@speech-sdk/core 0.6.2 → 0.8.0-alpha

package/dist/errors.js.map

@@ -1 +1 @@
-{"version":3,"file":"errors.js","sourceRoot":"","sources":["../src/errors.ts"],"names":[],"mappings":"AAAA,MAAM,OAAO,cAAe,SAAQ,KAAK;IACvC,YAAY,OAAe,EAAE,OAA6B;QACxD,KAAK,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;QACxB,IAAI,CAAC,IAAI,GAAG,gBAAgB,CAAC;IAC/B,CAAC;CACF;AAED,MAAM,OAAO,QAAS,SAAQ,cAAc;IACjC,UAAU,CAAS;IACnB,YAAY,CAAW;IACvB,KAAK,CAAS;IAEvB,YACE,OAAe,EACf,OAKC;QAED,KAAK,CAAC,OAAO,EAAE,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,CAAC,CAAC;QACzC,IAAI,CAAC,IAAI,GAAG,UAAU,CAAC;QACvB,IAAI,CAAC,UAAU,GAAG,OAAO,CAAC,UAAU,CAAC;QACrC,IAAI,CAAC,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC;QAC3B,IAAI,CAAC,YAAY,GAAG,OAAO,CAAC,YAAY,CAAC;IAC3C,CAAC;CACF;AAED,MAAM,OAAO,sBAAuB,SAAQ,cAAc;IACxD,YAAY,OAAgB;QAC1B,KAAK,CAAC,OAAO,IAAI,gCAAgC,CAAC,CAAC;QACnD,IAAI,CAAC,IAAI,GAAG,wBAAwB,CAAC;IACvC,CAAC;CACF;AAED,MAAM,OAAO,0BAA2B,SAAQ,cAAc;IAC5D,YAAY,KAAa;QACvB,KAAK,CACH,iCAAiC,KAAK,iCAAiC,CACxE,CAAC;QACF,IAAI,CAAC,IAAI,GAAG,4BAA4B,CAAC;IAC3C,CAAC;CACF;AAED,MAAM,OAAO,gCAAiC,SAAQ,cAAc;IAClE,YAAY,KAAa;QACvB,KAAK,CACH,kCAAkC,KAAK,gEAAgE,CACxG,CAAC;QACF,IAAI,CAAC,IAAI,GAAG,kCAAkC,CAAC;IACjD,CAAC;CACF"}
+{"version":3,"file":"errors.js","sourceRoot":"","sources":["../src/errors.ts"],"names":[],"mappings":"AAAA,MAAM,OAAO,cAAe,SAAQ,KAAK;IACvC,YAAY,OAAe,EAAE,OAA6B;QACxD,KAAK,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;QACxB,IAAI,CAAC,IAAI,GAAG,gBAAgB,CAAC;IAC/B,CAAC;CACF;AAED,MAAM,OAAO,QAAS,SAAQ,cAAc;IACjC,UAAU,CAAS;IACnB,YAAY,CAAW;IAChC,qEAAqE;IAC5D,IAAI,CAAU;IAEvB,YACE,OAAe,EACf,OAKC;QAED,KAAK,CAAC,OAAO,EAAE,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,CAAC,CAAC;QACzC,IAAI,CAAC,IAAI,GAAG,UAAU,CAAC;QACvB,IAAI,CAAC,UAAU,GAAG,OAAO,CAAC,UAAU,CAAC;QACrC,IAAI,CAAC,YAAY,GAAG,OAAO,CAAC,YAAY,CAAC;QACzC,IAAI,CAAC,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC;IAC3B,CAAC;CACF;AAED,MAAM,OAAO,sBAAuB,SAAQ,cAAc;IACxD,YAAY,OAAgB;QAC1B,KAAK,CAAC,OAAO,IAAI,gCAAgC,CAAC,CAAC;QACnD,IAAI,CAAC,IAAI,GAAG,wBAAwB,CAAC;IACvC,CAAC;CACF;AAED,MAAM,OAAO,0BAA2B,SAAQ,cAAc;IAC5D,YAAY,KAAa;QACvB,KAAK,CACH,iCAAiC,KAAK,iCAAiC,CACxE,CAAC;QACF,IAAI,CAAC,IAAI,GAAG,4BAA4B,CAAC;IAC3C,CAAC;CACF;AAED,MAAM,OAAO,gCAAiC,SAAQ,cAAc;IAClE,YAAY,KAAa;QACvB,KAAK,CACH,kCAAkC,KAAK,gEAAgE,CACxG,CAAC;QACF,IAAI,CAAC,IAAI,GAAG,kCAAkC,CAAC;IACjD,CAAC;CACF;AAED,MAAM,OAAO,iBAAkB,SAAQ,cAAc;IACnD,YAAY,OAAe;QACzB,KAAK,CAAC,OAAO,CAAC,CAAC;QACf,IAAI,CAAC,IAAI,GAAG,mBAAmB,CAAC;IAClC,CAAC;CACF;AAED,MAAM,OAAO,kBAAmB,SAAQ,cAAc;IAC3C,YAAY,CAAS;IACrB,MAAM,CAAS;IAExB,YAAY,OAAiD;QAC3D,KAAK,CACH,GAAG,OAAO,CAAC,YAAY,8DAA8D,OAAO,CAAC,MAAM,wBAAwB,CAC5H,CAAC;QACF,IAAI,CAAC,IAAI,GAAG,oBAAoB,CAAC;QACjC,IAAI,CAAC,YAAY,GAAG,OAAO,CAAC,YAAY,CAAC;QACzC,IAAI,CAAC,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC;IAC/B,CAAC;CACF;AAED,MAAM,OAAO,wBAAyB,SAAQ,cAAc;IAC1D,YAAY,OAIX;QACC,KAAK,CACH,GAAG,OAAO,CAAC,QAAQ,6CAA6C;YAC9D,OAAO,OAAO,CAAC,MAAM,uBAAuB,OAAO,CAAC,WAAW,aAAa;YAC5E,2DAA2D;YAC3D,qGAAqG,CACxG,CAAC;QACF,IAAI,CAAC,IAAI,GAAG,0BAA0B,CAAC;IACzC,CAAC;CACF"}

package/dist/generate-conversation.d.ts

@@ -1,7 +1,8 @@
 import type { GenerateConversationOptions } from "./conversation/types.js";
-import type { Voice } from "./speech-provider.js";
-import type { SpeechResult } from "./speech-result.js";
-export { ConversationInputError, DialogueConstraintError, StitchUnsupportedError, } from "./conversation/errors.js";
+import { type Voice } from "./speech-provider.js";
