@cartesia/cartesia-js 1.0.2 → 1.0.3

package/.turbo/turbo-build.log

@@ -5,32 +5,19 @@ $ tsup src/ --format cjs,esm --dts
 CLI Target: es6
 CJS Build start
 ESM Build start
-CJS dist/types/index.cjs   764.00 B
-CJS dist/voices/index.cjs  5.36 KB
-CJS dist/index.cjs         24.67 KB
-CJS dist/lib/client.cjs    4.07 KB
-CJS dist/lib/constants.cjs 1.43 KB
-CJS dist/lib/index.cjs     20.57 KB
-CJS dist/react/index.cjs   32.23 KB
-CJS dist/react/utils.cjs   1.80 KB
-CJS dist/tts/index.cjs     18.92 KB
-CJS dist/tts/player.cjs    6.96 KB
-CJS dist/tts/source.cjs    8.15 KB
-CJS dist/tts/utils.cjs     4.33 KB
-CJS dist/tts/websocket.cjs 18.61 KB
-CJS ⚡️ Build success in 75ms
-ESM dist/index.js          499.00 B
-ESM dist/lib/constants.js  183.00 B
 ESM dist/chunk-FXPGR372.js 0 B
+ESM dist/index.js          499.00 B
+ESM dist/lib/client.js     132.00 B
 ESM dist/react/index.js    7.62 KB
 ESM dist/tts/index.js      261.00 B
+ESM dist/lib/constants.js  183.00 B
 ESM dist/lib/index.js      322.00 B
-ESM dist/chunk-NVOCUUOF.js 353.00 B
-ESM dist/chunk-UCYL2SOX.js 1.48 KB
+ESM dist/chunk-ZF6HASZT.js 353.00 B
+ESM dist/chunk-YFN6TFR4.js 1.74 KB
 ESM dist/react/utils.js    109.00 B
 ESM dist/chunk-NQVZNVOU.js 337.00 B
-ESM dist/chunk-EDAAHENY.js 439.00 B
-ESM dist/chunk-IZBPLCGW.js 7.13 KB
+ESM dist/chunk-NWCW6C7H.js 439.00 B
+ESM dist/chunk-EUW2435M.js 7.17 KB
 ESM dist/chunk-PISCPZK4.js 1.02 KB
 ESM dist/chunk-2BFEKY3F.js 366.00 B
 ESM dist/tts/player.js     174.00 B
@@ -38,16 +25,29 @@ $ tsup src/ --format cjs,esm --dts
 ESM dist/chunk-BHY7MNGT.js 2.11 KB
 ESM dist/chunk-6YQ6KDIQ.js 5.32 KB
 ESM dist/chunk-GHY2WEOK.js 2.70 KB
+ESM dist/voices/index.js   174.00 B
+ESM dist/types/index.js    31.00 B
 ESM dist/tts/source.js     144.00 B
 ESM dist/tts/utils.js      426.00 B
 ESM dist/tts/websocket.js  242.00 B
-ESM dist/types/index.js    31.00 B
-ESM dist/voices/index.js   174.00 B
-ESM dist/lib/client.js     132.00 B
-ESM ⚡️ Build success in 77ms
+ESM ⚡️ Build success in 72ms
+CJS dist/index.cjs         24.98 KB
+CJS dist/lib/client.cjs    4.07 KB
+CJS dist/lib/constants.cjs 1.43 KB
+CJS dist/lib/index.cjs     20.87 KB
+CJS dist/react/index.cjs   32.54 KB
+CJS dist/react/utils.cjs   1.80 KB
+CJS dist/tts/index.cjs     18.96 KB
+CJS dist/tts/player.cjs    6.96 KB
+CJS dist/tts/utils.cjs     4.33 KB
+CJS dist/tts/source.cjs    8.15 KB
+CJS dist/types/index.cjs   764.00 B
+CJS dist/tts/websocket.cjs 18.65 KB
+CJS dist/voices/index.cjs  5.62 KB
+CJS ⚡️ Build success in 73ms
 DTS Build start
-DTS ⚡️ Build success in 8516ms
-DTS dist/index.d.cts         817.00 B
+DTS ⚡️ Build success in 8468ms
+DTS dist/index.d.cts         866.00 B
 DTS dist/lib/constants.d.cts 564.00 B
 DTS dist/lib/index.d.cts     410.00 B
 DTS dist/react/index.d.cts   1.09 KB
@@ -57,10 +57,10 @@ $ tsup src/ --format cjs,esm --dts
 DTS dist/tts/utils.d.cts     2.68 KB
 DTS dist/tts/websocket.d.cts 2.94 KB
 DTS dist/tts/source.d.cts    2.55 KB
-DTS dist/voices/index.d.cts  472.00 B
+DTS dist/voices/index.d.cts  573.00 B
 DTS dist/lib/client.d.cts    293.00 B
-DTS dist/types/index.d.cts   3.25 KB
-DTS dist/index.d.ts          809.00 B
+DTS dist/types/index.d.cts   3.50 KB
+DTS dist/index.d.ts          858.00 B
 DTS dist/lib/constants.d.ts  564.00 B
 DTS dist/lib/index.d.ts      404.00 B
 DTS dist/react/index.d.ts    1.09 KB
@@ -70,6 +70,6 @@ $ tsup src/ --format cjs,esm --dts
