@cartesia/cartesia-js 1.0.3 → 1.1.0

package/dist/tts/websocket.cjs

@@ -22,18 +22,6 @@ var __spreadValues = (a, b) => {
   return a;
 };
 var __spreadProps = (a, b) => __defProps(a, __getOwnPropDescs(b));
-var __objRest = (source, exclude) => {
-  var target = {};
-  for (var prop in source)
-    if (__hasOwnProp.call(source, prop) && exclude.indexOf(prop) < 0)
-      target[prop] = source[prop];
-  if (source != null && __getOwnPropSymbols)
-    for (var prop of __getOwnPropSymbols(source)) {
-      if (exclude.indexOf(prop) < 0 && __propIsEnum.call(source, prop))
-        target[prop] = source[prop];
-    }
-  return target;
-};
 var __export = (target, all) => {
   for (var name in all)
     __defProp(target, name, { get: all[name], enumerable: true });
@@ -247,6 +235,37 @@ var Source = class {
       return read;
     });
   }
+  /**
+   * Seek in the buffer.
+   *
+   * @param offset The offset to seek to.
+   * @param whence The position to seek from.
+   * @returns The new position in the buffer.
+   * @throws {Error} If the seek is invalid.
+   */
+  seek(offset, whence) {
+    return __async(this, null, function* () {
+      let position = __privateGet(this, _readIndex);
+      switch (whence) {
+        case "start":
+          position = offset;
+          break;
+        case "current":
+          position += offset;
+          break;
+        case "end":
+          position = __privateGet(this, _writeIndex) + offset;
+          break;
+        default:
+          throw new Error(`Invalid seek mode: ${whence}`);
+      }
+      if (position < 0 || position > __privateGet(this, _writeIndex)) {
+        throw new Error("Seek out of bounds");
+      }
+      __privateSet(this, _readIndex, position);
+      return position;
+    });
+  }
   /**
    * Get the number of samples in a given duration.
    *
@@ -380,7 +399,7 @@ var WebSocket = class extends Client {
   /**
    * Send a message over the WebSocket to start a stream.
    *
-   * @param inputs - Stream options. Defined in the StreamRequest type.
+   * @param inputs - Generation parameters. Defined in the StreamRequest type.
    * @param options - Options for the stream.
    * @param options.timeout - The maximum time to wait for a chunk before cancelling the stream.
    *                          If set to `0`, the stream will not time out.
@@ -388,9 +407,8 @@ var WebSocket = class extends Client {
    * @returns An Emittery instance that emits messages from the WebSocket.
    * @returns An abort function that can be called to cancel the stream.
    */
-  send(_a, { timeout = 0 } = {}) {
-    var inputs = __objRest(_a, []);
-    var _a2, _b, _c, _d;
+  send(inputs, { timeout = 0 } = {}) {
+    var _a, _b, _c, _d;
     if (!__privateGet(this, _isConnected)) {
       throw new Error("Not connected to WebSocket. Call .connect() first.");
     }
@@ -404,7 +422,7 @@ var WebSocket = class extends Client {
         sample_rate: __privateGet(this, _sampleRate2)
       };
     }
-    (_a2 = this.socket) == null ? void 0 : _a2.send(
+    (_a = this.socket) == null ? void 0 : _a.send(
       JSON.stringify(__spreadValues({}, inputs))
     );
     const emitter = new import_emittery2.default();
@@ -474,6 +492,32 @@ var WebSocket = class extends Client {
       stop: streamCompleteController.abort.bind(streamCompleteController)
     });
   }
+  /**
+   * Continue a stream.
+   *
+   * @param inputs - Generation parameters. Defined in the StreamRequest type, but must include a `context_id` field. `continue` is set to true by default.
+   */
+  continue(inputs) {
+    var _a;
+    if (!__privateGet(this, _isConnected)) {
+      throw new Error("Not connected to WebSocket. Call .connect() first.");
+    }
+    if (!inputs.context_id) {
+      throw new Error("context_id is required to continue a context.");
+    }
+    if (!inputs.output_format) {
+      inputs.output_format = {
+        container: __privateGet(this, _container2),
+        encoding: __privateGet(this, _encoding2),
+        sample_rate: __privateGet(this, _sampleRate2)
+      };
+    }
+    (_a = this.socket) == null ? void 0 : _a.send(
+      JSON.stringify(__spreadValues({
+        continue: true
+      }, inputs))
+    );
+  }
   /**
    * Authenticate and connect to a Cartesia streaming WebSocket.
    *

