assistantText)
+            if self.isAudio, let durationSeconds = self.att.durationSeconds {
+                Text(openClawVoiceNoteDurationLabel(durationSeconds))
+                    .font(OpenClawChatTypography.footnote)
+                    .foregroundStyle(
+                        self.isUser
+                            ? OpenClawChatTheme.userText.opacity(0.72)
+                            : OpenClawChatTheme.assistantText.opacity(0.72))
+            }
             Spacer()
         }
         .padding(10)
         .background(self.isUser ? Color.white.opacity(0.2) : Color.black.opacity(0.04))
         .clipShape(RoundedRectangle(cornerRadius: 12, style: .continuous))
     }
+
+    private var isAudio: Bool {
+        self.att.mimeType?.hasPrefix("audio/") == true
+    }
 }
 
 private struct ToolCallCard: View {
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatModels.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatModels.swift
index 8b29a7b8ee7..09103abe8e4 100644
--- a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatModels.swift
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatModels.swift
@@ -116,6 +116,7 @@ public struct OpenClawChatMessageContent: Codable, Hashable, Sendable {
     public let thinkingSignature: String?
     public let mimeType: String?
     public let fileName: String?
+    public let durationSeconds: Double?
     public let content: AnyCodable?
 
     // Tool-call fields (when `type == "toolCall"` or similar)
@@ -130,6 +131,7 @@ public struct OpenClawChatMessageContent: Codable, Hashable, Sendable {
         thinkingSignature: String? = nil,
         mimeType: String?,
         fileName: String?,
+        durationSeconds: Double? = nil,
         content: AnyCodable?,
         id: String? = nil,
         name: String? = nil,
@@ -141,6 +143,7 @@ public struct OpenClawChatMessageContent: Codable, Hashable, Sendable {
         self.thinkingSignature = thinkingSignature
         self.mimeType = mimeType
         self.fileName = fileName
+        self.durationSeconds = durationSeconds
         self.content = content
         self.id = id
         self.name = name
@@ -154,6 +157,7 @@ public struct OpenClawChatMessageContent: Codable, Hashable, Sendable {
         case thinkingSignature
         case mimeType
         case fileName
+        case durationSeconds
         case content
         case id
         case name
@@ -168,6 +172,7 @@ public struct OpenClawChatMessageContent: Codable, Hashable, Sendable {
         self.thinkingSignature = try container.decodeIfPresent(String.self, forKey: .thinkingSignature)
         self.mimeType = try container.decodeIfPresent(String.self, forKey: .mimeType)
         self.fileName = try container.decodeIfPresent(String.self, forKey: .fileName)
+        self.durationSeconds = try container.decodeIfPresent(Double.self, forKey: .durationSeconds)
         self.id = try container.decodeIfPresent(String.self, forKey: .id)
         self.name = try container.decodeIfPresent(String.self, forKey: .name)
         self.arguments = try container.decodeIfPresent(AnyCodable.self, forKey: .arguments)
@@ -496,6 +501,7 @@ public struct OpenClawPendingAttachment: Identifiable {
     public let mimeType: String
     public let type: String
     public let preview: OpenClawPlatformImage?
+    public let durationSeconds: Double?
 
     public init(
         url: URL?,
@@ -503,7 +509,8 @@ public struct OpenClawPendingAttachment: Identifiable {
         fileName: String,
         mimeType: String,
         type: String = "file",
-        preview: OpenClawPlatformImage?)
+        preview: OpenClawPlatformImage?,
+        durationSeconds: Double? = nil)
     {
         self.url = url
         self.data = data
@@ -511,6 +518,7 @@ public struct OpenClawPendingAttachment: Identifiable {
         self.mimeType = mimeType
         self.type = type
         self.preview = preview
+        self.durationSeconds = durationSeconds
     }
 }
 
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatTranscriptCache.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatTranscriptCache.swift
index 81a7b3bf662..faae4b2e2fe 100644
--- a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatTranscriptCache.swift
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatTranscriptCache.swift
@@ -1085,6 +1085,7 @@ public actor OpenClawChatSQLiteTranscriptCache: OpenClawChatTranscriptCache,
                         thinkingSignature: nil,
                         mimeType: item.mimeType,
                         fileName: item.fileName,
+                        durationSeconds: item.durationSeconds,
                         content: nil,
                         id: item.id,
                         name: item.name,
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift
index 2bfa87127b9..4e42fc5cdc4 100644
--- a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatView.swift
@@ -98,6 +98,7 @@ public struct OpenClawChatView: View {
     private let emptyAssistantIntro: String?
     private let emptyAssistantPrompts: [StarterPrompt]
     private let talkControl: OpenClawChatTalkControl?
+    private let voiceNoteControl: OpenClawChatVoiceNoteControl?
     private let speech: OpenClawChatSpeechController?
