@supertone/supertone 0.1.0 → 0.1.2

package/src/lib/custom_utils/text_utils.ts

@@ -7,6 +7,94 @@
 
 import { DEFAULT_MAX_TEXT_LENGTH } from "./constants.js";
 
+/**
+ * Check if text contains spaces (to determine if word-based splitting is possible)
+ *
+ * @param text - Text to check
+ * @returns true if text contains spaces
+ */
+function hasSpaces(text: string): boolean {
+	return /\s/.test(text);
+}
+
+/**
+ * Split text by words, ensuring each chunk is under maxLength.
+ * Used for languages with spaces (English, Korean, etc.)
+ *
+ * @param text - Text to split
+ * @param maxLength - Maximum length of each chunk
+ * @returns Array of text chunks
+ */
+function splitByWords(text: string, maxLength: number): string[] {
+	const words = text.split(/(\s+)/);
+	const chunks: string[] = [];
+	let currentChunk = "";
+
+	for (const word of words) {
+		if (currentChunk.length + word.length <= maxLength) {
+			currentChunk += word;
+		} else {
+			if (currentChunk.trim()) {
+				chunks.push(currentChunk.trim());
+			}
+			// If a single word exceeds maxLength, split by characters
+			if (word.trim().length > maxLength) {
+				const charChunks = splitByCharacters(word.trim(), maxLength);
+				chunks.push(...charChunks);
+				currentChunk = "";
+			} else {
+				currentChunk = word;
+			}
+		}
+	}
+
+	if (currentChunk.trim()) {
+		chunks.push(currentChunk.trim());
+	}
+
+	return chunks;
+}
+
+/**
+ * Split text by characters, ensuring each chunk is under maxLength.
+ * Used for languages without spaces (Japanese, Chinese, etc.)
+ *
+ * @param text - Text to split
+ * @param maxLength - Maximum length of each chunk
+ * @returns Array of text chunks
+ */
+function splitByCharacters(text: string, maxLength: number): string[] {
+	const chunks: string[] = [];
+
+	for (let i = 0; i < text.length; i += maxLength) {
+		chunks.push(text.slice(i, i + maxLength));
+	}
+
+	return chunks;
+}
+
+/**
+ * Split a single chunk that exceeds maxLength into smaller chunks.
+ * Uses word-based splitting for texts with spaces, character-based for texts without.
+ *
+ * @param chunk - Text chunk to split
+ * @param maxLength - Maximum length of each chunk
+ * @returns Array of text chunks, all under maxLength
+ */
+function splitOversizedChunk(chunk: string, maxLength: number): string[] {
+	if (chunk.length <= maxLength) {
+		return [chunk];
+	}
+
+	// Check if text has spaces (word-based splitting possible)
+	if (hasSpaces(chunk)) {
+		return splitByWords(chunk, maxLength);
+	}
+
+	// No spaces: use character-based splitting (Japanese, Chinese, etc.)
+	return splitByCharacters(chunk, maxLength);
+}
+
 /**
  * Split input text into sentence chunks suitable for TTS processing.
  *
@@ -15,9 +103,16 @@ import { DEFAULT_MAX_TEXT_LENGTH } from "./constants.js";
  * It handles various punctuation patterns and provides graceful fallback to
  * word/character boundaries when necessary.
  *
+ * Chunking Strategy:
+ * 1. First, split by sentence boundaries (punctuation: .!?;:)
+ * 2. Merge sentences into chunks up to maxLength
+ * 3. If a sentence exceeds maxLength:
+ *    - For text with spaces: split by words
+ *    - For text without spaces (Japanese, etc.): split by characters
+ *
  * @param text - Input text to be segmented
  * @param maxLength - Maximum length of each chunk
- * @returns Array of text chunks
+ * @returns Array of text chunks, each guaranteed to be <= maxLength
  */
 export function chunkText(
 	text: string,
@@ -27,28 +122,43 @@ export function chunkText(
 		return [text];
 	}
 