package/dist/tts/websocket.d.cts

@@ -1,7 +1,7 @@
 import * as emittery from 'emittery';
 import { WebSocket as WebSocket$1 } from 'partysocket';
 import { Client } from '../lib/client.cjs';
-import { WebSocketOptions, StreamRequest, StreamOptions, WordTimestamps, EmitteryCallbacks, ConnectionEventData } from '../types/index.cjs';
+import { WebSocketOptions, StreamRequest, StreamOptions, WordTimestamps, ContinueRequest, EmitteryCallbacks, ConnectionEventData } from '../types/index.cjs';
 import Source from './source.cjs';
 
 declare class WebSocket extends Client {
@@ -16,7 +16,7 @@ declare class WebSocket extends Client {
     /**
      * Send a message over the WebSocket to start a stream.
      *
-     * @param inputs - Stream options. Defined in the StreamRequest type.
+     * @param inputs - Generation parameters. Defined in the StreamRequest type.
      * @param options - Options for the stream.
      * @param options.timeout - The maximum time to wait for a chunk before cancelling the stream.
      *                          If set to `0`, the stream will not time out.
@@ -24,7 +24,7 @@ declare class WebSocket extends Client {
      * @returns An Emittery instance that emits messages from the WebSocket.
      * @returns An abort function that can be called to cancel the stream.
      */
-    send({ ...inputs }: StreamRequest, { timeout }?: StreamOptions): {
+    send(inputs: StreamRequest, { timeout }?: StreamOptions): {
         stop: {
             (reason?: any): void;
             (reason?: any): void;
@@ -47,6 +47,12 @@ declare class WebSocket extends Client {
         }[Name_3]>;
         source: Source;
     };
+    /**
+     * Continue a stream.
+     *
+     * @param inputs - Generation parameters. Defined in the StreamRequest type, but must include a `context_id` field. `continue` is set to true by default.
+     */
+    continue(inputs: ContinueRequest): void;
     /**
      * Authenticate and connect to a Cartesia streaming WebSocket.
      *

package/dist/tts/websocket.d.ts

@@ -1,7 +1,7 @@
 import * as emittery from 'emittery';
 import { WebSocket as WebSocket$1 } from 'partysocket';
 import { Client } from '../lib/client.js';
-import { WebSocketOptions, StreamRequest, StreamOptions, WordTimestamps, EmitteryCallbacks, ConnectionEventData } from '../types/index.js';
+import { WebSocketOptions, StreamRequest, StreamOptions, WordTimestamps, ContinueRequest, EmitteryCallbacks, ConnectionEventData } from '../types/index.js';
 import Source from './source.js';
 
 declare class WebSocket extends Client {
@@ -16,7 +16,7 @@ declare class WebSocket extends Client {
     /**
      * Send a message over the WebSocket to start a stream.
      *
-     * @param inputs - Stream options. Defined in the StreamRequest type.
+     * @param inputs - Generation parameters. Defined in the StreamRequest type.
      * @param options - Options for the stream.
      * @param options.timeout - The maximum time to wait for a chunk before cancelling the stream.
      *                          If set to `0`, the stream will not time out.
@@ -24,7 +24,7 @@ declare class WebSocket extends Client {
      * @returns An Emittery instance that emits messages from the WebSocket.
      * @returns An abort function that can be called to cancel the stream.
      */
-    send({ ...inputs }: StreamRequest, { timeout }?: StreamOptions): {
+    send(inputs: StreamRequest, { timeout }?: StreamOptions): {
         stop: {
             (reason?: any): void;
             (reason?: any): void;
@@ -47,6 +47,12 @@ declare class WebSocket extends Client {
         }[Name_3]>;
         source: Source;
     };
+    /**
+     * Continue a stream.
+     *
+     * @param inputs - Generation parameters. Defined in the StreamRequest type, but must include a `context_id` field. `continue` is set to true by default.
+     */
+    continue(inputs: ContinueRequest): void;
     /**
      * Authenticate and connect to a Cartesia streaming WebSocket.
      *

package/dist/tts/websocket.js

@@ -1,11 +1,11 @@
 import {
   WebSocket
-} from "../chunk-EUW2435M.js";
-import "../chunk-PISCPZK4.js";
+} from "../chunk-L3OMQKWL.js";
+import "../chunk-6PWLZAMS.js";
 import "../chunk-2BFEKY3F.js";
-import "../chunk-BHY7MNGT.js";
-import "../chunk-6YQ6KDIQ.js";
-import "../chunk-GHY2WEOK.js";
+import "../chunk-VCZESWYA.js";
+import "../chunk-VVDJR3OA.js";
+import "../chunk-WIFMLPT5.js";
 export {
   WebSocket as default
 };

