import { z } from "zod";
export declare const TTS_CHATTERBOX_LANGUAGES: readonly ["en", "es", "fr", "de", "it", "pt", "nl", "pl", "tr", "sv", "da", "fi", "no", "el", "ms", "sw", "ar", "ko"];
export declare const TTS_SUPERTONIC_LANGUAGES: readonly ["en", "es", "fr", "pt", "ko"];
export declare const TTS_LANGUAGES: readonly ["en", "es", "fr", "de", "it", "pt", "nl", "pl", "tr", "sv", "da", "fi", "no", "el", "ms", "sw", "ar", "ko"];
export declare const ttsChatterboxRuntimeConfigSchema: z.ZodObject<{
    ttsEngine: z.ZodLiteral<"chatterbox">;
    language: z.ZodEnum<{
        en: "en";
        ar: "ar";
        de: "de";
        es: "es";
        fi: "fi";
        fr: "fr";
        it: "it";
        ko: "ko";
        nl: "nl";
        pl: "pl";
        pt: "pt";
        da: "da";
        el: "el";
        ms: "ms";
        no: "no";
        sv: "sv";
        tr: "tr";
        sw: "sw";
    }>;
    voice: z.ZodOptional<z.ZodString>;
    useGPU: z.ZodOptional<z.ZodBoolean>;
}, z.core.$strip>;
export declare const ttsSupertonicRuntimeConfigSchema: z.ZodObject<{
    ttsEngine: z.ZodLiteral<"supertonic">;
    language: z.ZodEnum<{
        en: "en";
        es: "es";
        fr: "fr";
        ko: "ko";
        pt: "pt";
    }>;
    voice: z.ZodOptional<z.ZodString>;
    ttsSpeed: z.ZodOptional<z.ZodNumber>;
    ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
    useGPU: z.ZodOptional<z.ZodBoolean>;
}, z.core.$strip>;
export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
    ttsEngine: z.ZodLiteral<"chatterbox">;
    language: z.ZodEnum<{
        en: "en";
        ar: "ar";
        de: "de";
        es: "es";
        fi: "fi";
        fr: "fr";
        it: "it";
        ko: "ko";
        nl: "nl";
        pl: "pl";
        pt: "pt";
        da: "da";
        el: "el";
        ms: "ms";
        no: "no";
        sv: "sv";
        tr: "tr";
        sw: "sw";
    }>;
    voice: z.ZodOptional<z.ZodString>;
    useGPU: z.ZodOptional<z.ZodBoolean>;
}, z.core.$strip>, z.ZodObject<{
    ttsEngine: z.ZodLiteral<"supertonic">;
    language: z.ZodEnum<{
        en: "en";
        es: "es";
        fr: "fr";
        ko: "ko";
        pt: "pt";
    }>;
    voice: z.ZodOptional<z.ZodString>;
    ttsSpeed: z.ZodOptional<z.ZodNumber>;
    ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
    useGPU: z.ZodOptional<z.ZodBoolean>;
}, z.core.$strip>]>;
export declare const ttsChatterboxLoadConfigSchema: z.ZodObject<{
    ttsEngine: z.ZodLiteral<"chatterbox">;
    language: z.ZodEnum<{
        en: "en";
        ar: "ar";
        de: "de";
        es: "es";
        fi: "fi";
        fr: "fr";
        it: "it";
        ko: "ko";
        nl: "nl";
        pl: "pl";
        pt: "pt";
        da: "da";
        el: "el";
        ms: "ms";
        no: "no";
        sv: "sv";
        tr: "tr";
        sw: "sw";
    }>;
    voice: z.ZodOptional<z.ZodString>;
    useGPU: z.ZodOptional<z.ZodBoolean>;
    s3genModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
        src: z.ZodString;
        name: z.ZodOptional<z.ZodString>;
        modelId: z.ZodOptional<z.ZodString>;
        registryPath: z.ZodOptional<z.ZodString>;
        registrySource: z.ZodOptional<z.ZodString>;
        blobCoreKey: z.ZodOptional<z.ZodString>;
        blobIndex: z.ZodOptional<z.ZodNumber>;
        engine: z.ZodOptional<z.ZodString>;
        expectedSize: z.ZodOptional<z.ZodNumber>;
        sha256Checksum: z.ZodOptional<z.ZodString>;
        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
            embeddings: "embeddings";
            diffusion: "diffusion";
            tts: "tts";
            ocr: "ocr";
            vla: "vla";
            "llamacpp-completion": "llamacpp-completion";
            "whispercpp-transcription": "whispercpp-transcription";
            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
            "llamacpp-embedding": "llamacpp-embedding";
            "nmtcpp-translation": "nmtcpp-translation";
            "onnx-tts": "onnx-tts";
            "tts-ggml": "tts-ggml";
            "parakeet-transcription": "parakeet-transcription";