 DTS dist/tts/utils.d.ts      2.68 KB
 DTS dist/tts/websocket.d.ts  2.94 KB
 DTS dist/tts/source.d.ts     2.55 KB
-DTS dist/voices/index.d.ts   470.00 B
+DTS dist/voices/index.d.ts   571.00 B
 DTS dist/lib/client.d.ts     292.00 B
-DTS dist/types/index.d.ts    3.25 KB
+DTS dist/types/index.d.ts    3.50 KB

package/CHANGELOG.md

@@ -1,5 +1,11 @@
 # @cartesia/cartesia-js
 
+## 1.0.3
+
+### Patch Changes
+
+- 9b157d6: Support Bun by adding socket binaryType
+
 ## 1.0.2
 
 ### Patch Changes

package/README.md

@@ -52,6 +52,17 @@ console.log(voices);
 const voice = await cartesia.voices.get("<voice-id>");
 console.log(voice);
 
+// Clone a voice from a file.
+const clonedVoiceEmbedding = await cartesia.voices.clone({
+	mode: "clip",
+	clip: myFile, // Pass a File object or a Blob.
+});
+
+// Mix voices together.
+const mixedVoiceEmbedding = await cartesia.voices.mix({
+	voices: [{ id: "<voice-id-1>", weight: 0.6 }, { id: "<voice-id-2>", weight: 0.4 }],
+});
+
 // Create a voice.
 const newVoice = await cartesia.voices.create({
 	name: "Tim",
@@ -59,12 +70,6 @@ const newVoice = await cartesia.voices.create({
 	embedding: Array(192).fill(1.0),
 });
 console.log(newVoice);
-
-// Clone a voice from a file.
-const clonedVoice = await cartesia.voices.clone({
-	mode: "clip",
-	clip: myFile, // Pass a File object or a Blob.
-});
 ```
 
 ### TTS over WebSocket

package/dist/{chunk-IZBPLCGW.js → chunk-EUW2435M.js}

@@ -169,6 +169,7 @@ var WebSocket = class extends Client {
         url.searchParams.set("cartesia_version", CARTESIA_VERSION);
         return url.toString();
       }));
+      this.socket.binaryType = "arraybuffer";
       this.socket.onopen = () => {
         __privateSet(this, _isConnected, true);
         emitter.emit("open");

package/dist/{chunk-EDAAHENY.js → chunk-NWCW6C7H.js}

@@ -1,6 +1,6 @@
 import {
   WebSocket
-} from "./chunk-IZBPLCGW.js";
+} from "./chunk-EUW2435M.js";
 import {
   Client
 } from "./chunk-PISCPZK4.js";

package/dist/{chunk-UCYL2SOX.js → chunk-YFN6TFR4.js}

@@ -54,6 +54,16 @@ var Voices = class extends Client {
       throw new Error("Invalid mode for clone()");
     });
   }
+  mix(options) {
+    return __async(this, null, function* () {
+      const request = options;
+      const response = yield this._fetch("/voices/mix", {
+        method: "POST",
+        body: JSON.stringify(request)
+      });
+      return response.json();
+    });
+  }
 };
 
 export {

package/dist/{chunk-NVOCUUOF.js → chunk-ZF6HASZT.js}

@@ -1,9 +1,9 @@
 import {
   Voices
-} from "./chunk-UCYL2SOX.js";
+} from "./chunk-YFN6TFR4.js";
 import {
   TTS
-} from "./chunk-EDAAHENY.js";
+} from "./chunk-NWCW6C7H.js";
 import {
   Client
 } from "./chunk-PISCPZK4.js";

package/dist/index.cjs

@@ -510,6 +510,7 @@ var WebSocket = class extends Client {
         url.searchParams.set("cartesia_version", CARTESIA_VERSION);
         return url.toString();
       }));
+      this.socket.binaryType = "arraybuffer";
       this.socket.onopen = () => {
         __privateSet(this, _isConnected, true);
         emitter.emit("open");
@@ -639,6 +640,16 @@ var Voices = class extends Client {
       throw new Error("Invalid mode for clone()");
     });
   }
+  mix(options) {
+    return __async(this, null, function* () {
+      const request = options;
+      const response = yield this._fetch("/voices/mix", {
+        method: "POST",