 
     private enum ScrollFollowTarget: Equatable {
@@ -150,6 +151,7 @@ public struct OpenClawChatView: View {
         emptyAssistantIntro: String? = nil,
         emptyAssistantPrompts: [StarterPrompt] = [],
         talkControl: OpenClawChatTalkControl? = nil,
+        voiceNoteControl: OpenClawChatVoiceNoteControl? = nil,
         speech: OpenClawChatSpeechController? = nil)
     {
         _viewModel = State(initialValue: viewModel)
@@ -170,6 +172,7 @@ public struct OpenClawChatView: View {
         self.emptyAssistantIntro = emptyAssistantIntro
         self.emptyAssistantPrompts = emptyAssistantPrompts
         self.talkControl = talkControl
+        self.voiceNoteControl = voiceNoteControl
         self.speech = speech
     }
 
@@ -231,7 +234,8 @@ public struct OpenClawChatView: View {
             isComposerEnabled: self.isComposerEnabled,
             isAttachmentInputEnabled: self.isAttachmentInputEnabled,
             messagePlaceholder: self.messagePlaceholder,
-            talkControl: self.talkControl)
+            talkControl: self.talkControl,
+            voiceNoteControl: self.voiceNoteControl)
     }
 
     private var messageList: some View {
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+Attachments.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+Attachments.swift
index d0a6793a263..1a95ce8ad67 100644
--- a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+Attachments.swift
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+Attachments.swift
@@ -9,6 +9,35 @@ import UIKit
 #endif
 
 extension OpenClawChatViewModel {
+    /// Stages a recorded m4a voice note and removes its temporary file.
+    func addVoiceNoteAttachment(fileURL: URL, durationSeconds: Double) async {
+        defer { try? FileManager.default.removeItem(at: fileURL) }
+
+        let data: Data
+        do {
+            data = try await Task.detached(priority: .userInitiated) {
+                try Data(contentsOf: fileURL)
+            }.value
+        } catch {
+            self.errorText = String(localized: "Could not attach voice note: \(error.localizedDescription)")
+            return
+        }
+
+        guard data.count <= Self.maxAttachmentBytes else {
+            self.errorText = String(localized: "Voice note exceeds the 5 MB attachment limit")
+            return
+        }
+
+        self.attachments.append(
+            OpenClawPendingAttachment(
+                url: nil,
+                data: data,
+                fileName: fileURL.lastPathComponent,
+                mimeType: "audio/mp4",
+                preview: nil,
+                durationSeconds: max(0, durationSeconds)))
+    }
+
     func loadAttachments(urls: [URL]) async {
         for url in urls {
             do {
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+HistoryReconciliation.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+HistoryReconciliation.swift
index 14a88e56aed..ce28a6c3a98 100644
--- a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+HistoryReconciliation.swift
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel+HistoryReconciliation.swift
@@ -25,6 +25,7 @@ extension OpenClawChatViewModel {
                 thinkingSignature: content.thinkingSignature,
                 mimeType: content.mimeType,
                 fileName: content.fileName,
+                durationSeconds: content.durationSeconds,
                 content: content.content,
                 id: content.id,
                 name: content.name,
@@ -133,7 +134,9 @@ extension OpenClawChatViewModel {
         OpenClawChatMessage(
             id: existing.id,
             role: incoming.role,
-            content: incoming.content,
+            content: self.preservingLocalAudioDurations(
+                in: incoming.content,
+                from: existing.content),
             timestamp: incoming.timestamp ?? existing.timestamp,
             idempotencyKey: incoming.idempotencyKey,
             toolCallId: incoming.toolCallId,
@@ -143,6 +146,38 @@ extension OpenClawChatViewModel {
             errorMessage: incoming.errorMessage)
     }
 
+    private static func preservingLocalAudioDurations(
+        in incoming: [OpenClawChatMessageContent],
+        from existing: [OpenClawChatMessageContent]) -> [OpenClawChatMessageContent]
+    {
+        let localDurations = existing
+            .filter { $0.mimeType?.hasPrefix("audio/") == true }
+            .map(\.durationSeconds)
+        var audioIndex = 0
+        return incoming.map { content in
+            guard content.mimeType?.hasPrefix("audio/") == true else { return content }
+            defer { audioIndex += 1 }
+            guard content.durationSeconds == nil,
+                  localDurations.indices.contains(audioIndex),
+                  let localDuration = localDurations[audioIndex]
+            else {
+                return content
+            }
+            return OpenClawChatMessageContent(
+                type: content.type,
+                text: content.text,
+                thinking: content.thinking,
+                thinkingSignature: content.thinkingSignature,
+                mimeType: content.mimeType,
+                fileName: content.fileName,
+                durationSeconds: localDuration,
+                content: content.content,
+                id: content.id,
+                name: content.name,
+                arguments: content.arguments)
+        }
+    }
+
     func currentRunMessageScope() -> RunMessageScope {
         RunMessageScope(
             session: self.currentSessionSnapshot(),
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel.swift
index 0f03f229ec7..c419952d1f9 100644
--- a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel.swift
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/ChatViewModel.swift
@@ -1232,7 +1232,7 @@ public final class OpenClawChatViewModel {
                 fileName: att.fileName,
                 content: att.data.base64EncodedString())
         }
-        for att in encodedAttachments {
+        for (attachment, att) in zip(draftAttachments, encodedAttachments) {
             userContent.append(
                 OpenClawChatMessageContent(
                     type: att.type,
@@ -1241,6 +1241,7 @@ public final class OpenClawChatViewModel {
                     thinkingSignature: nil,
                     mimeType: att.mimeType,
                     fileName: att.fileName,
+                    durationSeconds: attachment.durationSeconds,
                     content: AnyCodable(att.content),
                     id: nil,
                     name: nil,
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteComposerViews.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteComposerViews.swift
new file mode 100644
index 00000000000..b4a55b579aa
--- /dev/null
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteComposerViews.swift
@@ -0,0 +1,107 @@
+import SwiftUI
+
+/// Host-provided voice-note recording state for the shared chat composer.