            "onnx-ocr": "onnx-ocr";
            "sdcpp-generation": "sdcpp-generation";
            "ggml-vla": "ggml-vla";
            "ggml-classification": "ggml-classification";
            llm: "llm";
            whisper: "whisper";
            bci: "bci";
            nmt: "nmt";
            parakeet: "parakeet";
            classification: "classification";
        }>, z.ZodLiteral<"vad">]>>;
    }, z.core.$strip>]>>;
    referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
        src: z.ZodString;
        name: z.ZodOptional<z.ZodString>;
        modelId: z.ZodOptional<z.ZodString>;
        registryPath: z.ZodOptional<z.ZodString>;
        registrySource: z.ZodOptional<z.ZodString>;
        blobCoreKey: z.ZodOptional<z.ZodString>;
        blobIndex: z.ZodOptional<z.ZodNumber>;
        engine: z.ZodOptional<z.ZodString>;
        expectedSize: z.ZodOptional<z.ZodNumber>;
        sha256Checksum: z.ZodOptional<z.ZodString>;
        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
            embeddings: "embeddings";
            diffusion: "diffusion";
            tts: "tts";
            ocr: "ocr";
            vla: "vla";
            "llamacpp-completion": "llamacpp-completion";
            "whispercpp-transcription": "whispercpp-transcription";
            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
            "llamacpp-embedding": "llamacpp-embedding";
            "nmtcpp-translation": "nmtcpp-translation";
            "onnx-tts": "onnx-tts";
            "tts-ggml": "tts-ggml";
            "parakeet-transcription": "parakeet-transcription";
            "onnx-ocr": "onnx-ocr";
            "sdcpp-generation": "sdcpp-generation";
            "ggml-vla": "ggml-vla";
            "ggml-classification": "ggml-classification";
            llm: "llm";
            whisper: "whisper";
            bci: "bci";
            nmt: "nmt";
            parakeet: "parakeet";
            classification: "classification";
        }>, z.ZodLiteral<"vad">]>>;
    }, z.core.$strip>]>>;
}, z.core.$strip>;
export declare const ttsSupertonicLoadConfigSchema: z.ZodObject<{
    ttsEngine: z.ZodLiteral<"supertonic">;
    language: z.ZodEnum<{
        en: "en";
        es: "es";
        fr: "fr";
        ko: "ko";
        pt: "pt";
    }>;
    voice: z.ZodOptional<z.ZodString>;
    ttsSpeed: z.ZodOptional<z.ZodNumber>;
    ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
    useGPU: z.ZodOptional<z.ZodBoolean>;
}, z.core.$strip>;
export declare const ttsLoadConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
    ttsEngine: z.ZodLiteral<"chatterbox">;
    language: z.ZodEnum<{
        en: "en";
        ar: "ar";
        de: "de";
        es: "es";
        fi: "fi";
        fr: "fr";
        it: "it";
        ko: "ko";
        nl: "nl";
        pl: "pl";
        pt: "pt";
        da: "da";
        el: "el";
        ms: "ms";
        no: "no";
        sv: "sv";
        tr: "tr";
        sw: "sw";
    }>;
    voice: z.ZodOptional<z.ZodString>;
    useGPU: z.ZodOptional<z.ZodBoolean>;
    s3genModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
        src: z.ZodString;
        name: z.ZodOptional<z.ZodString>;
        modelId: z.ZodOptional<z.ZodString>;
        registryPath: z.ZodOptional<z.ZodString>;
        registrySource: z.ZodOptional<z.ZodString>;
        blobCoreKey: z.ZodOptional<z.ZodString>;
        blobIndex: z.ZodOptional<z.ZodNumber>;
        engine: z.ZodOptional<z.ZodString>;
        expectedSize: z.ZodOptional<z.ZodNumber>;
        sha256Checksum: z.ZodOptional<z.ZodString>;
        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
            embeddings: "embeddings";
            diffusion: "diffusion";
            tts: "tts";
            ocr: "ocr";
            vla: "vla";
            "llamacpp-completion": "llamacpp-completion";
            "whispercpp-transcription": "whispercpp-transcription";
            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
            "llamacpp-embedding": "llamacpp-embedding";