+        body: JSON.stringify(request)
+      });
+      return response.json();
+    });
+  }
 };
 
 // src/lib/index.ts

package/dist/index.d.cts

@@ -1,5 +1,5 @@
 export { Cartesia, Cartesia as default } from './lib/index.cjs';
-export { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps } from './types/index.cjs';
+export { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, MixVoicesOptions, MixVoicesResponse, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, VoiceToMix, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps } from './types/index.cjs';
 export { default as WebPlayer } from './tts/player.cjs';
 export { default as Source } from './tts/source.cjs';
 export { default as WebSocket } from './tts/websocket.cjs';

package/dist/index.d.ts

@@ -1,5 +1,5 @@
 export { Cartesia, Cartesia as default } from './lib/index.js';
-export { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps } from './types/index.js';
+export { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, MixVoicesOptions, MixVoicesResponse, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, VoiceToMix, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps } from './types/index.js';
 export { default as WebPlayer } from './tts/player.js';
 export { default as Source } from './tts/source.js';
 export { default as WebSocket } from './tts/websocket.js';

package/dist/index.js

@@ -1,12 +1,12 @@
 import "./chunk-FXPGR372.js";
 import {
   Cartesia
-} from "./chunk-NVOCUUOF.js";
-import "./chunk-UCYL2SOX.js";
-import "./chunk-EDAAHENY.js";
+} from "./chunk-ZF6HASZT.js";
+import "./chunk-YFN6TFR4.js";
+import "./chunk-NWCW6C7H.js";
 import {
   WebSocket
-} from "./chunk-IZBPLCGW.js";
+} from "./chunk-EUW2435M.js";
 import "./chunk-PISCPZK4.js";
 import "./chunk-2BFEKY3F.js";
 import {

package/dist/lib/index.cjs

@@ -493,6 +493,7 @@ var WebSocket = class extends Client {
         url.searchParams.set("cartesia_version", CARTESIA_VERSION);
         return url.toString();
       }));
+      this.socket.binaryType = "arraybuffer";
       this.socket.onopen = () => {
         __privateSet(this, _isConnected, true);
         emitter.emit("open");
@@ -622,6 +623,16 @@ var Voices = class extends Client {
       throw new Error("Invalid mode for clone()");
     });
   }
+  mix(options) {
+    return __async(this, null, function* () {
+      const request = options;
+      const response = yield this._fetch("/voices/mix", {
+        method: "POST",
+        body: JSON.stringify(request)
+      });
+      return response.json();
+    });
+  }
 };
 
 // src/lib/index.ts

package/dist/lib/index.js

@@ -1,9 +1,9 @@
 import {
   Cartesia
-} from "../chunk-NVOCUUOF.js";
-import "../chunk-UCYL2SOX.js";
-import "../chunk-EDAAHENY.js";
-import "../chunk-IZBPLCGW.js";
+} from "../chunk-ZF6HASZT.js";
+import "../chunk-YFN6TFR4.js";
+import "../chunk-NWCW6C7H.js";
+import "../chunk-EUW2435M.js";
 import "../chunk-PISCPZK4.js";
 import "../chunk-2BFEKY3F.js";
 import "../chunk-BHY7MNGT.js";

package/dist/react/index.cjs

@@ -507,6 +507,7 @@ var WebSocket = class extends Client {
         url.searchParams.set("cartesia_version", CARTESIA_VERSION);
         return url.toString();
       }));
+      this.socket.binaryType = "arraybuffer";
       this.socket.onopen = () => {
         __privateSet(this, _isConnected, true);
         emitter.emit("open");