+import type { ConversationResult } from "./speech-result.js";
+export { ConversationInputError, DialogueConstraintError, MixedDispatchError, StitchUnsupportedError, } from "./conversation/errors.js";
 export type { ConversationTurn, GenerateConversationOptions, } from "./conversation/types.js";
-export declare function generateConversation<V extends Voice = Voice>(options: GenerateConversationOptions<V>): Promise<SpeechResult>;
+export type { ConversationResult } from "./speech-result.js";
+export declare function generateConversation<V extends Voice = Voice>(options: GenerateConversationOptions<V>): Promise<ConversationResult>;
 //# sourceMappingURL=generate-conversation.d.ts.map

package/dist/generate-conversation.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"generate-conversation.d.ts","sourceRoot":"","sources":["../src/generate-conversation.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EAAE,2BAA2B,EAAE,MAAM,yBAAyB,CAAC;AAK3E,OAAO,KAAK,EAAiB,KAAK,EAAE,MAAM,sBAAsB,CAAC;AACjE,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAIvD,OAAO,EACL,sBAAsB,EACtB,uBAAuB,EACvB,sBAAsB,GACvB,MAAM,0BAA0B,CAAC;AAClC,YAAY,EACV,gBAAgB,EAChB,2BAA2B,GAC5B,MAAM,yBAAyB,CAAC;AAMjC,wBAAsB,oBAAoB,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK,EAChE,OAAO,EAAE,2BAA2B,CAAC,CAAC,CAAC,GACtC,OAAO,CAAC,YAAY,CAAC,CAqFvB"}
+{"version":3,"file":"generate-conversation.d.ts","sourceRoot":"","sources":["../src/generate-conversation.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EAEV,2BAA2B,EAC5B,MAAM,yBAAyB,CAAC;AAUjC,OAAO,EAGL,KAAK,KAAK,EACX,MAAM,sBAAsB,CAAC;AAC9B,OAAO,KAAK,EAAE,kBAAkB,EAAE,MAAM,oBAAoB,CAAC;AAK7D,OAAO,EACL,sBAAsB,EACtB,uBAAuB,EACvB,kBAAkB,EAClB,sBAAsB,GACvB,MAAM,0BAA0B,CAAC;AAClC,YAAY,EACV,gBAAgB,EAChB,2BAA2B,GAC5B,MAAM,yBAAyB,CAAC;AACjC,YAAY,EAAE,kBAAkB,EAAE,MAAM,oBAAoB,CAAC;AAM7D,wBAAsB,oBAAoB,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK,EAChE,OAAO,EAAE,2BAA2B,CAAC,CAAC,CAAC,GACtC,OAAO,CAAC,kBAAkB,CAAC,CAwG7B"}

package/dist/generate-conversation.js

@@ -3,34 +3,63 @@ import { computeAudioDuration } from "./audio-duration.js";
 import { chooseConversationPath } from "./conversation/dispatch.js";
 import { ConversationInputError } from "./conversation/errors.js";
 import { validateConversationInput } from "./conversation/validate.js";
+import { getDefaultSTTFallback } from "./default-stt-fallback.js";
+import { deriveTimestampsViaSTT } from "./derive-timestamps.js";
 import { ApiError, NoSpeechGeneratedError } from "./errors.js";
+import { debug } from "./logger.js";
+import { isRetriableApiError } from "./provider-utils.js";
 import { resolveModel } from "./resolve-provider.js";
+import { modelDeclaresNativeTimestamps, } from "./speech-provider.js";
 import { DefaultGeneratedAudioFile } from "./speech-result.js";
 // biome-ignore lint/performance/noBarrelFile: public entry point — re-export error classes so callers get fn + types + errors from one import
-export { ConversationInputError, DialogueConstraintError, StitchUnsupportedError, } from "./conversation/errors.js";
+export { ConversationInputError, DialogueConstraintError, MixedDispatchError, StitchUnsupportedError, } from "./conversation/errors.js";
 const DEFAULT_GAP_MS = 300;
 const DEFAULT_MAX_CONCURRENCY = 6;
 const DEFAULT_MAX_RETRIES = 2;
 export async function generateConversation(options) {
     validateConversationInput(options);
+    // Cache string-model resolutions so turns share one provider instance — dispatch compares by reference.