            "nmtcpp-translation": "nmtcpp-translation";
            "onnx-tts": "onnx-tts";
            "tts-ggml": "tts-ggml";
            "parakeet-transcription": "parakeet-transcription";
            "onnx-ocr": "onnx-ocr";
            "sdcpp-generation": "sdcpp-generation";
            "ggml-vla": "ggml-vla";
            "ggml-classification": "ggml-classification";
            llm: "llm";
            whisper: "whisper";
            bci: "bci";
            nmt: "nmt";
            parakeet: "parakeet";
            classification: "classification";
        }>, z.ZodLiteral<"vad">]>>;
    }, z.core.$strip>]>>;
    referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
        src: z.ZodString;
        name: z.ZodOptional<z.ZodString>;
        modelId: z.ZodOptional<z.ZodString>;
        registryPath: z.ZodOptional<z.ZodString>;
        registrySource: z.ZodOptional<z.ZodString>;
        blobCoreKey: z.ZodOptional<z.ZodString>;
        blobIndex: z.ZodOptional<z.ZodNumber>;
        engine: z.ZodOptional<z.ZodString>;
        expectedSize: z.ZodOptional<z.ZodNumber>;
        sha256Checksum: z.ZodOptional<z.ZodString>;
        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
            embeddings: "embeddings";
            diffusion: "diffusion";
            tts: "tts";
            ocr: "ocr";
            vla: "vla";
            "llamacpp-completion": "llamacpp-completion";
            "whispercpp-transcription": "whispercpp-transcription";
            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
            "llamacpp-embedding": "llamacpp-embedding";
            "nmtcpp-translation": "nmtcpp-translation";
            "onnx-tts": "onnx-tts";
            "tts-ggml": "tts-ggml";
            "parakeet-transcription": "parakeet-transcription";
            "onnx-ocr": "onnx-ocr";
            "sdcpp-generation": "sdcpp-generation";
            "ggml-vla": "ggml-vla";
            "ggml-classification": "ggml-classification";
            llm: "llm";
            whisper: "whisper";
            bci: "bci";
            nmt: "nmt";
            parakeet: "parakeet";
            classification: "classification";
        }>, z.ZodLiteral<"vad">]>>;
    }, z.core.$strip>]>>;
}, z.core.$strip>, z.ZodObject<{
    ttsEngine: z.ZodLiteral<"supertonic">;
    language: z.ZodEnum<{
        en: "en";
        es: "es";
        fr: "fr";
        ko: "ko";
        pt: "pt";
    }>;
    voice: z.ZodOptional<z.ZodString>;
    ttsSpeed: z.ZodOptional<z.ZodNumber>;
    ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
    useGPU: z.ZodOptional<z.ZodBoolean>;
}, z.core.$strip>]>;
export declare const LEGACY_TTS_ONNX_MODEL_CONFIG_FIELDS: readonly ["ttsSupertonicMultilingual", "ttsTokenizerSrc", "ttsSpeechEncoderSrc", "ttsEmbedTokensSrc", "ttsConditionalDecoderSrc", "ttsLanguageModelSrc", "ttsTextEncoderSrc", "ttsDurationPredictorSrc", "ttsVectorEstimatorSrc", "ttsVocoderSrc", "ttsUnicodeIndexerSrc", "ttsTtsConfigSrc", "ttsVoiceStyleSrc"];
export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
    [x: string]: z.ZodOptional<z.ZodUnknown>;
}, z.core.$strict>, z.ZodObject<{
    [x: string]: z.ZodOptional<z.ZodUnknown>;
}, z.core.$strict>]>;
export declare const ttsClientParamsSchema: z.ZodObject<{
    modelId: z.ZodString;
    inputType: z.ZodDefault<z.ZodString>;
    text: z.ZodString;
    stream: z.ZodDefault<z.ZodBoolean>;
    sentenceStream: z.ZodDefault<z.ZodBoolean>;
    sentenceStreamLocale: z.ZodOptional<z.ZodString>;
    sentenceStreamMaxChunkScalars: z.ZodOptional<z.ZodNumber>;
}, z.core.$strip>;
export declare const ttsRequestSchema: z.ZodObject<{
    modelId: z.ZodString;
    inputType: z.ZodDefault<z.ZodString>;
    text: z.ZodString;
    stream: z.ZodDefault<z.ZodBoolean>;
    sentenceStream: z.ZodDefault<z.ZodBoolean>;
    sentenceStreamLocale: z.ZodOptional<z.ZodString>;
    sentenceStreamMaxChunkScalars: z.ZodOptional<z.ZodNumber>;
    type: z.ZodLiteral<"textToSpeech">;
}, z.core.$strip>;
export declare const ttsStatsSchema: z.ZodObject<{
    audioDuration: z.ZodOptional<z.ZodNumber>;