package/dist/types/index.d.cts

@@ -38,9 +38,13 @@ type StreamRequest = {
     context_id?: string;
     continue?: boolean;
     duration?: number;
-    language?: string;
+    language?: Language;
     add_timestamps?: boolean;
 };
+type ContinueRequest = StreamRequest & {
+    context_id: string;
+};
+type Language = "en" | "es" | "fr" | "de" | "ja" | "zh" | "pt" | (string & {});
 type StreamOptions = {
     timeout?: number;
 };
@@ -83,6 +87,14 @@ type CloneOptions = {
     clip: Blob;
     enhance?: boolean;
 };
+type LocalizeOptions = {
+    mode: "embedding";
+    embedding: number[];
+} & {
+    language: Language;
+    dialect: string & {};
+    original_speaker_gender: "male" | "female" | (string & {});
+};
 interface VoiceToMix {
     id?: string;
     embedding?: number[];
@@ -99,13 +111,16 @@ type Voice = {
     is_public: boolean;
     user_id: string;
     created_at: string;
-    language: string;
+    language: Language;
 };
 type CreateVoice = Pick<Voice, "name" | "description" | "embedding"> & Partial<Omit<Voice, "name" | "description" | "embedding">>;
 type UpdateVoice = Partial<Pick<Voice, "name" | "description" | "embedding">>;
 type CloneResponse = {
     embedding: number[];
 };
+type LocalizeResponse = {
+    embedding: number[];
+};
 type MixVoicesResponse = {
     embedding: number[];
 };
@@ -123,4 +138,4 @@ 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, MixVoicesOptions, MixVoicesResponse, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, VoiceToMix, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps };
+export type { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, ContinueRequest, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, Language, LocalizeOptions, LocalizeResponse, 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

@@ -38,9 +38,13 @@ type StreamRequest = {
     context_id?: string;
     continue?: boolean;
     duration?: number;
-    language?: string;
+    language?: Language;
     add_timestamps?: boolean;
 };
+type ContinueRequest = StreamRequest & {
+    context_id: string;
+};
+type Language = "en" | "es" | "fr" | "de" | "ja" | "zh" | "pt" | (string & {});
 type StreamOptions = {
     timeout?: number;
 };
@@ -83,6 +87,14 @@ type CloneOptions = {
     clip: Blob;
     enhance?: boolean;
 };
+type LocalizeOptions = {
+    mode: "embedding";
+    embedding: number[];
+} & {
+    language: Language;
+    dialect: string & {};
+    original_speaker_gender: "male" | "female" | (string & {});
+};
 interface VoiceToMix {
     id?: string;
     embedding?: number[];
@@ -99,13 +111,16 @@ type Voice = {
     is_public: boolean;
     user_id: string;
     created_at: string;
-    language: string;
+    language: Language;
 };
 type CreateVoice = Pick<Voice, "name" | "description" | "embedding"> & Partial<Omit<Voice, "name" | "description" | "embedding">>;
 type UpdateVoice = Partial<Pick<Voice, "name" | "description" | "embedding">>;
 type CloneResponse = {
     embedding: number[];
 };
+type LocalizeResponse = {
+    embedding: number[];
+};
 type MixVoicesResponse = {
     embedding: number[];
 };
@@ -123,4 +138,4 @@ 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, MixVoicesOptions, MixVoicesResponse, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, VoiceToMix, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps };
+export type { Chunk, ClientOptions, CloneOptions, CloneResponse, ConnectionEventData, ContinueRequest, CreateVoice, EmitteryCallbacks, Emotion, EmotionControl, Encoding, Intensity, Language, LocalizeOptions, LocalizeResponse, MixVoicesOptions, MixVoicesResponse, Sentinel, SourceEventData, StreamOptions, StreamRequest, TypedArray, UpdateVoice, Voice, VoiceOptions, VoiceSpecifier, VoiceToMix, WebSocketBaseResponse, WebSocketChunkResponse, WebSocketErrorResponse, WebSocketOptions, WebSocketResponse, WebSocketTimestampsResponse, WordTimestamps };