+    const stringResolutionCache = new Map();
+    const resolveOnce = (model) => {
+        if (typeof model !== "string") {
+            return resolveModel(model, {
+                apiKey: options.apiKey,
+            });
+        }
+        const cached = stringResolutionCache.get(model);
+        if (cached) {
+            return cached;
+        }
+        const fresh = resolveModel(model, {
+            apiKey: options.apiKey,
+        });
+        stringResolutionCache.set(model, fresh);
+        return fresh;
+    };
+    const topLevelResolved = options.model == null ? undefined : resolveOnce(options.model);
     const resolvedPerTurn = options.turns.map((turn) => {
-        const model = turn.model ?? options.model;
+        if (turn.model == null && topLevelResolved) {
+            return topLevelResolved;
+        }
+        const model = turn.model;
         if (!model) {
             throw new Error("generateConversation: model is required");
         }
-        return resolveModel(model, { apiKey: options.apiKey });
+        return resolveOnce(model);
     });
     const path = chooseConversationPath({
         resolvedPerTurn,
         turns: options.turns,
     });
+    if (path.kind === "gateway") {
+        return await runGateway({
+            options,
+            resolvedPerTurn: path.resolvedPerTurn,
+            maxRetries: options.maxRetries ?? DEFAULT_MAX_RETRIES,
+        });
+    }
     if (path.kind === "native") {
-        // The native-dialogue path renders the entire script in a single provider
-        // API call, so per-turn providerOptions have no well-defined meaning —
-        // silently collapsing them to a single blob would lie to the caller. Fail
-        // loudly and let them move providerOptions to the top level (where it's
-        // forwarded once to the dialogue call) or pick a model that routes
-        // through the stitch path.
+        // Native-dialogue is one API call; per-turn providerOptions can't be honored — fail loudly.
         const turnWithOpts = options.turns.findIndex((t) => t.providerOptions !== undefined);
         if (turnWithOpts !== -1) {
             throw new ConversationInputError(`turns[${turnWithOpts}].providerOptions is set, but ${path.resolved.provider.id}/${path.resolved.modelId} dispatched to the native dialogue path, which renders all turns in one API call. Per-turn providerOptions are not supported on this path; move them to the top-level providerOptions instead.`);
@@ -41,9 +70,7 @@ export async function generateConversation(options) {
             maxRetries: options.maxRetries ?? DEFAULT_MAX_RETRIES,
         });
     }
-    // Lazy-load the stitch pipeline so callers whose dispatch always picks
-    // native (e.g. a Jellypod gateway provider that handles concatenation
-    // server-side) never bundle pcm-concat / audio-utils / mediabunny WAV mux.
+    // Lazy-load so native-only callers don't bundle pcm-concat / mediabunny.
     const { runStitch } = await import("./conversation/stitch.js");
     const stitched = await runStitch({
         resolvedPerTurn,
@@ -54,21 +81,17 @@ export async function generateConversation(options) {
         gapMs: options.gapMs ?? DEFAULT_GAP_MS,
         maxConcurrency: options.maxConcurrency ?? DEFAULT_MAX_CONCURRENCY,
         maxRetries: options.maxRetries ?? DEFAULT_MAX_RETRIES,
-        normalizeVolume: options.normalizeVolume ?? true,
         volumeDbfs: options.volumeDbfs,
         abortSignal: options.abortSignal,
         headers: options.headers,
+        timestamps: options.timestamps ?? false,
     });
     if (stitched.audio.length === 0) {
         throw new NoSpeechGeneratedError();
     }
-    const providers = Array.from(new Set(resolvedPerTurn.map((r) => r.provider.id)));
-    const models = Array.from(new Set(resolvedPerTurn.map((r) => r.modelId)));
     const metadata = {
         latencyMs: stitched.metadata.latencyMs,
         inputChars: stitched.metadata.inputChars,
-        provider: providers.length === 1 ? providers[0] : providers.join(","),
-        model: models.length === 1 ? models[0] : models.join(","),
         ...(stitched.metadata.audioDurationMs != null && {
             audioDurationMs: stitched.metadata.audioDurationMs,
         }),
@@ -81,6 +104,76 @@ export async function generateConversation(options) {
         metadata,
         providerMetadata: { turns: stitched.providerMetadataPerTurn },
         warnings: stitched.warnings.length > 0 ? [...stitched.warnings] : undefined,
+        timestamps: stitched.timestamps,
+    };
+}
+async function runGateway(args) {
+    const { options, resolvedPerTurn, maxRetries } = args;
+    const start = performance.now();
+    const provider = resolvedPerTurn[0]
+        .provider;
+    const includeTimestamps = options.timestamps ?? false;
+    // Pick wire shape: shared model when every turn resolved to the same modelId
+    // (covers both `options.model` and the case where every turn happened to
+    // resolve identically); per-turn model when they diverge.