+public struct OpenClawChatVoiceNoteControl {
+    public var recorder: OpenClawVoiceNoteRecorder
+    public var isTalkActive: Bool
+
+    public init(recorder: OpenClawVoiceNoteRecorder, isTalkActive: Bool) {
+        self.recorder = recorder
+        self.isTalkActive = isTalkActive
+    }
+}
+
+struct OpenClawVoiceNoteButton: View {
+    let control: OpenClawChatVoiceNoteControl
+    let compact: Bool
+    let isComposerEnabled: Bool
+    let isAttachmentInputEnabled: Bool
+
+    var isRecordingEnabled: Bool {
+        self.isComposerEnabled
+            && self.isAttachmentInputEnabled
+            && !self.control.isTalkActive
+            && !self.control.recorder.isRequestingPermission
+    }
+
+    var body: some View {
+        Button {
+            Task { await self.control.recorder.start() }
+        } label: {
+            if self.control.recorder.isRequestingPermission {
+                ProgressView()
+                    .controlSize(.mini)
+            } else {
+                Image(systemName: "mic")
+                    .font(OpenClawChatTypography.display(size: 15, weight: .semibold, relativeTo: .subheadline))
+            }
+        }
+        .help("Record Voice Note")
+        .accessibilityLabel("Record voice note")
+        .accessibilityIdentifier("chat-voice-note-record")
+        .modifier(VoiceNoteButtonChrome(compact: self.compact))
+        .controlSize(.small)
+        .foregroundStyle(.secondary)
+        .contentShape(Rectangle())
+        .disabled(!self.isRecordingEnabled)
+    }
+}
+
+struct OpenClawVoiceNoteRecordingRow: View {
+    let recorder: OpenClawVoiceNoteRecorder
+
+    var body: some View {
+        HStack(spacing: 10) {
+            Circle()
+                .fill(OpenClawChatTheme.danger)
+                .frame(width: 9, height: 9)
+
+            Text("Recording")
+                .font(OpenClawChatTypography.footnoteSemiBold)
+
+            Text(openClawVoiceNoteDurationLabel(self.recorder.elapsedSeconds))
+                .font(OpenClawChatTypography.mono(size: 13, relativeTo: .footnote))
+                .foregroundStyle(.secondary)
+
+            Spacer(minLength: 8)
+
+            Button {
+                self.recorder.cancel()
+            } label: {
+                Image(systemName: "xmark")
+            }
+            .buttonStyle(.bordered)
+            .controlSize(.small)
+            .accessibilityLabel("Cancel voice note")
+
+            Button {
+                self.recorder.finish()
+            } label: {
+                Image(systemName: "checkmark")
+            }
+            .buttonStyle(.borderedProminent)
+            .controlSize(.small)
+            .accessibilityLabel("Finish voice note")
+        }
+        .padding(.horizontal, 14)
+        .padding(.vertical, 10)
+        .background(
+            RoundedRectangle(cornerRadius: 14, style: .continuous)
+                .fill(OpenClawChatTheme.composerField)
+                .overlay(
+                    RoundedRectangle(cornerRadius: 14, style: .continuous)
+                        .strokeBorder(OpenClawChatTheme.composerBorder)))
+    }
+}
+
+private struct VoiceNoteButtonChrome: ViewModifier {
+    let compact: Bool
+
+    func body(content: Content) -> some View {
+        if self.compact {
+            content.buttonStyle(.plain)
+        } else {
+            content.buttonStyle(.bordered)
+        }
+    }
+}
diff --git a/apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteRecorder.swift b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteRecorder.swift
new file mode 100644
index 00000000000..0b488148830
--- /dev/null
+++ b/apps/shared/OpenClawKit/Sources/OpenClawChatUI/VoiceNoteRecorder.swift
@@ -0,0 +1,305 @@
+import AVFAudio
+import Foundation
+import Observation
+
+private let voiceNoteMaximumDurationSeconds: TimeInterval = 180
+
+/// Audio capture operations driven by the voice-note recorder state machine.
+@MainActor
+public protocol VoiceNoteAudioCapture: AnyObject {
+    /// Requests microphone access when recording is first used.
+    func requestPermission() async -> Bool
+
+    /// Starts writing an audio recording to the supplied URL.
+    func start(url: URL) throws
+
+    /// Stops capture and returns the recorded duration in seconds.
+    func stop() -> TimeInterval
+
+    /// Stops and discards the active capture.
+    func cancel()
+}
+
+/// A completed voice-note recording ready to stage as a chat attachment.
+public struct OpenClawVoiceNoteRecording: Equatable, Sendable {
+    public let fileURL: URL
+    public let durationSeconds: TimeInterval
+
+    public init(fileURL: URL, durationSeconds: TimeInterval) {
+        self.fileURL = fileURL
+        self.durationSeconds = durationSeconds
+    }
+}
+
+/// Main-actor voice-note recorder with explicit permission and capture states.
