@qvac/sdk 0.10.2 → 0.12.0

package/dist/schemas/text-to-speech.d.ts

@@ -8,6 +8,8 @@ export declare const ttsChatterboxRuntimeConfigSchema: z.ZodObject<{
         es: "es";
         it: "it";
     }>;
+    voice: z.ZodOptional<z.ZodString>;
+    useGPU: z.ZodOptional<z.ZodBoolean>;
 }, z.core.$strip>;
 export declare const ttsSupertonicRuntimeConfigSchema: z.ZodObject<{
     ttsEngine: z.ZodLiteral<"supertonic">;
@@ -17,9 +19,10 @@ export declare const ttsSupertonicRuntimeConfigSchema: z.ZodObject<{
         es: "es";
         it: "it";
     }>;
+    voice: z.ZodOptional<z.ZodString>;
     ttsSpeed: z.ZodOptional<z.ZodNumber>;
     ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
-    ttsSupertonicMultilingual: z.ZodOptional<z.ZodBoolean>;
+    useGPU: z.ZodOptional<z.ZodBoolean>;
 }, z.core.$strip>;
 export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
     ttsEngine: z.ZodLiteral<"chatterbox">;
@@ -29,6 +32,8 @@ export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
         es: "es";
         it: "it";
     }>;
+    voice: z.ZodOptional<z.ZodString>;
+    useGPU: z.ZodOptional<z.ZodBoolean>;
 }, z.core.$strip>, z.ZodObject<{
     ttsEngine: z.ZodLiteral<"supertonic">;
     language: z.ZodEnum<{
@@ -37,11 +42,12 @@ export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
         es: "es";
         it: "it";
     }>;
+    voice: z.ZodOptional<z.ZodString>;
     ttsSpeed: z.ZodOptional<z.ZodNumber>;
     ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
-    ttsSupertonicMultilingual: z.ZodOptional<z.ZodBoolean>;
+    useGPU: z.ZodOptional<z.ZodBoolean>;
 }, z.core.$strip>]>;
-export declare const ttsChatterboxConfigSchema: z.ZodObject<{
+export declare const ttsChatterboxLoadConfigSchema: z.ZodObject<{
     ttsEngine: z.ZodLiteral<"chatterbox">;
     language: z.ZodEnum<{
         en: "en";
@@ -49,7 +55,9 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
         es: "es";
         it: "it";
     }>;
-    ttsTokenizerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
+    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>;
@@ -62,525 +70,29 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsSpeechEncoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsEmbedTokensSrc: 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";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsConditionalDecoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsLanguageModelSrc: 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";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    referenceAudioSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-}, z.core.$strip>;
-export declare const ttsSupertonicConfigSchema: z.ZodObject<{
-    ttsEngine: z.ZodLiteral<"supertonic">;
-    language: z.ZodEnum<{
-        en: "en";
-        de: "de";
-        es: "es";
-        it: "it";
-    }>;
-    ttsSpeed: z.ZodOptional<z.ZodNumber>;
-    ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
-    ttsSupertonicMultilingual: z.ZodOptional<z.ZodBoolean>;
-    ttsTextEncoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsDurationPredictorSrc: 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";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsVectorEstimatorSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsVocoderSrc: 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";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsUnicodeIndexerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsTtsConfigSrc: 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";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsVoiceStyleSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-}, z.core.$strip>;
-export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
-    ttsEngine: z.ZodLiteral<"chatterbox">;
-    language: z.ZodEnum<{
-        en: "en";
-        de: "de";
-        es: "es";
-        it: "it";
-    }>;
-    ttsTokenizerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsSpeechEncoderSrc: 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";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
             diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsEmbedTokensSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
             tts: "tts";
             ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsConditionalDecoderSrc: 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";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "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";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsLanguageModelSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
+    }, z.core.$strip>]>>;
+    referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
         src: z.ZodString;
         name: z.ZodOptional<z.ZodString>;
         modelId: z.ZodOptional<z.ZodString>;
@@ -593,54 +105,30 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
+            diffusion: "diffusion";
             tts: "tts";
             ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    referenceAudioSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-}, z.core.$strip>, z.ZodObject<{
+    }, z.core.$strip>]>>;
+}, z.core.$strip>;
+export declare const ttsSupertonicLoadConfigSchema: z.ZodObject<{
     ttsEngine: z.ZodLiteral<"supertonic">;
     language: z.ZodEnum<{
         en: "en";
@@ -648,40 +136,22 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
         es: "es";
         it: "it";
     }>;
+    voice: z.ZodOptional<z.ZodString>;
     ttsSpeed: z.ZodOptional<z.ZodNumber>;
     ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
-    ttsSupertonicMultilingual: z.ZodOptional<z.ZodBoolean>;
-    ttsTextEncoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsDurationPredictorSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
+    useGPU: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+export declare const ttsLoadConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
+    ttsEngine: z.ZodLiteral<"chatterbox">;
+    language: z.ZodEnum<{
+        en: "en";
+        de: "de";
+        es: "es";
+        it: "it";
+    }>;
+    voice: z.ZodOptional<z.ZodString>;
+    useGPU: z.ZodOptional<z.ZodBoolean>;
+    s3genModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
         src: z.ZodString;
         name: z.ZodOptional<z.ZodString>;
         modelId: z.ZodOptional<z.ZodString>;
@@ -694,84 +164,29 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
             diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsVectorEstimatorSrc: 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";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
             tts: "tts";
             ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsVocoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsUnicodeIndexerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
+    }, z.core.$strip>]>>;
+    referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
         src: z.ZodString;
         name: z.ZodOptional<z.ZodString>;
         modelId: z.ZodOptional<z.ZodString>;
@@ -784,84 +199,47 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
             diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsTtsConfigSrc: 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";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
             tts: "tts";
             ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
-    ttsVoiceStyleSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>;
+    }, z.core.$strip>]>>;
+}, z.core.$strip>, z.ZodObject<{
+    ttsEngine: z.ZodLiteral<"supertonic">;
+    language: z.ZodEnum<{
+        en: "en";
+        de: "de";
+        es: "es";
+        it: "it";
+    }>;
+    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.$strict>]>;
 export declare const ttsClientParamsSchema: z.ZodObject<{
     modelId: z.ZodString;
     inputType: z.ZodDefault<z.ZodString>;
@@ -933,12 +311,17 @@ export declare const textToSpeechStreamResponseSchema: z.ZodObject<{
     sentenceChunk: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 export type TtsLanguage = (typeof TTS_LANGUAGES)[number];
-export type TtsChatterboxConfig = z.infer<typeof ttsChatterboxConfigSchema>;
-export type TtsSupertonicConfig = z.infer<typeof ttsSupertonicConfigSchema>;
-export type TtsConfig = z.infer<typeof ttsConfigSchema>;
+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>;