@cartesia/cartesia-js 1.1.0 → 1.2.1

package/package.json

@@ -4,7 +4,7 @@
 		"name": "Cartesia",
 		"url": "https://cartesia.ai"
 	},
-	"version": "1.1.0",
+	"version": "1.2.1",
 	"description": "Client for the Cartesia API.",
 	"type": "module",
 	"module": "./dist/index.js",
@@ -27,29 +27,22 @@
 		"cross-fetch": "^4.0.0",
 		"emittery": "^1.0.3",
 		"human-id": "^4.1.1",
-		"partysocket": "^1.0.1"
+		"partysocket": "^1.0.1",
+		"react": "^18.3.1"
 	},
 	"publishConfig": {
 		"access": "public"
 	},
 	"scripts": {
 		"build": "tsup src/ --format cjs,esm --dts",
-		"dev": "bun run build -- --watch"
-	},
-	"peerDependencies": {
-		"react": "^18.2.0",
-		"@types/react": "^18.2.58"
-	},
-	"peerDependenciesMeta": {
-		"react": {
-			"optional": true
-		},
-		"@types/react": {
-			"optional": true
-		}
+		"dev": "bun run build -- --watch",
+		"release": "rm -rf dist && bun run build && npm publish"
 	},
 	"devDependencies": {
-		"@repo/config-typescript": "workspace:*",
-		"tsup": "^8.0.2"
+		"@biomejs/biome": "^1.9.4",
+		"@types/node": "^22.7.9",
+		"@types/react": "^18.3.12",
+		"tsup": "^8.0.2",
+		"typescript": "^5.6.3"
 	}
 }

package/src/lib/index.ts

@@ -1,16 +1,19 @@
 import TTS from "../tts";
 import type { ClientOptions } from "../types";
+import VoiceChanger from "../voice-changer";
 import Voices from "../voices";
 import { Client } from "./client";
 
 export class Cartesia extends Client {
 	tts: TTS;
 	voices: Voices;
+	voiceChanger: VoiceChanger;
 
 	constructor(options: ClientOptions = {}) {
 		super(options);
 
 		this.tts = new TTS(options);
 		this.voices = new Voices(options);
+		this.voiceChanger = new VoiceChanger(options);
 	}
 }

package/src/tts/index.ts

@@ -1,5 +1,5 @@
 import { Client } from "../lib/client";
-import type { WebSocketOptions } from "../types";
+import type { BytesRequest, WebSocketOptions } from "../types";
 import WebSocket from "./websocket";
 
 export default class TTS extends Client {
@@ -14,4 +14,19 @@ export default class TTS extends Client {
 			baseUrl: this.baseUrl,
 		});
 	}
+
+	/**
+	 * Generate audio bytes from text.
+	 *
+	 * @param options - The options for the request.
+	 * @returns {Promise<ArrayBuffer>} A promise that resolves to an ArrayBuffer containing the audio bytes.
+	 */
+	async bytes(options: BytesRequest): Promise<ArrayBuffer> {
+		const response = await this._fetch("/tts/bytes", {
+			method: "POST",
+			body: JSON.stringify(options),
+		});
+
+		return response.arrayBuffer();
+	}
 }

package/src/types/index.ts

@@ -56,6 +56,11 @@ export type StreamRequest = {
 	add_timestamps?: boolean;
 };
 
+export type BytesRequest = Omit<
+	StreamRequest,
+	"continue" | "add_timestamps" | "context_id"
+>;
+
 export type ContinueRequest = StreamRequest & {
 	context_id: string;
 };
@@ -126,6 +131,28 @@ export type CloneOptions =
 			enhance?: boolean;
 	  };
 
+export type VoiceChangerOptions = {
+	clip: File;
+	voice: { id: string }; // match VoiceSpecifier shape, but only id is supported for now
+	output_format:
+		| {
+				container: "mp3";
+				bit_rate: number;
+				sample_rate: number;
+		  }
+		| {
+				container: "wav";
+				encoding: Encoding;
+				sample_rate: number;
+				bit_rate: number;
+		  }
+		| {
+				container: "raw";
+				encoding: Encoding;
+				sample_rate: number;
+		  };
+};
+
 export type LocalizeOptions = {
 	mode: "embedding";
 	embedding: number[];
@@ -167,6 +194,10 @@ export type CloneResponse = {
 	embedding: number[];
 };
 
+export type VoiceChangerBytesResponse = {
+	buffer: ArrayBuffer;
+};
+
 export type LocalizeResponse = {
 	embedding: number[];
 };

package/src/voice-changer/index.ts

@@ -0,0 +1,37 @@
+import { Client } from "../lib/client";
+import type { VoiceChangerBytesResponse, VoiceChangerOptions } from "../types";
+
+export default class VoiceChanger extends Client {
+	async bytes(
+		options: VoiceChangerOptions,
+	): Promise<VoiceChangerBytesResponse> {
+		const formData = new FormData();
+		formData.append("clip", options.clip); // TODO: handle Blobs that are not Files
+		formData.append("voice[id]", options.voice.id);
+
+		const fmt = options.output_format;
+		formData.append("output_format[container]", fmt.container);
+		if ("encoding" in fmt) {
+			formData.append("output_format[encoding]", fmt.encoding);
+		}
+		if ("bit_rate" in fmt) {
+			formData.append("output_format[bit_rate]", fmt.bit_rate.toString());
+		}
+		if ("sample_rate" in fmt) {
+			formData.append("output_format[sample_rate]", fmt.sample_rate.toString());
+		}
+
+		const response = await this._fetch("/voice-changer/bytes", {
+			method: "POST",
+			body: formData,
+		});
+
+		if (!response.ok) {
+			throw new Error(
+				`Voice changer error! status: ${response.status}, message: ${await response.text()}`,
+			);
+		}
+
+		return { buffer: await response.arrayBuffer() };
+	}
+}