    totalSamples: z.ZodOptional<z.ZodNumber>;
}, z.core.$strip>;
export declare const ttsResponseSchema: z.ZodObject<{
    type: z.ZodLiteral<"textToSpeech">;
    buffer: z.ZodArray<z.ZodNumber>;
    done: z.ZodDefault<z.ZodBoolean>;
    stats: z.ZodOptional<z.ZodObject<{
        audioDuration: z.ZodOptional<z.ZodNumber>;
        totalSamples: z.ZodOptional<z.ZodNumber>;
    }, z.core.$strip>>;
    chunkIndex: z.ZodOptional<z.ZodNumber>;
    sentenceChunk: z.ZodOptional<z.ZodString>;
}, z.core.$strip>;
declare const textToSpeechStreamRequestBaseSchema: z.ZodObject<{
    modelId: z.ZodString;
    inputType: z.ZodDefault<z.ZodString>;
    accumulateSentences: z.ZodOptional<z.ZodBoolean>;
    sentenceDelimiterPreset: z.ZodOptional<z.ZodEnum<{
        latin: "latin";
        cjk: "cjk";
        multilingual: "multilingual";
    }>>;
    maxBufferScalars: z.ZodOptional<z.ZodNumber>;
    flushAfterMs: z.ZodOptional<z.ZodNumber>;
}, z.core.$strip>;
export declare const textToSpeechStreamRequestSchema: z.ZodObject<{
    modelId: z.ZodString;
    inputType: z.ZodDefault<z.ZodString>;
    accumulateSentences: z.ZodOptional<z.ZodBoolean>;
    sentenceDelimiterPreset: z.ZodOptional<z.ZodEnum<{
        latin: "latin";
        cjk: "cjk";
        multilingual: "multilingual";
    }>>;
    maxBufferScalars: z.ZodOptional<z.ZodNumber>;
    flushAfterMs: z.ZodOptional<z.ZodNumber>;
    type: z.ZodLiteral<"textToSpeechStream">;
}, z.core.$strip>;
export declare const textToSpeechStreamResponseSchema: z.ZodObject<{
    type: z.ZodLiteral<"textToSpeechStream">;
    buffer: z.ZodArray<z.ZodNumber>;
    done: z.ZodDefault<z.ZodBoolean>;
    stats: z.ZodOptional<z.ZodObject<{
        audioDuration: z.ZodOptional<z.ZodNumber>;
        totalSamples: z.ZodOptional<z.ZodNumber>;
    }, z.core.$strip>>;
    chunkIndex: z.ZodOptional<z.ZodNumber>;
    sentenceChunk: z.ZodOptional<z.ZodString>;
}, z.core.$strip>;
export type TtsLanguage = (typeof TTS_LANGUAGES)[number];
export type TtsChatterboxLanguage = (typeof TTS_CHATTERBOX_LANGUAGES)[number];
export type TtsSupertonicLanguage = (typeof TTS_SUPERTONIC_LANGUAGES)[number];
export type TtsChatterboxLoadConfig = z.infer<typeof ttsChatterboxLoadConfigSchema>;
export type TtsSupertonicLoadConfig = z.infer<typeof ttsSupertonicLoadConfigSchema>;
export type TtsLoadConfig = z.infer<typeof ttsLoadConfigSchema>;
/** @deprecated Use {@link TtsChatterboxLoadConfig} */
export type TtsChatterboxConfig = TtsChatterboxLoadConfig;
/** @deprecated Use {@link TtsSupertonicLoadConfig} */
export type TtsSupertonicConfig = TtsSupertonicLoadConfig;
export type TtsChatterboxRuntimeConfig = z.infer<typeof ttsChatterboxRuntimeConfigSchema>;
export type TtsSupertonicRuntimeConfig = z.infer<typeof ttsSupertonicRuntimeConfigSchema>;
export type TtsRuntimeConfig = z.infer<typeof ttsRuntimeConfigSchema>;
export type TtsConfig = z.infer<typeof ttsConfigSchema>;
export type TtsClientParamsInput = z.input<typeof ttsClientParamsSchema>;
export type TtsClientParams = z.output<typeof ttsClientParamsSchema>;
export type TtsRequest = z.infer<typeof ttsRequestSchema>;
export type TtsResponse = z.infer<typeof ttsResponseSchema>;
export type TtsStats = z.infer<typeof ttsStatsSchema>;
export type TtsSentenceChunkUpdate = {
    buffer: number[];
    chunkIndex?: number;
    sentenceChunk?: string;
};
export type TextToSpeechStreamRequest = z.infer<typeof textToSpeechStreamRequestSchema>;
export type TextToSpeechStreamResponse = z.infer<typeof textToSpeechStreamResponseSchema>;
export type TextToSpeechStreamClientParams = z.infer<typeof textToSpeechStreamRequestBaseSchema>;
export interface TextToSpeechStreamResult {
    bufferStream: AsyncGenerator<number>;
    chunkUpdates?: AsyncGenerator<TtsSentenceChunkUpdate>;
    buffer: Promise<number[]>;
    done: Promise<boolean>;
}
export interface TextToSpeechStreamSession {
    write(textFragment: string | Buffer): void;
    end(): void;
    destroy(): void;
    [Symbol.asyncIterator](): AsyncIterator<TextToSpeechStreamResponse>;
}
export {};
//# sourceMappingURL=text-to-speech.d.ts.map