-	// Split by sentence boundaries
-	const sentences = text.split(/([.!?;:]+\s*)/);
+	// Step 1: Split by sentence boundaries (including various punctuation marks)
+	// Includes Western punctuation (.!?;:) and CJK punctuation (。！？；：)
+	const sentences = text.split(/([.!?;:。！？；：]+\s*)/);
 
-	const chunks: string[] = [];
+	const preliminaryChunks: string[] = [];
 	let currentChunk = "";
 
+	// Step 2: Merge sentences into chunks up to maxLength
 	for (const sentence of sentences) {
 		if (currentChunk.length + sentence.length <= maxLength) {
 			currentChunk += sentence;
 		} else {
 			if (currentChunk) {
-				chunks.push(currentChunk);
+				preliminaryChunks.push(currentChunk);
 			}
 			currentChunk = sentence;
 		}
 	}
 
 	if (currentChunk) {
-		chunks.push(currentChunk);
+		preliminaryChunks.push(currentChunk);
 	}
 
-	return chunks;
+	// Step 3: Handle oversized chunks (split by words or characters)
+	const finalChunks: string[] = [];
+	for (const chunk of preliminaryChunks) {
+		if (chunk.length <= maxLength) {
+			finalChunks.push(chunk);
+		} else {
+			// Chunk exceeds maxLength, need to split further
+			const subChunks = splitOversizedChunk(chunk, maxLength);
+			finalChunks.push(...subChunks);
+		}
+	}
+
+	// Filter out empty chunks
+	return finalChunks.filter((chunk) => chunk.length > 0);
 }
 
 /**

package/src/models/apiconverttexttospeechusingcharacterrequest.ts

@@ -22,8 +22,26 @@ export const APIConvertTextToSpeechUsingCharacterRequestLanguage = {
   En: "en",
   Ko: "ko",
   Ja: "ja",
+  Bg: "bg",
+  Cs: "cs",
+  Da: "da",
+  El: "el",
   Es: "es",
+  Et: "et",
+  Fi: "fi",
+  Hu: "hu",
+  It: "it",
+  Nl: "nl",
+  Pl: "pl",
   Pt: "pt",
+  Ro: "ro",
+  Ar: "ar",
+  De: "de",
+  Fr: "fr",
+  Hi: "hi",
+  Id: "id",
+  Ru: "ru",
+  Vi: "vi",
 } as const;
 /**
  * The language code of the text
@@ -32,6 +50,22 @@ export type APIConvertTextToSpeechUsingCharacterRequestLanguage = ClosedEnum<
   typeof APIConvertTextToSpeechUsingCharacterRequestLanguage
 >;
 
+/**
+ * The model type to use for the text-to-speech conversion
+ */
+export const APIConvertTextToSpeechUsingCharacterRequestModel = {
+  SonaSpeech1: "sona_speech_1",
+  SonaSpeech2: "sona_speech_2",
+  SonaSpeech2t: "sona_speech_2t",
+  SupertonicApi1: "supertonic_api_1",
+} as const;
+/**
+ * The model type to use for the text-to-speech conversion
+ */
+export type APIConvertTextToSpeechUsingCharacterRequestModel = ClosedEnum<
+  typeof APIConvertTextToSpeechUsingCharacterRequestModel
+>;
+
 /**
  * The desired output format of the audio file (wav, mp3). Default is wav.
  */
@@ -61,7 +95,7 @@ export type APIConvertTextToSpeechUsingCharacterRequest = {
   /**
    * The model type to use for the text-to-speech conversion
    */
-  model?: string | undefined;
+  model?: APIConvertTextToSpeechUsingCharacterRequestModel | undefined;
   /**
    * The desired output format of the audio file (wav, mp3). Default is wav.
    */
@@ -98,6 +132,29 @@ export namespace APIConvertTextToSpeechUsingCharacterRequestLanguage$ {
     APIConvertTextToSpeechUsingCharacterRequestLanguage$outboundSchema;
 }
 