package/dist/voices/index.cjs

@@ -161,10 +161,18 @@ var Voices = class extends Client {
   }
   mix(options) {
     return __async(this, null, function* () {
-      const request = options;
       const response = yield this._fetch("/voices/mix", {
         method: "POST",
-        body: JSON.stringify(request)
+        body: JSON.stringify(options)
+      });
+      return response.json();
+    });
+  }
+  localize(options) {
+    return __async(this, null, function* () {
+      const response = yield this._fetch("/voices/localize", {
+        method: "POST",
+        body: JSON.stringify(options)
       });
       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, MixVoicesOptions, MixVoicesResponse } from '../types/index.cjs';
+import { Voice, CreateVoice, UpdateVoice, CloneOptions, CloneResponse, MixVoicesOptions, MixVoicesResponse, LocalizeOptions, LocalizeResponse } from '../types/index.cjs';
 import 'emittery';
 
 declare class Voices extends Client {
@@ -9,6 +9,7 @@ declare class Voices extends Client {
     update(id: string, voice: UpdateVoice): Promise<Voice>;
     clone(options: CloneOptions): Promise<CloneResponse>;
     mix(options: MixVoicesOptions): Promise<MixVoicesResponse>;
+    localize(options: LocalizeOptions): Promise<LocalizeResponse>;
 }
 
 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, MixVoicesOptions, MixVoicesResponse } from '../types/index.js';
+import { Voice, CreateVoice, UpdateVoice, CloneOptions, CloneResponse, MixVoicesOptions, MixVoicesResponse, LocalizeOptions, LocalizeResponse } from '../types/index.js';
 import 'emittery';
 
 declare class Voices extends Client {
@@ -9,6 +9,7 @@ declare class Voices extends Client {
     update(id: string, voice: UpdateVoice): Promise<Voice>;
     clone(options: CloneOptions): Promise<CloneResponse>;
     mix(options: MixVoicesOptions): Promise<MixVoicesResponse>;
+    localize(options: LocalizeOptions): Promise<LocalizeResponse>;
 }
 
 export { Voices as default };

package/dist/voices/index.js

@@ -1,9 +1,9 @@
 import {
   Voices
-} from "../chunk-YFN6TFR4.js";
-import "../chunk-PISCPZK4.js";
+} from "../chunk-CWIJUBG6.js";
+import "../chunk-6PWLZAMS.js";
 import "../chunk-2BFEKY3F.js";
-import "../chunk-GHY2WEOK.js";
+import "../chunk-WIFMLPT5.js";
 export {
   Voices as default
 };

package/package.json

@@ -4,7 +4,7 @@
 		"name": "Cartesia",
 		"url": "https://cartesia.ai"
 	},
-	"version": "1.0.3",
+	"version": "1.1.0",
 	"description": "Client for the Cartesia API.",
 	"type": "module",
 	"module": "./dist/index.js",

package/src/react/index.ts

@@ -201,6 +201,10 @@ export function useTTS({
 				await player.current.stop();
 			}
 
+			if (playbackStatus === "finished") {
+				websocketReturn.current.source.seek(0, "start");
+			}
+
 			setPlaybackStatus("playing");
 
 			const unsubscribes = [];

package/src/tts/source.ts

@@ -124,6 +124,41 @@ export default class Source {
 		return read;
 	}
 
+	/**
+	 * Seek in the buffer.
+	 *
+	 * @param offset The offset to seek to.
+	 * @param whence The position to seek from.
+	 * @returns The new position in the buffer.
+	 * @throws {Error} If the seek is invalid.
+	 */
+	async seek(
+		offset: number,
+		whence: "start" | "current" | "end",
+	): Promise<number> {
+		let position = this.#readIndex;
+		switch (whence) {
+			case "start":
+				position = offset;
+				break;
+			case "current":
+				position += offset;
+				break;
+			case "end":
+				position = this.#writeIndex + offset;
+				break;
+			default:
+				throw new Error(`Invalid seek mode: ${whence}`);
+		}
+
+		if (position < 0 || position > this.#writeIndex) {
+			throw new Error("Seek out of bounds");
+		}
+
+		this.#readIndex = position;
+		return position;
+	}
+
 	/**
 	 * Get the number of samples in a given duration.
 	 *