+@MainActor
+@Observable
+public final class OpenClawVoiceNoteRecorder {
+    public enum State: Equatable {
+        case idle
+        case requestingPermission
+        case recording(startedAt: Date, fileURL: URL)
+        case finished(recording: OpenClawVoiceNoteRecording)
+        case failed(message: String)
+    }
+
+    public static let maximumDurationSeconds = voiceNoteMaximumDurationSeconds
+
+    public private(set) var state: State = .idle
+    public private(set) var elapsedSeconds: TimeInterval = 0
+
+    @ObservationIgnored public var onRecordingActiveChanged: (@MainActor (Bool) -> Void)?
+
+    @ObservationIgnored private let capture: any VoiceNoteAudioCapture
+    @ObservationIgnored private let durationLimit: TimeInterval
+    @ObservationIgnored private let timerIntervalNanoseconds: UInt64
+    @ObservationIgnored private let now: () -> Date
+    @ObservationIgnored private var timerTask: Task<Void, Never>?
+
+    /// Creates a recorder backed by the system audio recorder.
+    public convenience init() {
+        self.init(capture: OpenClawVoiceNoteAudioCapture())
+    }
+
+    /// Creates a recorder with an injectable audio capture seam.
+    public init(
+        capture: any VoiceNoteAudioCapture,
+        durationLimit: TimeInterval = OpenClawVoiceNoteRecorder.maximumDurationSeconds,
+        timerIntervalNanoseconds: UInt64 = 100_000_000,
+        now: @escaping () -> Date = Date.init)
+    {
+        self.capture = capture
+        self.durationLimit = durationLimit
+        self.timerIntervalNanoseconds = timerIntervalNanoseconds
+        self.now = now
+    }
+
+    deinit {
+        self.timerTask?.cancel()
+    }
+
+    public var isRecording: Bool {
+        if case .recording = self.state { return true }
+        return false
+    }
+
+    public var isRequestingPermission: Bool {
+        self.state == .requestingPermission
+    }
+
+    public var errorMessage: String? {
+        guard case let .failed(message) = self.state else { return nil }
+        return message
+    }
+
+    public var completedRecording: OpenClawVoiceNoteRecording? {
+        guard case let .finished(recording) = self.state else { return nil }
+        return recording
+    }
+
+    /// Requests permission if needed and starts a new recording.
+    @discardableResult
+    public func start() async -> Bool {
+        guard self.state == .idle || self.errorMessage != nil else { return false }
+
+        self.elapsedSeconds = 0
+        self.state = .requestingPermission
+        guard await self.capture.requestPermission() else {
+            self.fail(message: String(localized: "Microphone access is required. Enable it in Settings."))
+            return false
+        }
+        guard self.state == .requestingPermission else { return false }
+
+        let fileURL = self.makeTemporaryFileURL()
+        self.onRecordingActiveChanged?(true)
+        do {
+            try self.capture.start(url: fileURL)
+        } catch {
+            try? FileManager.default.removeItem(at: fileURL)
+            self.capture.cancel()
+            self.onRecordingActiveChanged?(false)
+            self.fail(message: String(localized: "Could not start recording: \(error.localizedDescription)"))
+            return false
+        }
+
+        self.state = .recording(startedAt: self.now(), fileURL: fileURL)
+        self.startTimer()
+        return true
+    }
+
+    /// Finishes the active recording and publishes its attachment handoff.
+    @discardableResult
+    public func finish() -> OpenClawVoiceNoteRecording? {
+        guard case let .recording(_, fileURL) = self.state else { return nil }
+
+        self.timerTask?.cancel()
+        self.timerTask = nil
+        let duration = max(0, self.capture.stop())
+        let recording = OpenClawVoiceNoteRecording(fileURL: fileURL, durationSeconds: duration)
+        self.elapsedSeconds = duration
+        self.state = .finished(recording: recording)
+        self.onRecordingActiveChanged?(false)
+        return recording
+    }
+
+    /// Cancels permission or capture and removes any temporary audio file.
+    public func cancel() {
+        let fileURL: URL? = switch self.state {
+        case let .recording(_, fileURL):
+            fileURL
+        case let .finished(recording):
+            recording.fileURL
+        default:
+            nil
+        }
+
+        self.timerTask?.cancel()
+        self.timerTask = nil
+        self.capture.cancel()
+        if let fileURL {
+            try? FileManager.default.removeItem(at: fileURL)
+        }
+        let wasRecording = self.isRecording
+        self.elapsedSeconds = 0
+        self.state = .idle
+        if wasRecording {
+            self.onRecordingActiveChanged?(false)
+        }
+    }
+
+    /// Clears a completed handoff after the composer has staged it.