+/** @internal */
+export const APIConvertTextToSpeechUsingCharacterRequestModel$inboundSchema:
+  z.ZodNativeEnum<typeof APIConvertTextToSpeechUsingCharacterRequestModel> = z
+    .nativeEnum(APIConvertTextToSpeechUsingCharacterRequestModel);
+
+/** @internal */
+export const APIConvertTextToSpeechUsingCharacterRequestModel$outboundSchema:
+  z.ZodNativeEnum<typeof APIConvertTextToSpeechUsingCharacterRequestModel> =
+    APIConvertTextToSpeechUsingCharacterRequestModel$inboundSchema;
+
+/**
+ * @internal
+ * @deprecated This namespace will be removed in future versions. Use schemas and types that are exported directly from this module.
+ */
+export namespace APIConvertTextToSpeechUsingCharacterRequestModel$ {
+  /** @deprecated use `APIConvertTextToSpeechUsingCharacterRequestModel$inboundSchema` instead. */
+  export const inboundSchema =
+    APIConvertTextToSpeechUsingCharacterRequestModel$inboundSchema;
+  /** @deprecated use `APIConvertTextToSpeechUsingCharacterRequestModel$outboundSchema` instead. */
+  export const outboundSchema =
+    APIConvertTextToSpeechUsingCharacterRequestModel$outboundSchema;
+}
+
 /** @internal */
 export const APIConvertTextToSpeechUsingCharacterRequestOutputFormat$inboundSchema:
   z.ZodNativeEnum<
@@ -133,7 +190,8 @@ export const APIConvertTextToSpeechUsingCharacterRequest$inboundSchema:
     text: z.string(),
     language: APIConvertTextToSpeechUsingCharacterRequestLanguage$inboundSchema,
     style: z.string().optional(),
-    model: z.string().default("sona_speech_1"),
+    model: APIConvertTextToSpeechUsingCharacterRequestModel$inboundSchema
+      .default("sona_speech_1"),
     output_format:
       APIConvertTextToSpeechUsingCharacterRequestOutputFormat$inboundSchema
         .default("wav"),
@@ -169,7 +227,8 @@ export const APIConvertTextToSpeechUsingCharacterRequest$outboundSchema:
     language:
       APIConvertTextToSpeechUsingCharacterRequestLanguage$outboundSchema,
     style: z.string().optional(),
-    model: z.string().default("sona_speech_1"),
+    model: APIConvertTextToSpeechUsingCharacterRequestModel$outboundSchema
+      .default("sona_speech_1"),
     outputFormat:
       APIConvertTextToSpeechUsingCharacterRequestOutputFormat$outboundSchema
         .default("wav"),

package/src/models/predictttsdurationusingcharacterrequest.ts

@@ -22,8 +22,26 @@ export const PredictTTSDurationUsingCharacterRequestLanguage = {
   En: "en",
   Ko: "ko",
   Ja: "ja",
+  Bg: "bg",
+  Cs: "cs",
+  Da: "da",
+  El: "el",
   Es: "es",
+  Et: "et",
+  Fi: "fi",
+  Hu: "hu",
+  It: "it",
+  Nl: "nl",
+  Pl: "pl",
   Pt: "pt",
+  Ro: "ro",
+  Ar: "ar",
+  De: "de",
+  Fr: "fr",
+  Hi: "hi",
+  Id: "id",
+  Ru: "ru",
+  Vi: "vi",
 } as const;
 /**
  * Language code of the voice
@@ -32,6 +50,22 @@ export type PredictTTSDurationUsingCharacterRequestLanguage = ClosedEnum<
   typeof PredictTTSDurationUsingCharacterRequestLanguage
 >;
 