+    const firstModelId = resolvedPerTurn[0].modelId;
+    const allSameModel = resolvedPerTurn.every((r) => r.modelId === firstModelId);
+    // Object-shaped voices aren't supported on the gateway conversation path.
+    const wireTurns = options.turns.map((t, i) => {
+        if (typeof t.voice !== "string") {
+            throw new Error(`speech-gateway/conversation: gateway conversation path requires string voices; turns[${i}].voice is an object.`);
+        }
+        return {
+            ...(allSameModel ? {} : { model: resolvedPerTurn[i].modelId }),
+            voice: t.voice,
+            text: t.text,
+            ...(t.providerOptions && { providerOptions: t.providerOptions }),
+        };
+    });
+    const result = await pRetry(() => provider.generateConversation({
+        ...(allSameModel && { modelId: firstModelId }),
+        turns: wireTurns,
+        gapMs: options.gapMs ?? DEFAULT_GAP_MS,
+        volumeDbfs: options.volumeDbfs,
+        providerOptions: options.providerOptions,
+        abortSignal: options.abortSignal,
+        headers: options.headers,
+        includeTimestamps,
+    }), {
+        retries: maxRetries,
+        signal: options.abortSignal,
+        shouldRetry: ({ error }) => {
+            if (error instanceof ApiError && !isRetriableApiError(error)) {
+                return false;
+            }
+            return true;
+        },
+    });
+    const latencyMs = Math.round(performance.now() - start);
+    if (result.audio.length === 0) {
+        throw new NoSpeechGeneratedError();
+    }
+    const audio = new DefaultGeneratedAudioFile({
+        data: result.audio,
+        mediaType: result.mediaType,
+    });
+    const audioDurationMs = await computeAudioDuration(audio.uint8Array, result.mediaType);
+    const timestamps = includeTimestamps ? (result.timestamps ?? []) : undefined;
+    const warnings = result.warnings;
+    const inputChars = options.turns.reduce((n, t) => n + t.text.length, 0);
+    const metadata = {
+        latencyMs,
+        inputChars,
+        ...(audioDurationMs != null && { audioDurationMs }),
+    };
+    return {
+        audio,
+        metadata,
+        ...(result.providerMetadata !== undefined && {
+            providerMetadata: result.providerMetadata,
+        }),
+        timestamps,
+        warnings: warnings && warnings.length > 0 ? warnings : undefined,
     };
 }
 async function runNative(args) {
@@ -90,37 +183,41 @@ async function runNative(args) {
         throw new Error(`generateConversation: ${resolved.provider.id}/${resolved.modelId} dispatched to native but generateDialogue missing`);
     }
     const generateDialogue = resolved.provider.generateDialogue.bind(resolved.provider);
-    // When normalization is requested and the provider exposes a decodable
-    // PCM/WAV mode via getStitchOptions, force the dialogue request into that
-    // mode so we can re-RMS-level the output. Otherwise the dialogue runs
-    // unchanged and emerges in whatever format the provider mixes natively
-    // (often MP3) — we surface that via a warning.
-    const normalize = options.normalizeVolume ?? true;
-    const stitchOpts = normalize
-        ? resolved.provider.getStitchOptions?.(resolved.modelId)
-        : undefined;
+    // Force decodable PCM/WAV via getStitchOptions for normalization; if unavailable, emit the provider's mixed audio and warn.
+    const stitchOpts = resolved.provider.getStitchOptions?.(resolved.modelId);
     const warnings = [];
-    if (normalize && !stitchOpts) {
-        warnings.push(`${resolved.provider.id}/${resolved.modelId}: native dialogue path returns the provider's mixed audio without volume normalization. Pass normalizeVolume:false to silence this warning.`);
+    if (!stitchOpts) {
+        warnings.push(`${resolved.provider.id}/${resolved.modelId}: native dialogue path returns the provider's mixed audio without volume normalization (no decodable PCM/WAV mode).`);
     }
-    // Stitch-mode options are applied last so they override user-supplied
-    // providerOptions that would otherwise break the decoder (e.g. a caller
-    // requesting `response_format: "mp3"` while normalization is on). Same
-    // precedence as the stitch path's per-turn merge.