package/src/tts/websocket.ts

@@ -5,6 +5,7 @@ import { Client } from "../lib/client";
 import { CARTESIA_VERSION, constructApiUrl } from "../lib/constants";
 import type {
 	ConnectionEventData,
+	ContinueRequest,
 	EmitteryCallbacks,
 	StreamOptions,
 	StreamRequest,
@@ -45,7 +46,7 @@ export default class WebSocket extends Client {
 	/**
 	 * Send a message over the WebSocket to start a stream.
 	 *
-	 * @param inputs - Stream options. Defined in the StreamRequest type.
+	 * @param inputs - Generation parameters. Defined in the StreamRequest type.
 	 * @param options - Options for the stream.
 	 * @param options.timeout - The maximum time to wait for a chunk before cancelling the stream.
 	 *                          If set to `0`, the stream will not time out.
@@ -53,7 +54,7 @@ export default class WebSocket extends Client {
 	 * @returns An Emittery instance that emits messages from the WebSocket.
 	 * @returns An abort function that can be called to cancel the stream.
 	 */
-	send({ ...inputs }: StreamRequest, { timeout = 0 }: StreamOptions = {}) {
+	send(inputs: StreamRequest, { timeout = 0 }: StreamOptions = {}) {
 		if (!this.#isConnected) {
 			throw new Error("Not connected to WebSocket. Call .connect() first.");
 		}
@@ -151,6 +152,36 @@ export default class WebSocket extends Client {
 		};
 	}
 
+	/**
+	 * Continue a stream.
+	 *
+	 * @param inputs - Generation parameters. Defined in the StreamRequest type, but must include a `context_id` field. `continue` is set to true by default.
+	 */
+	continue(inputs: ContinueRequest) {
+		if (!this.#isConnected) {
+			throw new Error("Not connected to WebSocket. Call .connect() first.");
+		}
+
+		if (!inputs.context_id) {
+			throw new Error("context_id is required to continue a context.");
+		}
+		if (!inputs.output_format) {
+			inputs.output_format = {
+				container: this.#container,
+				encoding: this.#encoding,
+				sample_rate: this.#sampleRate,
+			};
+		}
+
+		// Send continue request.
+		this.socket?.send(
+			JSON.stringify({
+				continue: true,
+				...inputs,
+			}),
+		);
+	}
+
 	/**
 	 * Generate a unique ID suitable for a streaming context.
 	 *

package/src/types/index.ts

@@ -52,10 +52,24 @@ export type StreamRequest = {
 	context_id?: string;
 	continue?: boolean;
 	duration?: number;
-	language?: string;
+	language?: Language;
 	add_timestamps?: boolean;
 };
 
+export type ContinueRequest = StreamRequest & {
+	context_id: string;
+};
+
+export type Language =
+	| "en"
+	| "es"
+	| "fr"
+	| "de"
+	| "ja"
+	| "zh"
+	| "pt"
+	| (string & {});
+
 export type StreamOptions = {
 	timeout?: number;
 };
@@ -112,6 +126,15 @@ export type CloneOptions =
 			enhance?: boolean;
 	  };
 
+export type LocalizeOptions = {
+	mode: "embedding";
+	embedding: number[];
+} & {
+	language: Language;
+	dialect: string & {};
+	original_speaker_gender: "male" | "female" | (string & {});
+};
+
 export interface VoiceToMix {
 	id?: string;
 	embedding?: number[];
@@ -130,7 +153,7 @@ export type Voice = {
 	is_public: boolean;
 	user_id: string;
 	created_at: string;
-	language: string;
+	language: Language;
 };
 
 export type CreateVoice = Pick<Voice, "name" | "description" | "embedding"> &
@@ -144,6 +167,10 @@ export type CloneResponse = {
 	embedding: number[];
 };
 
+export type LocalizeResponse = {
+	embedding: number[];
+};
+
 export type MixVoicesResponse = {
 	embedding: number[];
 };

package/src/voices/index.ts

@@ -3,6 +3,8 @@ import type {
 	CloneOptions,
 	CloneResponse,
 	CreateVoice,
+	LocalizeOptions,
+	LocalizeResponse,
 	MixVoicesOptions,
 	MixVoicesResponse,
 	UpdateVoice,
@@ -55,13 +57,20 @@ export default class Voices extends Client {
 	}
 
 	async mix(options: MixVoicesOptions): Promise<MixVoicesResponse> {
-		const request: MixVoicesOptions = options;
-
 		const response = await this._fetch("/voices/mix", {
 			method: "POST",
-			body: JSON.stringify(request),
+			body: JSON.stringify(options),
 		});
 
 		return response.json() as Promise<MixVoicesResponse>;
 	}
+
+	async localize(options: LocalizeOptions): Promise<LocalizeResponse> {
+		const response = await this._fetch("/voices/localize", {
+			method: "POST",
+			body: JSON.stringify(options),
+		});
+
+		return response.json() as Promise<LocalizeResponse>;
+	}
 }