@cartesia/cartesia-js 1.0.2 → 1.1.0

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[];
 	};
 };
@@ -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 +126,25 @@ 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[];
+	weight: number;
+}
+
+export interface MixVoicesOptions {
+	voices: VoiceToMix[];
+}
+
 export type Voice = {
 	id: string;
 	name: string;
@@ -120,7 +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"> &
@@ -134,6 +167,14 @@ export type CloneResponse = {
 	embedding: number[];
 };
 
+export type LocalizeResponse = {
+	embedding: number[];
+};
+
+export type MixVoicesResponse = {
+	embedding: number[];
+};
+
 export type WebSocketOptions = {
 	container?: string;
 	encoding?: string;

package/src/voices/index.ts

@@ -3,6 +3,10 @@ import type {
 	CloneOptions,
 	CloneResponse,
 	CreateVoice,
+	LocalizeOptions,
+	LocalizeResponse,
+	MixVoicesOptions,
+	MixVoicesResponse,
 	UpdateVoice,
 	Voice,
 } from "../types";
@@ -51,4 +55,22 @@ export default class Voices extends Client {
 
 		throw new Error("Invalid mode for clone()");
 	}
+
+	async mix(options: MixVoicesOptions): Promise<MixVoicesResponse> {
+		const response = await this._fetch("/voices/mix", {
+			method: "POST",
+			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>;
+	}
 }