+    public func clearCompletedRecording() {
+        guard case .finished = self.state else { return }
+        self.state = .idle
+        self.elapsedSeconds = 0
+    }
+
+    private func startTimer() {
+        self.timerTask?.cancel()
+        self.timerTask = Task { [weak self] in
+            guard let self else { return }
+            while !Task.isCancelled {
+                try? await Task.sleep(nanoseconds: self.timerIntervalNanoseconds)
+                guard !Task.isCancelled else { return }
+                guard case let .recording(startedAt, _) = self.state else { return }
+                self.elapsedSeconds = max(0, self.now().timeIntervalSince(startedAt))
+                if self.elapsedSeconds >= self.durationLimit {
+                    self.finish()
+                    return
+                }
+            }
+        }
+    }
+
+    private func fail(message: String) {
+        self.timerTask?.cancel()
+        self.timerTask = nil
+        let wasRecording = self.isRecording
+        self.elapsedSeconds = 0
+        self.state = .failed(message: message)
+        if wasRecording {
+            self.onRecordingActiveChanged?(false)
+        }
+    }
+
+    private func makeTemporaryFileURL() -> URL {
+        let formatter = DateFormatter()
+        formatter.locale = Locale(identifier: "en_US_POSIX")
+        formatter.dateFormat = "yyyyMMdd-HHmmss"
+        let timestamp = formatter.string(from: self.now())
+        return FileManager.default.temporaryDirectory
+            .appendingPathComponent("voice-note-\(timestamp).m4a")
+    }
+}
+
+/// AVAudioRecorder-backed AAC voice-note capture.
+@MainActor
+public final class OpenClawVoiceNoteAudioCapture: VoiceNoteAudioCapture {
+    private var recorder: AVAudioRecorder?
+    private var ownsAudioSession = false
+
+    public init() {}
+
+    public func requestPermission() async -> Bool {
+        #if os(iOS)
+        switch AVAudioApplication.shared.recordPermission {
+        case .granted:
+            return true
+        case .denied:
+            return false
+        case .undetermined:
+            return await withCheckedContinuation { continuation in
+                AVAudioApplication.requestRecordPermission { granted in
+                    continuation.resume(returning: granted)
+                }
+            }
+        @unknown default:
+            return false
+        }
+        #else
+        return false
+        #endif
+    }
+
+    public func start(url: URL) throws {
+        do {
+            #if os(iOS)
+            let session = AVAudioSession.sharedInstance()
+            try session.setCategory(.record, mode: .measurement, options: [.allowBluetoothHFP])
+            try session.setActive(true)
+            self.ownsAudioSession = true
+            #endif
+
+            let settings: [String: Any] = [
+                AVFormatIDKey: kAudioFormatMPEG4AAC,
+                AVSampleRateKey: 24000,
+                AVNumberOfChannelsKey: 1,
+                AVEncoderBitRateKey: 32000,
+                AVEncoderAudioQualityKey: AVAudioQuality.medium.rawValue,
+            ]
+            let recorder = try AVAudioRecorder(url: url, settings: settings)
+            guard recorder.record() else {
+                throw NSError(
+                    domain: "OpenClawVoiceNoteAudioCapture",
+                    code: 1,
+                    userInfo: [NSLocalizedDescriptionKey: "Audio recorder refused to start"])
+            }
+            self.recorder = recorder
+        } catch {
+            self.deactivateAudioSession()
+            throw error
+        }
+    }
+
+    public func stop() -> TimeInterval {
+        guard let recorder = self.recorder else { return 0 }
+        let duration = recorder.currentTime
+        recorder.stop()
+        self.recorder = nil
+        self.deactivateAudioSession()
+        return duration
+    }
+
+    public func cancel() {
+        self.recorder?.stop()
+        self.recorder = nil
+        self.deactivateAudioSession()
+    }
+
+    private func deactivateAudioSession() {
+        #if os(iOS)
+        guard self.ownsAudioSession else { return }
+        self.ownsAudioSession = false
+        try? AVAudioSession.sharedInstance().setActive(false, options: .notifyOthersOnDeactivation)
+        #endif
+    }
+}
+
+func openClawVoiceNoteDurationLabel(_ durationSeconds: TimeInterval) -> String {
+    guard durationSeconds.isFinite else { return "0:00" }
+    let boundedDuration = min(
+        max(0, durationSeconds),
+        voiceNoteMaximumDurationSeconds)
+    let totalSeconds = Int(boundedDuration)
+    return String(format: "%d:%02d", totalSeconds / 60, totalSeconds % 60)
+}
diff --git a/apps/shared/OpenClawKit/Tests/OpenClawKitTests/ChatViewModelAttachmentTests.swift b/apps/shared/OpenClawKit/Tests/OpenClawKitTests/ChatViewModelAttachmentTests.swift
index 74f1c902282..776bb16a6f1 100644
--- a/apps/shared/OpenClawKit/Tests/OpenClawKitTests/ChatViewModelAttachmentTests.swift
+++ b/apps/shared/OpenClawKit/Tests/OpenClawKitTests/ChatViewModelAttachmentTests.swift
@@ -6,7 +6,55 @@ import UniformTypeIdentifiers
 import XCTest
 @testable import OpenClawChatUI
 
+private actor AttachmentSendCapture {
+    private(set) var attachments: [OpenClawChatAttachmentPayload] = []
+
+    func store(_ attachments: [OpenClawChatAttachmentPayload]) {
+        self.attachments = attachments
+    }
+
+    func count() -> Int {
+        self.attachments.count
+    }
+
+    func first() -> OpenClawChatAttachmentPayload? {
+        self.attachments.first
+    }
+}
+
+private actor AttachmentHealthGate {
+    private var entered = false
+    private var released = false
+    private var continuation: CheckedContinuation<Void, Never>?