@@ -636,6 +637,16 @@ var Voices = class extends Client {
       throw new Error("Invalid mode for clone()");
     });
   }
+  mix(options) {
+    return __async(this, null, function* () {
+      const request = options;
+      const response = yield this._fetch("/voices/mix", {
+        method: "POST",
+        body: JSON.stringify(request)
+      });
+      return response.json();
+    });
+  }
 };
 
 // src/lib/index.ts

package/dist/react/index.js

@@ -1,12 +1,12 @@
 import {
   Cartesia
-} from "../chunk-NVOCUUOF.js";
-import "../chunk-UCYL2SOX.js";
+} from "../chunk-ZF6HASZT.js";
+import "../chunk-YFN6TFR4.js";
 import {
   pingServer
 } from "../chunk-NQVZNVOU.js";
-import "../chunk-EDAAHENY.js";
-import "../chunk-IZBPLCGW.js";
+import "../chunk-NWCW6C7H.js";
+import "../chunk-EUW2435M.js";
 import "../chunk-PISCPZK4.js";
 import "../chunk-2BFEKY3F.js";
 import {

package/dist/tts/index.cjs

@@ -493,6 +493,7 @@ var WebSocket = class extends Client {
         url.searchParams.set("cartesia_version", CARTESIA_VERSION);
         return url.toString();
       }));
+      this.socket.binaryType = "arraybuffer";
       this.socket.onopen = () => {
         __privateSet(this, _isConnected, true);
         emitter.emit("open");

package/dist/tts/index.js

@@ -1,7 +1,7 @@
 import {
   TTS
-} from "../chunk-EDAAHENY.js";
-import "../chunk-IZBPLCGW.js";
+} from "../chunk-NWCW6C7H.js";
+import "../chunk-EUW2435M.js";
 import "../chunk-PISCPZK4.js";
 import "../chunk-2BFEKY3F.js";
 import "../chunk-BHY7MNGT.js";

package/dist/tts/websocket.cjs

@@ -491,6 +491,7 @@ var WebSocket = class extends Client {
         url.searchParams.set("cartesia_version", CARTESIA_VERSION);
         return url.toString();
       }));
+      this.socket.binaryType = "arraybuffer";
       this.socket.onopen = () => {
         __privateSet(this, _isConnected, true);
         emitter.emit("open");

package/dist/tts/websocket.js

@@ -1,6 +1,6 @@
 import {
   WebSocket
-} from "../chunk-IZBPLCGW.js";
+} from "../chunk-EUW2435M.js";
 import "../chunk-PISCPZK4.js";
 import "../chunk-2BFEKY3F.js";
 import "../chunk-BHY7MNGT.js";

package/dist/types/index.d.cts

@@ -11,10 +11,10 @@ type ConnectionEventData = {
     close: never;
 };
 type VoiceSpecifier = {
-    mode: "id";
+    mode?: "id";
     id: string;
 } | {
-    mode: "embedding";
+    mode?: "embedding";
     embedding: number[];
 };
 type Emotion = "anger" | "sadness" | "positivity" | "curiosity" | "surprise";
@@ -22,7 +22,7 @@ type Intensity = "lowest" | "low" | "high" | "highest";
 type EmotionControl = Emotion | `${Emotion}:${Intensity}`;
 type VoiceOptions = VoiceSpecifier & {
     __experimental_controls?: {
-        speed?: "slowest" | "slow" | "normal" | "fast" | "fastest";
+        speed?: "slowest" | "slow" | "normal" | "fast" | "fastest" | number;
         emotion?: EmotionControl[];
     };
 };
@@ -83,6 +83,14 @@ type CloneOptions = {
     clip: Blob;
     enhance?: boolean;
 };
+interface VoiceToMix {
+    id?: string;
+    embedding?: number[];
+    weight: number;
+}
+interface MixVoicesOptions {
+    voices: VoiceToMix[];
+}
 type Voice = {
     id: string;
     name: string;
@@ -98,6 +106,9 @@ type UpdateVoice = Partial<Pick<Voice, "name" | "description" | "embedding">>;
 type CloneResponse = {
     embedding: number[];
 };
+type MixVoicesResponse = {
+    embedding: number[];
