fix(imessage): send TTS audio as voice messages

extensions/imessage/src/channel.runtime.ts

@@ -17,6 +17,7 @@ export async function sendIMessageOutbound(params: {
   text: string;
   mediaUrl?: string;
   mediaLocalRoots?: readonly string[];
+  audioAsVoice?: boolean;
   accountId?: string;
   deps?: { [channelId: string]: unknown };
   replyToId?: string;
@@ -36,6 +37,7 @@ export async function sendIMessageOutbound(params: {
     config: params.cfg,
     ...(params.mediaUrl ? { mediaUrl: params.mediaUrl } : {}),
     ...(params.mediaLocalRoots?.length ? { mediaLocalRoots: params.mediaLocalRoots } : {}),
+    ...(params.audioAsVoice ? { audioAsVoice: true } : {}),
     maxBytes,
     accountId: params.accountId ?? undefined,
     replyToId: params.replyToId ?? undefined,

extensions/imessage/src/channel.ts

@@ -113,11 +113,16 @@ const imessageMessageAdapter = defineChannelMessageAdapter({
         text: ctx.text,
         mediaUrl: ctx.mediaUrl,
         mediaLocalRoots: ctx.mediaLocalRoots,
+        audioAsVoice: ctx.audioAsVoice,
         accountId: ctx.accountId ?? undefined,
         deps: (ctx as typeof ctx & IMessageMessageContextExtras).deps,
         replyToId: ctx.replyToId ?? undefined,
       });
-      return toIMessageMessageSendResult(result, "media", ctx.replyToId);
+      return toIMessageMessageSendResult(
+        result,
+        ctx.audioAsVoice ? "voice" : "media",
+        ctx.replyToId,
+      );
     },
   },
 });
@@ -362,6 +367,7 @@ export const imessagePlugin: ChannelPlugin<ResolvedIMessageAccount, IMessageProb
           text,
           mediaUrl,
           mediaLocalRoots,
+          audioAsVoice,
           accountId,
           deps,
           replyToId,
@@ -374,6 +380,7 @@ export const imessagePlugin: ChannelPlugin<ResolvedIMessageAccount, IMessageProb
             text,
             mediaUrl,
             mediaLocalRoots,
+            audioAsVoice,
             accountId: accountId ?? undefined,
             deps,
             replyToId: replyToId ?? undefined,

extensions/imessage/src/send.test.ts

@@ -168,6 +168,38 @@ describe("sendMessageIMessage receipts", () => {
     expect(result.receipt.sentAt).toBeGreaterThan(0);
   });
 
+  it("sends audioAsVoice media through send-attachment audio transport", async () => {
+    const client = createClient({ message_id: 12345 });
+    const runCliJson = vi.fn().mockResolvedValueOnce({ messageId: "p:0/voice-guid" });
+
+    const result = await sendMessageIMessage("chat_guid:chat-1", "", {
+      config: IMESSAGE_TEST_CFG,
+      client,
+      mediaUrl: "/tmp/voice.caf",
+      audioAsVoice: true,
+      resolveAttachmentImpl: async () => ({ path: "/tmp/voice.caf", contentType: "audio/x-caf" }),
+      runCliJson,
+    });
+
+    expect(result.messageId).toBe("p:0/voice-guid");
+    expect(runCliJson.mock.calls).toEqual([
+      [
+        [
+          "send-attachment",
+          "--chat",
+          "chat-1",
+          "--file",
+          "/tmp/voice.caf",
+          "--audio",
+          "--transport",
+          "auto",
+        ],
+      ],
+    ]);
+    expect(result.receipt.parts.map((part) => part.kind)).toEqual(["voice"]);
+    expect(client["request"]).not.toHaveBeenCalled();
+  });
+
   it("resolves chat_id media-only payloads before using send-attachment", async () => {
     const client = createClient({ message_id: 12345 });
     const runCliJson = vi