+
+    func wait() async {
+        self.entered = true
+        guard !self.released else { return }
+        await withCheckedContinuation { continuation in
+            self.continuation = continuation
+        }
+    }
+
+    func hasEntered() -> Bool {
+        self.entered
+    }
+
+    func release() {
+        self.released = true
+        self.continuation?.resume()
+        self.continuation = nil
+    }
+}
+
 private struct AttachmentProcessingTransport: OpenClawChatTransport {
+    let capture: AttachmentSendCapture?
+    let healthGate: AttachmentHealthGate?
+
+    init(capture: AttachmentSendCapture? = nil, healthGate: AttachmentHealthGate? = nil) {
+        self.capture = capture
+        self.healthGate = healthGate
+    }
+
     func requestHistory(sessionKey _: String) async throws -> OpenClawChatHistoryPayload {
         throw NSError(domain: "ChatViewModelAttachmentTests", code: 1)
     }
@@ -15,13 +63,15 @@ private struct AttachmentProcessingTransport: OpenClawChatTransport {
         sessionKey _: String,
         message _: String,
         thinking _: String,
-        idempotencyKey _: String,
-        attachments _: [OpenClawChatAttachmentPayload]) async throws -> OpenClawChatSendResponse
+        idempotencyKey: String,
+        attachments: [OpenClawChatAttachmentPayload]) async throws -> OpenClawChatSendResponse
     {
-        throw NSError(domain: "ChatViewModelAttachmentTests", code: 2)
+        await self.capture?.store(attachments)
+        return OpenClawChatSendResponse(runId: idempotencyKey, status: "started")
     }
 
     func requestHealth(timeoutMs _: Int) async throws -> Bool {
+        await self.healthGate?.wait()
         true
     }
 
@@ -106,4 +156,160 @@ final class ChatViewModelAttachmentTests: XCTestCase {
         let errorText = await MainActor.run { viewModel.errorText }
         XCTAssertNil(errorText)
     }
+
+    func testVoiceNoteAttachmentStagesAudioAndDeletesTemporaryFile() async throws {
+        let fileURL = FileManager.default.temporaryDirectory
+            .appendingPathComponent("voice-note-20260706-120000.m4a")
+        let data = Data("voice-note-data".utf8)
+        try data.write(to: fileURL)
+        let viewModel = await MainActor.run {
+            OpenClawChatViewModel(sessionKey: "main", transport: AttachmentProcessingTransport())
+        }
+
+        await viewModel.addVoiceNoteAttachment(fileURL: fileURL, durationSeconds: 8.4)
+
+        let attachment = try await MainActor.run { () throws -> (Data, String, String, String, Double?, Bool) in
+            let attachment = try XCTUnwrap(viewModel.attachments.first)
+            return (
+                attachment.data,
+                attachment.fileName,
+                attachment.mimeType,
+                attachment.type,
+                attachment.durationSeconds,
+                attachment.preview == nil)
+        }
+        XCTAssertEqual(attachment.0, data)
+        XCTAssertEqual(attachment.1, "voice-note-20260706-120000.m4a")
+        XCTAssertEqual(attachment.2, "audio/mp4")
+        XCTAssertEqual(attachment.3, "file")
+        XCTAssertEqual(attachment.4, 8.4)
+        XCTAssertTrue(attachment.5)
+        XCTAssertFalse(FileManager.default.fileExists(atPath: fileURL.path))
+    }
+
+    func testOversizeVoiceNoteIsRejectedAndDeleted() async throws {
+        let fileURL = FileManager.default.temporaryDirectory
+            .appendingPathComponent("voice-note-oversize.m4a")
+        try Data(repeating: 0x41, count: 5_000_001).write(to: fileURL)
+        let viewModel = await MainActor.run {
+            OpenClawChatViewModel(sessionKey: "main", transport: AttachmentProcessingTransport())
+        }
+
+        await viewModel.addVoiceNoteAttachment(fileURL: fileURL, durationSeconds: 180)
+
+        let result = await MainActor.run { (viewModel.attachments.count, viewModel.errorText) }
+        XCTAssertEqual(result.0, 0)
+        XCTAssertEqual(result.1, "Voice note exceeds the 5 MB attachment limit")
+        XCTAssertFalse(FileManager.default.fileExists(atPath: fileURL.path))
+    }
+
+    func testVoiceNoteSendUsesExistingAttachmentPayloadAndOptimisticDuration() async throws {
+        let capture = AttachmentSendCapture()
+        let transport = AttachmentProcessingTransport(capture: capture)
+        let fileURL = FileManager.default.temporaryDirectory
+            .appendingPathComponent("voice-note-20260706-120001.m4a")
+        let data = Data("encoded-voice-note".utf8)
+        try data.write(to: fileURL)
+        let viewModel = await MainActor.run {
+            OpenClawChatViewModel(sessionKey: "main", transport: transport)
+        }
+