+};
 type WebSocketOptions = {
     container?: string;
     encoding?: string;
@@ -112,4 +123,4 @@ type SourceEventData = {
 type TypedArray = Float32Array | Int16Array | Uint8Array;
 type Encoding = "pcm_f32le" | "pcm_s16le" | "pcm_alaw" | "pcm_mulaw";
 
-export type { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps };
+export type { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, MixVoicesOptions, MixVoicesResponse, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, VoiceToMix, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps };

package/dist/types/index.d.ts

@@ -11,10 +11,10 @@ type ConnectionEventData = {
     close: never;
 };
 type VoiceSpecifier = {
-    mode: "id";
+    mode?: "id";
     id: string;
 } | {
-    mode: "embedding";
+    mode?: "embedding";
     embedding: number[];
 };
 type Emotion = "anger" | "sadness" | "positivity" | "curiosity" | "surprise";
@@ -22,7 +22,7 @@ type Intensity = "lowest" | "low" | "high" | "highest";
 type EmotionControl = Emotion | `${Emotion}:${Intensity}`;
 type VoiceOptions = VoiceSpecifier & {
     __experimental_controls?: {
-        speed?: "slowest" | "slow" | "normal" | "fast" | "fastest";
+        speed?: "slowest" | "slow" | "normal" | "fast" | "fastest" | number;
         emotion?: EmotionControl[];
     };
 };
@@ -83,6 +83,14 @@ type CloneOptions = {
     clip: Blob;
     enhance?: boolean;
 };
+interface VoiceToMix {
+    id?: string;
+    embedding?: number[];
+    weight: number;
+}
+interface MixVoicesOptions {
+    voices: VoiceToMix[];
+}
 type Voice = {
     id: string;
     name: string;
@@ -98,6 +106,9 @@ type UpdateVoice = Partial<Pick<Voice, "name" | "description" | "embedding">>;
 type CloneResponse = {
     embedding: number[];
 };
+type MixVoicesResponse = {
+    embedding: number[];
+};
 type WebSocketOptions = {
     container?: string;
     encoding?: string;
@@ -112,4 +123,4 @@ type SourceEventData = {
 type TypedArray = Float32Array | Int16Array | Uint8Array;
 type Encoding = "pcm_f32le" | "pcm_s16le" | "pcm_alaw" | "pcm_mulaw";
 
-export type { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps };
+export type { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, MixVoicesOptions, MixVoicesResponse, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, VoiceToMix, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps };

package/dist/voices/index.cjs

@@ -159,4 +159,14 @@ var Voices = class extends Client {
       throw new Error("Invalid mode for clone()");
     });
   }
+  mix(options) {
+    return __async(this, null, function* () {
+      const request = options;
+      const response = yield this._fetch("/voices/mix", {
+        method: "POST",
+        body: JSON.stringify(request)
+      });
+      return response.json();
+    });
+  }
 };

package/dist/voices/index.d.cts

@@ -1,5 +1,5 @@
 import { Client } from '../lib/client.cjs';
-import { Voice, CreateVoice, UpdateVoice, CloneOptions, CloneResponse } from '../types/index.cjs';
+import { Voice, CreateVoice, UpdateVoice, CloneOptions, CloneResponse, MixVoicesOptions, MixVoicesResponse } from '../types/index.cjs';
 import 'emittery';
 
 declare class Voices extends Client {
@@ -8,6 +8,7 @@ declare class Voices extends Client {
     create(voice: CreateVoice): Promise<Voice>;
     update(id: string, voice: UpdateVoice): Promise<Voice>;