+    // Stitch options must win — caller-supplied response_format would break the decoder.
     const dialogueProviderOptions = stitchOpts
         ? { ...options.providerOptions, ...stitchOpts.providerOptions }
         : options.providerOptions;
+    const requestTimestamps = options.timestamps ?? false;
+    const hasNativeDialogueTimestamps = modelDeclaresNativeTimestamps(resolved);
+    const shouldRequestNative = requestTimestamps && hasNativeDialogueTimestamps;
+    const dialogueId = `${resolved.provider.id}/${resolved.modelId}`;
+    if (!requestTimestamps) {
+        debug(`${dialogueId} (dialogue): timestamps: false — skipping alignment.`);
+    }
+    else if (shouldRequestNative) {
+        debug(`${dialogueId} (dialogue): timestamps: true — requesting native dialogue alignment.`);
+    }
+    else {
+        debug(`${dialogueId} (dialogue): timestamps: true but no native dialogue alignment — will transcribe mixed audio via STT after rendering (adds a round-trip).`);
+    }
     const result = await pRetry(() => generateDialogue({
         modelId: resolved.modelId,
         turns: options.turns.map((t) => ({ voice: t.voice, text: t.text })),
         providerOptions: dialogueProviderOptions,
         abortSignal: options.abortSignal,
         headers: options.headers,
+        includeTimestamps: shouldRequestNative,
     }), {
         retries: maxRetries,
         signal: options.abortSignal,
         shouldRetry: ({ error }) => {
-            if (error instanceof ApiError && error.statusCode < 500) {
+            if (error instanceof ApiError && !isRetriableApiError(error)) {
                 return false;
             }
             return true;
@@ -131,8 +228,7 @@ async function runNative(args) {
         throw new NoSpeechGeneratedError();
     }
     let audioBytes = result.audio;
-    // Prefer the stitch-mode mediaType over the provider's response header;
-    // some providers (e.g. Hume) omit the sample rate from content-type.
+    // Hume and others omit sample rate from content-type; prefer stitch mediaType.
     let outputMediaType = stitchOpts?.mediaType ?? result.mediaType;
     if (stitchOpts) {
         const { adjustVolume } = await import("./volume-adjust.js");
@@ -149,19 +245,76 @@ async function runNative(args) {
     });
     const computedDuration = await computeAudioDuration(audio.uint8Array, outputMediaType);
     const audioDurationMs = computedDuration ?? result.audioDurationMs;
+    const { timestamps, warnings: attributionWarnings } = await resolveNativeDialogueTimestamps({
+        requestTimestamps,
+        nativeTimestamps: result.timestamps,
+        audio: audio.uint8Array,
+        mediaType: outputMediaType,
+        ttsModel: `${resolved.provider.id}/${resolved.modelId}`,
+        resolved,
+        abortSignal: options.abortSignal,
+        turns: options.turns,
+    });
     const inputChars = options.turns.reduce((n, t) => n + t.text.length, 0);
     const metadata = {
         latencyMs,
         inputChars,
-        provider: resolved.provider.id,
-        model: resolved.modelId,
         ...(audioDurationMs != null && { audioDurationMs }),
     };
+    const mergedWarnings = warnings.length > 0 || attributionWarnings.length > 0
+        ? [...warnings, ...attributionWarnings]
+        : undefined;
     return {
         audio,
         metadata,
         providerMetadata: result.providerMetadata,
-        warnings: warnings.length > 0 ? warnings : undefined,
+        warnings: mergedWarnings,
+        timestamps,
+    };
+}
+async function resolveNativeDialogueTimestamps(args) {
+    if (!args.requestTimestamps) {
+        return { timestamps: undefined, warnings: [] };
+    }
+    // Either use native flat timestamps, or derive via STT fallback.
+    let flatTimestamps;
+    if (args.nativeTimestamps && args.nativeTimestamps.length > 0) {
+        flatTimestamps = args.nativeTimestamps;
+    }
+    else {
+        const fallback = args.resolved.fallbackSTT ?? (await getDefaultSTTFallback());
+        flatTimestamps = await deriveTimestampsViaSTT({
+            ttsModel: args.ttsModel,
+            audio: args.audio,
+            mediaType: args.mediaType,
+            timestampFallback: fallback,
+            abortSignal: args.abortSignal,
+        });
+    }
+    const { decodeToPcm16 } = await import("./conversation/pcm-concat.js");
+    const { detectSilenceGaps } = await import("./conversation/silence-detection.js");
+    const { attributeTimestamps } = await import("./conversation/attribute-timestamps.js");
+    let silenceGaps = [];
+    try {
+        const segment = decodeToPcm16(args.audio, args.mediaType);
+        const gaps = detectSilenceGaps(segment.pcm, {
+            sampleRate: segment.sampleRate,
+            minDurationMs: 150,
+        });
+        silenceGaps = gaps;
+    }
+    catch {
+        // Decoder couldn't read the audio (e.g., compressed format we can't decode locally).
+        // Tier 1 will be skipped; dispatcher falls through to Tier 2/3.