+        await viewModel.addVoiceNoteAttachment(fileURL: fileURL, durationSeconds: 21.2)
+        await MainActor.run { viewModel.send() }
+        try await waitUntil("voice note sent") {
+            await capture.count() == 1
+        }
+
+        let capturedPayload = await capture.first()
+        let payload = try XCTUnwrap(capturedPayload)
+        XCTAssertEqual(payload.type, "file")
+        XCTAssertEqual(payload.mimeType, "audio/mp4")
+        XCTAssertEqual(payload.fileName, "voice-note-20260706-120001.m4a")
+        XCTAssertEqual(payload.content, data.base64EncodedString())
+
+        let optimisticAudio = await MainActor.run {
+            viewModel.messages.last?.content.first { $0.mimeType == "audio/mp4" }
+        }
+        XCTAssertEqual(optimisticAudio?.type, "file")
+        XCTAssertEqual(optimisticAudio?.mimeType, "audio/mp4")
+        XCTAssertEqual(optimisticAudio?.durationSeconds, 21.2)
+    }
+
+    func testVoiceNoteSendKeepsCapturedDurationWhenDraftChangesDuringHealthCheck() async throws {
+        let capture = AttachmentSendCapture()
+        let healthGate = AttachmentHealthGate()
+        let transport = AttachmentProcessingTransport(capture: capture, healthGate: healthGate)
+        let draftAttachment = OpenClawPendingAttachment(
+            url: nil,
+            data: Data("draft-audio".utf8),
+            fileName: "draft.m4a",
+            mimeType: "audio/mp4",
+            preview: nil,
+            durationSeconds: 21.2)
+        let replacementAttachment = OpenClawPendingAttachment(
+            url: nil,
+            data: Data("replacement-audio".utf8),
+            fileName: "replacement.m4a",
+            mimeType: "audio/mp4",
+            preview: nil,
+            durationSeconds: 99)
+        let viewModel = await MainActor.run {
+            let viewModel = OpenClawChatViewModel(sessionKey: "main", transport: transport)
+            viewModel.attachments = [draftAttachment]
+            return viewModel
+        }
+
+        await MainActor.run { viewModel.send() }
+        try await waitUntil("health check started") {
+            await healthGate.hasEntered()
+        }
+        await MainActor.run {
+            viewModel.removeAttachment(draftAttachment.id)
+            viewModel.attachments.append(replacementAttachment)
+        }
+        await healthGate.release()
+        try await waitUntil("voice note sent") {
+            await capture.count() == 1
+        }
+
+        let optimisticAudio = await MainActor.run {
+            viewModel.messages.last?.content.first { $0.mimeType == "audio/mp4" }
+        }
+        XCTAssertEqual(optimisticAudio?.fileName, "draft.m4a")
+        XCTAssertEqual(optimisticAudio?.durationSeconds, 21.2)
+    }
+
+    @MainActor
+    func testCanonicalVoiceNotePreservesOptimisticDuration() throws {
+        let localAudio = OpenClawChatMessageContent(
+            type: "file",
+            text: nil,
+            mimeType: "audio/mp4",
+            fileName: "voice-note-local.m4a",
+            durationSeconds: 14.6,
+            content: AnyCodable("local"))
+        let canonicalAudio = OpenClawChatMessageContent(
+            type: "file",
+            text: nil,
+            mimeType: "audio/mp4",
+            fileName: "media-1.m4a",
+            content: AnyCodable("canonical"))
+        let existing = OpenClawChatMessage(
+            role: "user",
+            content: [localAudio],
+            timestamp: nil,
+            idempotencyKey: "run:user")
+        let incoming = OpenClawChatMessage(
+            role: "user",
+            content: [canonicalAudio],
+            timestamp: nil,
+            idempotencyKey: "run:user")
+
+        let adopted = OpenClawChatViewModel.adoptingCanonicalMessage(incoming, over: existing)
+
+        let audio = try XCTUnwrap(adopted.content.first)
+        XCTAssertEqual(audio.fileName, "media-1.m4a")
+        XCTAssertEqual(audio.content, AnyCodable("canonical"))
+        XCTAssertEqual(audio.durationSeconds, 14.6)
+    }
 }
diff --git a/apps/shared/OpenClawKit/Tests/OpenClawKitTests/VoiceNoteRecorderTests.swift b/apps/shared/OpenClawKit/Tests/OpenClawKitTests/VoiceNoteRecorderTests.swift
new file mode 100644
index 00000000000..5178cd014e2
--- /dev/null
+++ b/apps/shared/OpenClawKit/Tests/OpenClawKitTests/VoiceNoteRecorderTests.swift
@@ -0,0 +1,162 @@
+import Foundation
+import XCTest
+@testable import OpenClawChatUI
+
+@MainActor
+private final class FakeVoiceNoteAudioCapture: VoiceNoteAudioCapture {
+    var permissionGranted = true
+    var duration: TimeInterval = 12.5
+    var startError: Error?
+    var startCount = 0
+    var cancelCount = 0
+    var activeURL: URL?
+    var onStart: (() -> Void)?
