@qvac/sdk 0.8.2 → 0.9.0

package/dist/server/bare/plugins/onnx-tts/plugin.d.ts

@@ -30,6 +30,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -37,6 +38,7 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
         ttsSpeechEncoderSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -58,6 +60,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -65,6 +68,7 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
         ttsEmbedTokensSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -86,6 +90,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -93,6 +98,7 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
         ttsConditionalDecoderSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -114,6 +120,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -121,6 +128,7 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
         ttsLanguageModelSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -142,6 +150,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -149,6 +158,7 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
         referenceAudioSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -170,6 +180,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -177,6 +188,7 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
     }, import("zod/v4/core").$strip>, import("zod").ZodObject<{
@@ -189,7 +201,8 @@ export declare const ttsPlugin: {
         }>;
         ttsSpeed: import("zod").ZodOptional<import("zod").ZodNumber>;
         ttsNumInferenceSteps: import("zod").ZodOptional<import("zod").ZodNumber>;
-        ttsTokenizerSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+        ttsSupertonicMultilingual: import("zod").ZodOptional<import("zod").ZodBoolean>;
+        ttsTextEncoderSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
             src: import("zod").ZodString;
             name: import("zod").ZodOptional<import("zod").ZodString>;
             modelId: import("zod").ZodOptional<import("zod").ZodString>;
@@ -208,6 +221,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -215,9 +229,40 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
-        ttsTextEncoderSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+        ttsDurationPredictorSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+            src: import("zod").ZodString;
+            name: import("zod").ZodOptional<import("zod").ZodString>;
+            modelId: import("zod").ZodOptional<import("zod").ZodString>;
+            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
+            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
+            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
+            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            engine: import("zod").ZodOptional<import("zod").ZodString>;
+            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
+            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
+                "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";
+                embeddings: "embeddings";
+                nmt: "nmt";
+                parakeet: "parakeet";
+                tts: "tts";
+                ocr: "ocr";
+                diffusion: "diffusion";
+            }>, import("zod").ZodLiteral<"vad">]>>;
+        }, import("zod/v4/core").$strip>]>;
+        ttsVectorEstimatorSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
             src: import("zod").ZodString;
             name: import("zod").ZodOptional<import("zod").ZodString>;
             modelId: import("zod").ZodOptional<import("zod").ZodString>;
@@ -236,6 +281,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -243,9 +289,10 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
-        ttsLatentDenoiserSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+        ttsVocoderSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
             src: import("zod").ZodString;
             name: import("zod").ZodOptional<import("zod").ZodString>;
             modelId: import("zod").ZodOptional<import("zod").ZodString>;
@@ -264,6 +311,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -271,9 +319,10 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
-        ttsVoiceDecoderSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+        ttsUnicodeIndexerSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
             src: import("zod").ZodString;
             name: import("zod").ZodOptional<import("zod").ZodString>;
             modelId: import("zod").ZodOptional<import("zod").ZodString>;
@@ -292,6 +341,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -299,9 +349,10 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
-        ttsVoiceSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+        ttsTtsConfigSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
             src: import("zod").ZodString;
             name: import("zod").ZodOptional<import("zod").ZodString>;
             modelId: import("zod").ZodOptional<import("zod").ZodString>;
@@ -320,6 +371,7 @@ export declare const ttsPlugin: {
                 "onnx-tts": "onnx-tts";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
                 llm: "llm";
                 whisper: "whisper";
                 embeddings: "embeddings";
@@ -327,6 +379,37 @@ export declare const ttsPlugin: {
                 parakeet: "parakeet";
                 tts: "tts";