+    }
+    const result = attributeTimestamps({
+        timestamps: flatTimestamps,
+        turnTexts: args.turns.map((t) => t.text),
+        silenceGaps,
+    });
+    return {
+        timestamps: result.timestamps,
+        warnings: result.warnings,
     };
 }
 //# sourceMappingURL=generate-conversation.js.map

package/dist/generate-conversation.js.map

@@ -1 +1 @@
-{"version":3,"file":"generate-conversation.js","sourceRoot":"","sources":["../src/generate-conversation.ts"],"names":[],"mappings":"AAAA,OAAO,MAAM,MAAM,SAAS,CAAC;AAC7B,OAAO,EAAE,oBAAoB,EAAE,MAAM,qBAAqB,CAAC;AAC3D,OAAO,EAAE,sBAAsB,EAAE,MAAM,4BAA4B,CAAC;AACpE,OAAO,EAAE,sBAAsB,EAAE,MAAM,0BAA0B,CAAC;AAElE,OAAO,EAAE,yBAAyB,EAAE,MAAM,4BAA4B,CAAC;AACvE,OAAO,EAAE,QAAQ,EAAE,sBAAsB,EAAE,MAAM,aAAa,CAAC;AAE/D,OAAO,EAAE,YAAY,EAAE,MAAM,uBAAuB,CAAC;AAGrD,OAAO,EAAE,yBAAyB,EAAE,MAAM,oBAAoB,CAAC;AAE/D,8IAA8I;AAC9I,OAAO,EACL,sBAAsB,EACtB,uBAAuB,EACvB,sBAAsB,GACvB,MAAM,0BAA0B,CAAC;AAMlC,MAAM,cAAc,GAAG,GAAG,CAAC;AAC3B,MAAM,uBAAuB,GAAG,CAAC,CAAC;AAClC,MAAM,mBAAmB,GAAG,CAAC,CAAC;AAE9B,MAAM,CAAC,KAAK,UAAU,oBAAoB,CACxC,OAAuC;IAEvC,yBAAyB,CAAC,OAAO,CAAC,CAAC;IAEnC,MAAM,eAAe,GAAuB,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE;QACrE,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,IAAI,OAAO,CAAC,KAAK,CAAC;QAC1C,IAAI,CAAC,KAAK,EAAE,CAAC;YACX,MAAM,IAAI,KAAK,CAAC,yCAAyC,CAAC,CAAC;QAC7D,CAAC;QACD,OAAO,YAAY,CAAC,KAAK,EAAE,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAqB,CAAC;IAC7E,CAAC,CAAC,CAAC;IAEH,MAAM,IAAI,GAAG,sBAAsB,CAAC;QAClC,eAAe;QACf,KAAK,EAAE,OAAO,CAAC,KAAK;KACrB,CAAC,CAAC;IAEH,IAAI,IAAI,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;QAC3B,0EAA0E;QAC1E,uEAAuE;QACvE,0EAA0E;QAC1E,wEAAwE;QACxE,mEAAmE;QACnE,2BAA2B;QAC3B,MAAM,YAAY,GAAG,OAAO,CAAC,KAAK,CAAC,SAAS,CAC1C,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,eAAe,KAAK,SAAS,CACvC,CAAC;QACF,IAAI,YAAY,KAAK,CAAC,CAAC,EAAE,CAAC;YACxB,MAAM,IAAI,sBAAsB,CAC9B,SAAS,YAAY,iCAAiC,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,IAAI,CAAC,QAAQ,CAAC,OAAO,gMAAgM,CACzS,CAAC;QACJ,CAAC;QACD,OAAO,MAAM,SAAS,CAAC;YACrB,OAAO;YACP,QAAQ,EAAE,IAAI,CAAC,QAAQ;YACvB,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,mBAAmB;SACtD,CAAC,CAAC;IACL,CAAC;IAED,uEAAuE;IACvE,sEAAsE;IACtE,2EAA2E;IAC3E,MAAM,EAAE,SAAS,EAAE,GAAG,MAAM,MAAM,CAAC,0BAA0B,CAAC,CAAC;IAC/D,MAAM,QAAQ,GAAG,MAAM,SAAS,CAAC;QAC/B,eAAe;QACf,KAAK,EAAE,OAAO,CAAC,KAAK;QACpB,oBAAoB,EAAE,IAAI,CAAC,oBAAoB;QAC/C,uBAAuB,EAAE,OAAO,CAAC,eAAe;QAChD,MAAM,EAAE,OAAO,CAAC,MAAM;QACtB,KAAK,EAAE,OAAO,CAAC,KAAK,IAAI,cAAc;QACtC,cAAc,EAAE,OAAO,CAAC,cAAc,IAAI,uBAAuB;QACjE,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