+
+    func requestPermission() async -> Bool {
+        self.permissionGranted
+    }
+
+    func start(url: URL) throws {
+        self.onStart?()
+        if let startError { throw startError }
+        self.startCount += 1
+        self.activeURL = url
+        try Data("voice-note".utf8).write(to: url)
+    }
+
+    func stop() -> TimeInterval {
+        self.duration
+    }
+
+    func cancel() {
+        self.cancelCount += 1
+    }
+}
+
+final class VoiceNoteRecorderTests: XCTestCase {
+    @MainActor
+    func testStartAndFinishProduceRecordingWithDuration() async throws {
+        let capture = FakeVoiceNoteAudioCapture()
+        let recorder = OpenClawVoiceNoteRecorder(capture: capture)
+        var activeChanges: [Bool] = []
+        recorder.onRecordingActiveChanged = { activeChanges.append($0) }
+        capture.onStart = { XCTAssertEqual(activeChanges, [true]) }
+
+        let started = await recorder.start()
+        XCTAssertTrue(started)
+        guard case .recording = recorder.state else {
+            return XCTFail("Expected recording state")
+        }
+
+        let result = try XCTUnwrap(recorder.finish())
+        XCTAssertEqual(result.durationSeconds, 12.5)
+        XCTAssertEqual(recorder.state, .finished(recording: result))
+        XCTAssertEqual(recorder.completedRecording, result)
+        XCTAssertEqual(activeChanges, [true, false])
+        XCTAssertTrue(FileManager.default.fileExists(atPath: result.fileURL.path))
+        try FileManager.default.removeItem(at: result.fileURL)
+    }
+
+    @MainActor
+    func testCancelReturnsToIdleAndDeletesTemporaryFile() async throws {
+        let capture = FakeVoiceNoteAudioCapture()
+        let recorder = OpenClawVoiceNoteRecorder(capture: capture)
+
+        let started = await recorder.start()
+        XCTAssertTrue(started)
+        let fileURL = try XCTUnwrap(capture.activeURL)
+        XCTAssertTrue(FileManager.default.fileExists(atPath: fileURL.path))
+
+        recorder.cancel()
+
+        XCTAssertEqual(recorder.state, .idle)
+        XCTAssertFalse(FileManager.default.fileExists(atPath: fileURL.path))
+        XCTAssertEqual(capture.cancelCount, 1)
+    }
+
+    @MainActor
+    func testDurationCapAutoFinishes() async throws {
+        let capture = FakeVoiceNoteAudioCapture()
+        capture.duration = 0.25
+        let recorder = OpenClawVoiceNoteRecorder(
+            capture: capture,
+            durationLimit: 0,
+            timerIntervalNanoseconds: 1_000_000)
+
+        let started = await recorder.start()
+        XCTAssertTrue(started)
+        try await waitUntil("voice note auto-finished") {
+            await MainActor.run { recorder.completedRecording != nil }
+        }
+
+        let result = try XCTUnwrap(recorder.completedRecording)
+        XCTAssertEqual(result.durationSeconds, 0.25)
+        XCTAssertEqual(recorder.state, .finished(recording: result))
+        try FileManager.default.removeItem(at: result.fileURL)
+    }
+
+    @MainActor
+    func testPermissionDeniedBecomesUserVisibleFailure() async {
+        let capture = FakeVoiceNoteAudioCapture()
+        capture.permissionGranted = false
+        let recorder = OpenClawVoiceNoteRecorder(capture: capture)
+
+        let started = await recorder.start()
+        XCTAssertFalse(started)
+        guard case let .failed(message) = recorder.state else {
+            return XCTFail("Expected failure state")
+        }
+        XCTAssertTrue(message.contains("Microphone access"))
+        XCTAssertEqual(capture.startCount, 0)
+    }
+
+    @MainActor
+    func testCaptureStartFailureReleasesRecordingActivity() async {
+        let capture = FakeVoiceNoteAudioCapture()
+        capture.startError = NSError(domain: "VoiceNoteRecorderTests", code: 1)
+        let recorder = OpenClawVoiceNoteRecorder(capture: capture)
+        var activeChanges: [Bool] = []
+        recorder.onRecordingActiveChanged = { activeChanges.append($0) }
+
+        let started = await recorder.start()
+
+        XCTAssertFalse(started)
+        XCTAssertEqual(activeChanges, [true, false])
+        guard case .failed = recorder.state else {
+            return XCTFail("Expected failure state")
+        }
+    }
+
+    @MainActor
+    func testStartIsRefusedWhileAlreadyRecording() async {
+        let capture = FakeVoiceNoteAudioCapture()
+        let recorder = OpenClawVoiceNoteRecorder(capture: capture)
+
+        let firstStart = await recorder.start()
+        let secondStart = await recorder.start()
+        XCTAssertTrue(firstStart)
+        XCTAssertFalse(secondStart)
+        XCTAssertEqual(capture.startCount, 1)
+        recorder.cancel()
+    }
+
+    func testDurationLabelBoundsMalformedHistoryValues() {
+        XCTAssertEqual(openClawVoiceNoteDurationLabel(.infinity), "0:00")
+        XCTAssertEqual(openClawVoiceNoteDurationLabel(.nan), "0:00")
+        XCTAssertEqual(openClawVoiceNoteDurationLabel(1e100), "3:00")
+        XCTAssertEqual(openClawVoiceNoteDurationLabel(-1), "0:00")
+    }
+
+    @MainActor
+    func testRecordButtonRequiresAttachmentInput() {
+        let recorder = OpenClawVoiceNoteRecorder(capture: FakeVoiceNoteAudioCapture())
+        let control = OpenClawChatVoiceNoteControl(recorder: recorder, isTalkActive: false)
+        let button = OpenClawVoiceNoteButton(
+            control: control,
+            compact: false,
+            isComposerEnabled: true,
+            isAttachmentInputEnabled: false)
+
+        XCTAssertFalse(button.isRecordingEnabled)
+    }
+}