                 ocr: "ocr";
+                diffusion: "diffusion";
+            }>, import("zod").ZodLiteral<"vad">]>>;
+        }, import("zod/v4/core").$strip>]>;
+        ttsVoiceStyleSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+            src: import("zod").ZodString;
+            name: import("zod").ZodOptional<import("zod").ZodString>;
+            modelId: import("zod").ZodOptional<import("zod").ZodString>;
+            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
+            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
+            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
+            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            engine: import("zod").ZodOptional<import("zod").ZodString>;
+            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
+            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
+                "llamacpp-completion": "llamacpp-completion";
+                "whispercpp-transcription": "whispercpp-transcription";
+                "llamacpp-embedding": "llamacpp-embedding";
+                "nmtcpp-translation": "nmtcpp-translation";
+                "onnx-tts": "onnx-tts";
+                "parakeet-transcription": "parakeet-transcription";
+                "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
+                llm: "llm";
+                whisper: "whisper";
+                embeddings: "embeddings";
+                nmt: "nmt";
+                parakeet: "parakeet";
+                tts: "tts";
+                ocr: "ocr";
+                diffusion: "diffusion";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>;
     }, import("zod/v4/core").$strip>]>;
@@ -344,11 +427,13 @@ export declare const ttsPlugin: {
     } | {
         config: TtsSupertonicRuntimeConfig;
         artifacts: {
-            tokenizerPath: string;
             textEncoderPath: string;
-            latentDenoiserPath: string;
-            voiceDecoderPath: string;
-            voicePath: string;
+            durationPredictorPath: string;
+            vectorEstimatorPath: string;
+            vocoderPath: string;
+            unicodeIndexerPath: string;
+            ttsConfigPath: string;
+            voiceStylePath: string;
         };
     }>;
     createModel(params: CreateModelParams): PluginModelResult;
@@ -363,6 +448,10 @@ export declare const ttsPlugin: {
             type: import("zod").ZodLiteral<"textToSpeech">;
             buffer: import("zod").ZodArray<import("zod").ZodNumber>;
             done: import("zod").ZodDefault<import("zod").ZodBoolean>;
+            stats: import("zod").ZodOptional<import("zod").ZodObject<{
+                audioDuration: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalSamples: import("zod").ZodOptional<import("zod").ZodNumber>;
+            }, import("zod/v4/core").$strip>>;
         }, import("zod/v4/core").$strip>>;
     };
     logging: {

package/dist/server/bare/plugins/onnx-tts/plugin.js

@@ -5,6 +5,7 @@ import { definePlugin, defineHandler, ttsRequestSchema, ttsResponseSchema, Model
 import { createStreamLogger, registerAddonLogger } from "../../../../logging/index.js";
 import { TtsArtifactsRequiredError, TtsReferenceAudioRequiredError, } from "../../../../utils/errors-server.js";
 import { textToSpeech } from "./ops/text-to-speech.js";
+import { attachModelExecutionMs } from "../../../../profiling/model-execution.js";
 import { loadReferenceAudioAt24k } from "./wav-helper.js";
 async function resolveChatterboxConfig(config, ctx) {
     const { ttsTokenizerSrc, ttsSpeechEncoderSrc, ttsEmbedTokensSrc, ttsConditionalDecoderSrc, ttsLanguageModelSrc, referenceAudioSrc, language, } = config;
@@ -43,21 +44,25 @@ async function resolveChatterboxConfig(config, ctx) {
     };
 }
 async function resolveSupertonicConfig(config, ctx) {
-    const { ttsTokenizerSrc, ttsTextEncoderSrc, ttsLatentDenoiserSrc, ttsVoiceDecoderSrc, ttsVoiceSrc, ttsSpeed, ttsNumInferenceSteps, language, } = config;
-    if (!ttsTokenizerSrc ||
-        !ttsTextEncoderSrc ||
-        !ttsLatentDenoiserSrc ||
-        !ttsVoiceDecoderSrc ||
-        !ttsVoiceSrc) {
+    const { ttsTextEncoderSrc, ttsDurationPredictorSrc, ttsVectorEstimatorSrc, ttsVocoderSrc, ttsUnicodeIndexerSrc, ttsTtsConfigSrc, ttsVoiceStyleSrc, ttsSpeed, ttsNumInferenceSteps, ttsSupertonicMultilingual, language, } = config;
+    if (!ttsTextEncoderSrc ||
+        !ttsDurationPredictorSrc ||
+        !ttsVectorEstimatorSrc ||
+        !ttsVocoderSrc ||
+        !ttsUnicodeIndexerSrc ||
+        !ttsTtsConfigSrc ||
+        !ttsVoiceStyleSrc) {
         throw new TtsArtifactsRequiredError();
     }
     const resolve = ctx.resolveModelPath;
-    const [tokenizerPath, textEncoderPath, latentDenoiserPath, voiceDecoderPath, voicePath,] = await Promise.all([
-        resolve(ttsTokenizerSrc),
+    const [textEncoderPath, durationPredictorPath, vectorEstimatorPath, vocoderPath, unicodeIndexerPath, ttsConfigPath, voiceStylePath,] = await Promise.all([
         resolve(ttsTextEncoderSrc),
-        resolve(ttsLatentDenoiserSrc),
-        resolve(ttsVoiceDecoderSrc),
-        resolve(ttsVoiceSrc),
+        resolve(ttsDurationPredictorSrc),
+        resolve(ttsVectorEstimatorSrc),
+        resolve(ttsVocoderSrc),
+        resolve(ttsUnicodeIndexerSrc),
+        resolve(ttsTtsConfigSrc),
+        resolve(ttsVoiceStyleSrc),
     ]);
     return {