,mBAAmB;QACrD,eAAe,EAAE,OAAO,CAAC,eAAe,IAAI,IAAI;QAChD,UAAU,EAAE,OAAO,CAAC,UAAU;QAC9B,WAAW,EAAE,OAAO,CAAC,WAAW;QAChC,OAAO,EAAE,OAAO,CAAC,OAAO;KACzB,CAAC,CAAC;IAEH,IAAI,QAAQ,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAChC,MAAM,IAAI,sBAAsB,EAAE,CAAC;IACrC,CAAC;IAED,MAAM,SAAS,GAAG,KAAK,CAAC,IAAI,CAC1B,IAAI,GAAG,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,CACnD,CAAC;IACF,MAAM,MAAM,GAAG,KAAK,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;IAE1E,MAAM,QAAQ,GAAmB;QAC/B,SAAS,EAAE,QAAQ,CAAC,QAAQ,CAAC,SAAS;QACtC,UAAU,EAAE,QAAQ,CAAC,QAAQ,CAAC,UAAU;QACxC,QAAQ,EAAE,SAAS,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,IAAI,CAAC,GAAG,CAAC;QACrE,KAAK,EAAE,MAAM,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC;QACzD,GAAG,CAAC,QAAQ,CAAC,QAAQ,CAAC,eAAe,IAAI,IAAI,IAAI;YAC/C,eAAe,EAAE,QAAQ,CAAC,QAAQ,CAAC,eAAe;SACnD,CAAC;KACH,CAAC;IAEF,OAAO;QACL,KAAK,EAAE,IAAI,yBAAyB,CAAC;YACnC,IAAI,EAAE,QAAQ,CAAC,KAAK;YACpB,SAAS,EAAE,QAAQ,CAAC,SAAS;SAC9B,CAAC;QACF,QAAQ;QACR,gBAAgB,EAAE,EAAE,KAAK,EAAE,QAAQ,CAAC,uBAAuB,EAAE;QAC7D,QAAQ,EAAE,QAAQ,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,QAAQ,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,SAAS;KAC5E,CAAC;AACJ,CAAC;AAED,KAAK,UAAU,SAAS,CAAkB,IAIzC;IACC,MAAM,EAAE,OAAO,EAAE,QAAQ,EAAE,UAAU,EAAE,GAAG,IAAI,CAAC;IAC/C,MAAM,KAAK,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;IAEhC,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,gBAAgB,EAAE,CAAC;QACxC,MAAM,IAAI,KAAK,CACb,yBAAyB,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,oDAAoD,CACtH,CAAC;IACJ,CAAC;IAED,MAAM,gBAAgB,GAAG,QAAQ,CAAC,QAAQ,CAAC,gBAAgB,CAAC,IAAI,CAC9D,QAAQ,CAAC,QAAQ,CAClB,CAAC;IAEF,uEAAuE;IACvE,0EAA0E;IAC1E,sEAAsE;IACtE,uEAAuE;IACvE,+CAA+C;IAC/C,MAAM,SAAS,GAAG,OAAO,CAAC,eAAe,IAAI,IAAI,CAAC;IAClD,MAAM,UAAU,GAAG,SAAS;QAC1B,CAAC,CAAC,QAAQ,CAAC,QAAQ,CAAC,gBAAgB,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC;QACxD,CAAC,CAAC,SAAS,CAAC;IACd,MAAM,QAAQ,GAAa,EAAE,CAAC;IAC9B,IAAI,SAAS,IAAI,CAAC,UAAU,EAAE,CAAC;QAC7B,QAAQ,CAAC,IAAI,CACX,GAAG,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,6IAA6I,CACzL,CAAC;IACJ,CAAC;IAED,sEAAsE;IACtE,wEAAwE;IACxE,uEAAuE;IACvE,kDAAkD;IAClD,MAAM,uBAAuB,GAAG,UAAU;QACxC,CAAC,CAAC,EAAE,GAAG,OAAO,CAAC,eAAe,EAAE,GAAG,UAAU,CAAC,eAAe,EAAE;QAC/D,CAAC,CAAC,OAAO,CAAC,eAAe,CAAC;IAE5B,MAAM,MAAM,GAAG,MAAM,MAAM,CACzB,GAAG,EAAE,CACH,gBAAgB,CAAC;QACf,OAAO,EAAE,QAAQ,CAAC,OAAO;QACzB,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC;QACnE,eAAe,EAAE,uBAAuB;QACxC,WAAW,EAAE,OAAO,CAAC,WAAW;QAChC,OAAO,EAAE,OAAO,CAAC,OAAO;KACzB,CAAC,EACJ;QACE,OAAO,EAAE,UAAU;QACnB,MAAM,EAAE,OAAO,CAAC,WAAW;QAC3B,WAAW,EAAE,CAAC,EAAE,KAAK