extensions/imessage/src/send.ts

@@ -48,6 +48,7 @@ type IMessageSendOpts = {
   mediaUrl?: string;
   mediaLocalRoots?: readonly string[];
   mediaReadFile?: (filePath: string) => Promise<Buffer>;
+  audioAsVoice?: boolean;
   maxBytes?: number;
   timeoutMs?: number;
   chatId?: number;
@@ -729,6 +730,7 @@ async function trySendAttachmentForTarget(params: {
   target: ReturnType<typeof parseIMessageTarget>;
   service?: IMessageService;
   filePath: string;
+  audioAsVoice?: boolean;
   echoText?: string;
   runCliJson: (args: readonly string[]) => Promise<Record<string, unknown>>;
   resolveMessageGuidImpl?: IMessageSendOpts["resolveMessageGuidImpl"];
@@ -758,6 +760,7 @@ async function trySendAttachmentForTarget(params: {
       attachmentChatTarget,
       "--file",
       params.filePath,
+      ...(params.audioAsVoice ? ["--audio"] : []),
       "--transport",
       "auto",
     ]);
@@ -822,7 +825,7 @@ async function trySendAttachmentForTarget(params: {
     receipt: createIMessageSendReceipt({
       messageId,
       target: params.target,
-      kind: "media",
+      kind: params.audioAsVoice ? "voice" : "media",
     }),
   };
 }
@@ -915,6 +918,7 @@ export async function sendMessageIMessage(
       target,
       service,
       filePath,
+      audioAsVoice: opts.audioAsVoice,
       echoText: attachmentEchoText,
       runCliJson,
       resolveMessageGuidImpl: opts.resolveMessageGuidImpl,

extensions/imessage/src/shared.ts

@@ -83,6 +83,13 @@ export function createIMessagePluginBase(params: {
     capabilities: {
       chatTypes: ["direct", "group"],
       media: true,
+      tts: {
+        voice: {
+          synthesisTarget: "audio-file",
+          audioFileFormats: ["mp3", "caf", "audio/mpeg", "audio/x-caf"],
+          preferAudioFileFormat: "caf",
+        },
+      },
       reactions: true,
       edit: true,
       unsend: true,

extensions/imessage/src/test-plugin.test.ts

@@ -112,6 +112,14 @@ describe("createIMessageTestPlugin", () => {
     });
   });
 
+  it("declares native iMessage voice memo TTS delivery", () => {
+    expect(imessagePlugin.capabilities.tts?.voice).toStrictEqual({
+      synthesisTarget: "audio-file",
+      audioFileFormats: ["mp3", "caf", "audio/mpeg", "audio/x-caf"],
+      preferAudioFileFormat: "caf",
+    });
+  });
+
   it("preserves the local approval prompt suppressor through attached-result composition", () => {
     const suppressor = imessagePlugin.outbound?.shouldSuppressLocalPayloadPrompt;
     if (!suppressor) {
@@ -208,7 +216,7 @@ describe("createIMessageTestPlugin", () => {
     const sendIMessage = async (
       _to: string,
       _text: string,
-      opts?: { mediaUrl?: string; replyToId?: string },
+      opts?: { mediaUrl?: string; replyToId?: string; audioAsVoice?: boolean },
     ) => {
       const messageId = opts?.mediaUrl ? "imsg-media-1" : "imsg-text-1";
       return {
@@ -216,7 +224,7 @@ describe("createIMessageTestPlugin", () => {
         sentText: opts?.mediaUrl ? "<media:image>" : "hello",
         receipt: createMessageReceiptFromOutboundResults({
           results: [{ channel: "imessage", messageId }],
-          kind: opts?.mediaUrl ? "media" : "text",
+          kind: opts?.audioAsVoice ? "voice" : opts?.mediaUrl ? "media" : "text",
           ...(opts?.replyToId ? { replyToId: opts.replyToId } : {}),
         }),
       };
@@ -247,11 +255,13 @@ describe("createIMessageTestPlugin", () => {
             text: "caption",
             mediaUrl: "/tmp/image.png",
             mediaLocalRoots: ["/tmp"],
+            audioAsVoice: true,
             deps: { imessage: sendIMessage },
           } as Parameters<typeof sendMedia>[0] & {
             deps: { imessage: typeof sendIMessage };
           });
           expect(result.receipt.platformMessageIds).toEqual(["imsg-media-1"]);
+          expect(result.receipt.parts.map((part) => part.kind)).toEqual(["voice"]);
         },
         replyTo: async () => {
           const result = await sendText({

src/channels/plugins/tts-capabilities.test.ts

@@ -39,6 +39,7 @@ describe("resolveChannelTtsVoiceDelivery", () => {
               voice: {
                 synthesisTarget: "audio-file",
                 audioFileFormats: ["mp3", "caf", "audio/mpeg", "audio/x-caf"],
+                preferAudioFileFormat: "caf",
               },
             },
           }),
@@ -89,6 +90,7 @@ describe("resolveChannelTtsVoiceDelivery", () => {
     expect(resolveChannelTtsVoiceDelivery("imessage")).toEqual({
       synthesisTarget: "audio-file",
       audioFileFormats: ["mp3", "caf", "audio/mpeg", "audio/x-caf"],
+      preferAudioFileFormat: "caf",
     });
     expect(resolveChannelTtsVoiceDelivery("discord")).toEqual({
       synthesisTarget: "voice-note",