+/**
+ * The model type to use for the text-to-speech conversion
+ */
+export const PredictTTSDurationUsingCharacterRequestModel = {
+  SonaSpeech1: "sona_speech_1",
+  SonaSpeech2: "sona_speech_2",
+  SonaSpeech2t: "sona_speech_2t",
+  SupertonicApi1: "supertonic_api_1",
+} as const;
+/**
+ * The model type to use for the text-to-speech conversion
+ */
+export type PredictTTSDurationUsingCharacterRequestModel = ClosedEnum<
+  typeof PredictTTSDurationUsingCharacterRequestModel
+>;
+
 /**
  * The desired output format of the audio file (wav, mp3). Default is wav.
  */
@@ -62,7 +96,7 @@ export type PredictTTSDurationUsingCharacterRequest = {
   /**
    * The model type to use for the text-to-speech conversion
    */
-  model?: string | undefined;
+  model?: PredictTTSDurationUsingCharacterRequestModel | undefined;
   /**
    * The desired output format of the audio file (wav, mp3). Default is wav.
    */
@@ -95,6 +129,29 @@ export namespace PredictTTSDurationUsingCharacterRequestLanguage$ {
     PredictTTSDurationUsingCharacterRequestLanguage$outboundSchema;
 }
 
+/** @internal */
+export const PredictTTSDurationUsingCharacterRequestModel$inboundSchema:
+  z.ZodNativeEnum<typeof PredictTTSDurationUsingCharacterRequestModel> = z
+    .nativeEnum(PredictTTSDurationUsingCharacterRequestModel);
+
+/** @internal */
+export const PredictTTSDurationUsingCharacterRequestModel$outboundSchema:
+  z.ZodNativeEnum<typeof PredictTTSDurationUsingCharacterRequestModel> =
+    PredictTTSDurationUsingCharacterRequestModel$inboundSchema;
+
+/**
+ * @internal
+ * @deprecated This namespace will be removed in future versions. Use schemas and types that are exported directly from this module.
+ */
+export namespace PredictTTSDurationUsingCharacterRequestModel$ {
+  /** @deprecated use `PredictTTSDurationUsingCharacterRequestModel$inboundSchema` instead. */
+  export const inboundSchema =
+    PredictTTSDurationUsingCharacterRequestModel$inboundSchema;
+  /** @deprecated use `PredictTTSDurationUsingCharacterRequestModel$outboundSchema` instead. */
+  export const outboundSchema =
+    PredictTTSDurationUsingCharacterRequestModel$outboundSchema;
+}
+
 /** @internal */
 export const PredictTTSDurationUsingCharacterRequestOutputFormat$inboundSchema:
   z.ZodNativeEnum<typeof PredictTTSDurationUsingCharacterRequestOutputFormat> =
@@ -127,7 +184,9 @@ export const PredictTTSDurationUsingCharacterRequest$inboundSchema: z.ZodType<
   text: z.string(),
   language: PredictTTSDurationUsingCharacterRequestLanguage$inboundSchema,
   style: z.string().optional(),
-  model: z.string().default("sona_speech_1"),
+  model: PredictTTSDurationUsingCharacterRequestModel$inboundSchema.default(
+    "sona_speech_1",
+  ),
   output_format:
     PredictTTSDurationUsingCharacterRequestOutputFormat$inboundSchema.default(
       "wav",
@@ -159,7 +218,9 @@ export const PredictTTSDurationUsingCharacterRequest$outboundSchema: z.ZodType<
   text: z.string(),
   language: PredictTTSDurationUsingCharacterRequestLanguage$outboundSchema,
   style: z.string().optional(),
-  model: z.string().default("sona_speech_1"),
+  model: PredictTTSDurationUsingCharacterRequestModel$outboundSchema.default(
+    "sona_speech_1",
+  ),
   outputFormat:
     PredictTTSDurationUsingCharacterRequestOutputFormat$outboundSchema.default(
       "wav",