         config: {
@@ -65,13 +70,16 @@ async function resolveSupertonicConfig(config, ctx) {
             language,
             ttsSpeed,
             ttsNumInferenceSteps,
+            ttsSupertonicMultilingual,
         },
         artifacts: {
-            tokenizerPath,
             textEncoderPath,
-            latentDenoiserPath,
-            voiceDecoderPath,
-            voicePath,
+            durationPredictorPath,
+            vectorEstimatorPath,
+            vocoderPath,
+            unicodeIndexerPath,
+            ttsConfigPath,
+            voiceStylePath,
         },
     };
 }
@@ -95,51 +103,63 @@ function createChatterboxModel(modelId, config, artifacts) {
     const logger = createStreamLogger(modelId, ModelType.onnxTts);
     registerAddonLogger(modelId, ModelType.onnxTts, logger);
     const referenceAudio = loadReferenceAudioAt24k(referenceAudioPath);
-    const args = {
-        tokenizerPath,
-        speechEncoderPath,
-        embedTokensPath,
-        conditionalDecoderPath,
-        languageModelPath,
+    const model = new ONNXTTS({
+        files: {
+            tokenizerPath,
+            speechEncoderPath,
+            embedTokensPath,
+            conditionalDecoderPath,
+            languageModelPath,
+        },
+        engine: "chatterbox",
+        config: { language: config.language ?? "en", useGPU: false },
         referenceAudio,
         logger,
         opts: { stats: true },
-    };
-    const modelConfig = { language: config.language ?? "en", useGPU: false };
-    const model = new ONNXTTS(args, modelConfig);
+        exclusiveRun: true,
+    });
     return { model, loader: undefined };
 }
 function createSupertonicModel(modelId, config, artifacts) {
-    const tokenizerPath = artifacts["tokenizerPath"];
     const textEncoderPath = artifacts["textEncoderPath"];
-    const latentDenoiserPath = artifacts["latentDenoiserPath"];
-    const voiceDecoderPath = artifacts["voiceDecoderPath"];
-    const voicePath = artifacts["voicePath"];
-    if (!tokenizerPath ||
-        !textEncoderPath ||
-        !latentDenoiserPath ||
-        !voiceDecoderPath ||
-        !voicePath) {
+    const durationPredictorPath = artifacts["durationPredictorPath"];
+    const vectorEstimatorPath = artifacts["vectorEstimatorPath"];
+    const vocoderPath = artifacts["vocoderPath"];
+    const unicodeIndexerPath = artifacts["unicodeIndexerPath"];
+    const ttsConfigPath = artifacts["ttsConfigPath"];
+    const voiceStylePath = artifacts["voiceStylePath"];
+    if (!textEncoderPath ||
+        !durationPredictorPath ||
+        !vectorEstimatorPath ||
+        !vocoderPath ||
+        !unicodeIndexerPath ||
+        !ttsConfigPath ||
+        !voiceStylePath) {
         throw new TtsArtifactsRequiredError();
     }
     const logger = createStreamLogger(modelId, ModelType.onnxTts);
     registerAddonLogger(modelId, ModelType.onnxTts, logger);
-    const voicesDir = path.dirname(voicePath);
-    const voiceName = path.basename(voicePath).replace(/\.bin$/i, "") || "voice";
-    const args = {
-        tokenizerPath,
-        textEncoderPath,
-        latentDenoiserPath,
-        voiceDecoderPath,
-        voicesDir,
+    const voiceName = path.basename(voiceStylePath).replace(/\.json$/i, "") || "F1";
+    const model = new ONNXTTS({
+        files: {
+            textEncoderPath,
+            durationPredictorPath,
+            vectorEstimatorPath,
+            vocoderPath,
+            unicodeIndexerPath,
+            ttsConfigPath,
+            voiceStyleJsonPath: voiceStylePath,
+        },
+        engine: "supertonic",
         voiceName,
         speed: config.ttsSpeed ?? 1,
         numInferenceSteps: config.ttsNumInferenceSteps ?? 5,
+        supertonicMultilingual: config.ttsSupertonicMultilingual !== false,
+        config: { language: config.language ?? "en" },
         logger,
         opts: { stats: true },
-    };
-    const modelConfig = { language: config.language ?? "en" };
-    const model = new ONNXTTS(args, modelConfig);
+        exclusiveRun: true,
+    });
     return { model, loader: undefined };
 }
 export const ttsPlugin = definePlugin({
@@ -169,8 +189,27 @@ export const ttsPlugin = definePlugin({
             responseSchema: ttsResponseSchema,
             streaming: true,
             handler: async function* (request) {
-                for await (const response of textToSpeech(request)) {
-                    yield response;
+                const stream = textToSpeech(request);
+                try {
+                    let result = await stream.next();
+                    while (!result.done) {
+                        yield {
+                            type: "textToSpeech",
+                            buffer: result.value.buffer,
+                            done: false,
+                        };
+                        result = await stream.next();
+                    }
+                    const { modelExecutionMs, stats } = result.value;
+                    yield attachModelExecutionMs({
+                        type: "textToSpeech",
+                        buffer: [],
+                        done: true,
+                        ...(stats && { stats }),
+                    }, modelExecutionMs);
+                }
+                finally {
+                    await stream.return?.(undefined);
                 }
             },
         }),