package/.turbo/turbo-build.log

@@ -1,75 +0,0 @@
-$ tsup src/ --format cjs,esm --dts
-CLI Building entry: src/index.ts, src/lib/client.ts, src/lib/constants.ts, src/lib/index.ts, src/react/index.ts, src/react/utils.ts, src/tts/index.ts, src/tts/player.ts, src/tts/source.ts, src/tts/utils.ts, src/tts/websocket.ts, src/types/index.ts, src/voices/index.ts
-CLI Using tsconfig: tsconfig.json
-CLI tsup v8.0.2
-CLI Target: es6
-CJS Build start
-ESM Build start
-ESM dist/index.js          499.00 B
-ESM dist/chunk-FXPGR372.js 0 B
-ESM dist/lib/client.js     132.00 B
-ESM dist/react/index.js    7.73 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-4GEDAGVY.js 353.00 B
-ESM dist/react/utils.js    109.00 B
-ESM dist/chunk-CWIJUBG6.js 1.95 KB
-ESM dist/chunk-3FL2SNIR.js 337.00 B
-ESM dist/chunk-MUPVAEL7.js 439.00 B
-ESM dist/chunk-L3OMQKWL.js 7.95 KB
-ESM dist/chunk-6PWLZAMS.js 1.02 KB
-ESM dist/chunk-2BFEKY3F.js 366.00 B
-ESM dist/tts/player.js     174.00 B
-ESM dist/chunk-MNOPO7G6.js 3.76 KB
-ESM dist/chunk-VVDJR3OA.js 6.20 KB
-ESM dist/chunk-VCZESWYA.js 2.11 KB
-ESM dist/chunk-WIFMLPT5.js 2.27 KB
-ESM dist/types/index.js    31.00 B
-ESM dist/voices/index.js   174.00 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 ⚡️ Build success in 134ms
-CJS dist/index.cjs         26.45 KB
-CJS dist/lib/client.cjs    4.07 KB
-CJS dist/lib/constants.cjs 1.43 KB
-CJS dist/lib/index.cjs     22.35 KB
-CJS dist/react/index.cjs   34.12 KB
-CJS dist/tts/index.cjs     20.23 KB
-CJS dist/tts/player.cjs    6.96 KB
-CJS dist/react/utils.cjs   1.80 KB
-CJS dist/tts/utils.cjs     4.33 KB
-CJS dist/tts/source.cjs    9.03 KB
-CJS dist/types/index.cjs   764.00 B
-CJS dist/tts/websocket.cjs 19.92 KB
-CJS dist/voices/index.cjs  5.83 KB
-CJS ⚡️ Build success in 142ms
-DTS Build start
-DTS ⚡️ Build success in 7238ms
-DTS dist/index.d.cts         928.00 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
-DTS dist/react/utils.d.cts   240.00 B
-DTS dist/tts/index.d.cts     471.00 B
-DTS dist/tts/player.d.cts    1.20 KB
-DTS dist/tts/utils.d.cts     2.68 KB
-DTS dist/tts/websocket.d.cts 3.21 KB
-DTS dist/tts/source.d.cts    2.86 KB
-DTS dist/voices/index.d.cts  675.00 B
-DTS dist/lib/client.d.cts    293.00 B
-DTS dist/types/index.d.cts   3.96 KB
-DTS dist/index.d.ts          920.00 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
-DTS dist/react/utils.d.ts    240.00 B
-DTS dist/tts/index.d.ts      467.00 B
-DTS dist/tts/player.d.ts     1.20 KB
-DTS dist/tts/utils.d.ts      2.68 KB
-DTS dist/tts/websocket.d.ts  3.20 KB
-DTS dist/tts/source.d.ts     2.86 KB
-DTS dist/voices/index.d.ts   673.00 B
-DTS dist/lib/client.d.ts     292.00 B
-DTS dist/types/index.d.ts    3.96 KB

package/dist/chunk-MUPVAEL7.js

@@ -1,25 +0,0 @@
-import {
-  WebSocket
-} from "./chunk-L3OMQKWL.js";
-import {
-  Client
-} from "./chunk-6PWLZAMS.js";
-
-// src/tts/index.ts
-var TTS = class extends Client {
-  /**
-   * Get a WebSocket client for streaming audio from the TTS API.
-   *
-   * @returns {WebSocket} A Cartesia WebSocket client.
-   */
-  websocket(options) {
-    return new WebSocket(options, {
-      apiKey: this.apiKey,
-      baseUrl: this.baseUrl
-    });
-  }
-};
-
-export {
-  TTS
-};

package/tsconfig.json

@@ -1,3 +0,0 @@
-{
-	"extends": "@repo/config-typescript/react-library.json"
-}