     clone(options: CloneOptions): Promise<CloneResponse>;
+    mix(options: MixVoicesOptions): Promise<MixVoicesResponse>;
 }
 
 export { Voices as default };

package/dist/voices/index.d.ts

@@ -1,5 +1,5 @@
 import { Client } from '../lib/client.js';
-import { Voice, CreateVoice, UpdateVoice, CloneOptions, CloneResponse } from '../types/index.js';
+import { Voice, CreateVoice, UpdateVoice, CloneOptions, CloneResponse, MixVoicesOptions, MixVoicesResponse } from '../types/index.js';
 import 'emittery';
 
 declare class Voices extends Client {
@@ -8,6 +8,7 @@ declare class Voices extends Client {
     create(voice: CreateVoice): Promise<Voice>;
     update(id: string, voice: UpdateVoice): Promise<Voice>;
     clone(options: CloneOptions): Promise<CloneResponse>;
+    mix(options: MixVoicesOptions): Promise<MixVoicesResponse>;
 }
 
 export { Voices as default };

package/dist/voices/index.js

@@ -1,6 +1,6 @@
 import {
   Voices
-} from "../chunk-UCYL2SOX.js";
+} from "../chunk-YFN6TFR4.js";
 import "../chunk-PISCPZK4.js";
 import "../chunk-2BFEKY3F.js";
 import "../chunk-GHY2WEOK.js";

package/package.json

@@ -4,7 +4,7 @@
 		"name": "Cartesia",
 		"url": "https://cartesia.ai"
 	},
-	"version": "1.0.2",
+	"version": "1.0.3",
 	"description": "Client for the Cartesia API.",
 	"type": "module",
 	"module": "./dist/index.js",

package/src/tts/websocket.ts

@@ -182,6 +182,8 @@ export default class WebSocket extends Client {
 			url.searchParams.set("cartesia_version", CARTESIA_VERSION);
 			return url.toString();
 		});
+		this.socket.binaryType = "arraybuffer";
+
 		this.socket.onopen = () => {
 			this.#isConnected = true;
 			emitter.emit("open");

package/src/types/index.ts

@@ -16,11 +16,11 @@ export type ConnectionEventData = {
 
 export type VoiceSpecifier =
 	| {
-			mode: "id";
+			mode?: "id";
 			id: string;
 	  }
 	| {
-			mode: "embedding";
+			mode?: "embedding";
 			embedding: number[];
 	  };
 
@@ -35,7 +35,7 @@ export type EmotionControl = Emotion | `${Emotion}:${Intensity}`;
 
 export type VoiceOptions = VoiceSpecifier & {
 	__experimental_controls?: {
-		speed?: "slowest" | "slow" | "normal" | "fast" | "fastest";
+		speed?: "slowest" | "slow" | "normal" | "fast" | "fastest" | number;
 		emotion?: EmotionControl[];
 	};
 };
@@ -112,6 +112,16 @@ export type CloneOptions =
 			enhance?: boolean;
 	  };
 
+export interface VoiceToMix {
+	id?: string;
+	embedding?: number[];
+	weight: number;
+}
+
+export interface MixVoicesOptions {
+	voices: VoiceToMix[];
+}
+
 export type Voice = {
 	id: string;
 	name: string;
@@ -134,6 +144,10 @@ export type CloneResponse = {
 	embedding: number[];
 };
 
+export type MixVoicesResponse = {
+	embedding: number[];
+};
+
 export type WebSocketOptions = {
 	container?: string;
 	encoding?: string;

package/src/voices/index.ts

@@ -3,6 +3,8 @@ import type {
 	CloneOptions,
 	CloneResponse,
 	CreateVoice,
+	MixVoicesOptions,
+	MixVoicesResponse,
 	UpdateVoice,
 	Voice,
 } from "../types";
@@ -51,4 +53,15 @@ export default class Voices extends Client {
 
 		throw new Error("Invalid mode for clone()");
 	}
+
+	async mix(options: MixVoicesOptions): Promise<MixVoicesResponse> {
+		const request: MixVoicesOptions = options;
+
+		const response = await this._fetch("/voices/mix", {
+			method: "POST",
+			body: JSON.stringify(request),
+		});
+
+		return response.json() as Promise<MixVoicesResponse>;
+	}
 }