,EAAE,EAAE,EAAE;YACzB,IAAI,KAAK,YAAY,QAAQ,IAAI,KAAK,CAAC,UAAU,GAAG,GAAG,EAAE,CAAC;gBACxD,OAAO,KAAK,CAAC;YACf,CAAC;YACD,OAAO,IAAI,CAAC;QACd,CAAC;KACF,CACF,CAAC;IAEF,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC,CAAC;IAExD,IAAI,MAAM,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC9B,MAAM,IAAI,sBAAsB,EAAE,CAAC;IACrC,CAAC;IAED,IAAI,UAAU,GAAwB,MAAM,CAAC,KAAK,CAAC;IACnD,wEAAwE;IACxE,qEAAqE;IACrE,IAAI,eAAe,GAAG,UAAU,EAAE,SAAS,IAAI,MAAM,CAAC,SAAS,CAAC;IAEhE,IAAI,UAAU,EAAE,CAAC;QACf,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,MAAM,CAAC,oBAAoB,CAAC,CAAC;QAC5D,UAAU,GAAG,MAAM,YAAY,CAAC;YAC9B,KAAK,EAAE,MAAM,CAAC,KAAK;YACnB,SAAS,EAAE,UAAU,CAAC,SAAS;YAC/B,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,CAAC,EAAE;SACtC,CAAC,CAAC;QACH,eAAe,GAAG,WAAW,CAAC;IAChC,CAAC;IAED,MAAM,KAAK,GAAG,IAAI,yBAAyB,CAAC;QAC1C,IAAI,EAAE,UAAU;QAChB,SAAS,EAAE,eAAe;KAC3B,CAAC,CAAC;IAEH,MAAM,gBAAgB,GAAG,MAAM,oBAAoB,CACjD,KAAK,CAAC,UAAU,EAChB,eAAe,CAChB,CAAC;IACF,MAAM,eAAe,GAAG,gBAAgB,IAAI,MAAM,CAAC,eAAe,CAAC;IAEnE,MAAM,UAAU,GAAG,OAAO,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IAExE,MAAM,QAAQ,GAAmB;QAC/B,SAAS;QACT,UAAU;QACV,QAAQ,EAAE,QAAQ,CAAC,QAAQ,CAAC,EAAE;QAC9B,KAAK,EAAE,QAAQ,CAAC,OAAO;QACvB,GAAG,CAAC,eAAe,IAAI,IAAI,IAAI,EAAE,eAAe,EAAE,CAAC;KACpD,CAAC;IAEF,OAAO;QACL,KAAK;QACL,QAAQ;QACR,gBAAgB,EAAE,MAAM,CAAC,gBAAgB;QACzC,QAAQ,EAAE,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,SAAS;KACrD,CAAC;AACJ,CAAC"}
+{"version":3,"file":"generate-conversation.js","sourceRoot":"","sources":["../src/generate-conversation.ts"],"names":[],"mappings":"AAAA,OAAO,MAAM,MAAM,SAAS,CAAC;AAC7B,OAAO,EAAE,oBAAoB,EAAE,MAAM,qBAAqB,CAAC;AAC3D,OAAO,EAAE,sBAAsB,EAAE,MAAM,4BAA4B,CAAC;AACpE,OAAO,EAAE,sBAAsB,EAAE,MAAM,0BAA0B,CAAC;AAKlE,OAAO,EAAE,yBAAyB,EAAE,MAAM,4BAA4B,CAAC;AACvE,OAAO,EAAE,qBAAqB,EAAE,MAAM,2BAA2B,CAAC;AAClE,OAAO,EAAE,sBAAsB,EAAE,MAAM,wBAAwB,CAAC;AAChE,OAAO,EAAE,QAAQ,EAAE,sBAAsB,EAAE,MAAM,aAAa,CAAC;AAC/D,OAAO,EAAE,KAAK,EAAE,MAAM,aAAa,CAAC;AAEpC,OAAO,EAAE,mBAAmB,EAAE,MAAM,qBAAqB,CAAC;AAE1D,OAAO,EAAE,YAAY,EAAE,MAAM,uBAAuB,CAAC;AACrD,OAAO,EACL,6BAA6B,GAG9B,MAAM,sBAAsB,CAAC;AAE9B,OAAO,EAAE,yBAAyB,EAAE,MAAM,oBAAoB,CAAC;AAG/D,8IAA8I;AAC9I,OAAO,EACL,sBAAsB,EACtB,uBAAuB,EACvB,kBAAkB,EAClB,sBAAsB,GACvB,MAAM,0BAA0B,CAAC;AAOlC,MAAM,cAAc,GAAG,GAAG,CAAC;AAC3B,MAAM,uBAAuB,GAAG,CAAC,CAAC;AAClC,MAAM,mBAAmB,GAAG,CAAC,CAAC;AAE9B,MAAM,CAAC,KAAK,UAAU,oBAAoB,CACxC,OAAuC;IAEvC,yBAAyB,CAAC,OAAO,CAAC,CAAC;IAEnC,wGAAwG;IACxG,MAAM,qBAAqB,GAAG,IAAI,GAAG,EAA4B,CAAC;IAClE,MAAM,WAAW,GAAG,CAAC,KAAgC,EAAoB,EAAE;QACzE,IAAI,OAAO,KAAK,KAAK,QAAQ,EAAE,CAAC;YAC9B,OAAO,YAAY,CAAC,KAAK,EAAE;gBACzB,MAAM,EAAE,OAAO,CAAC,MAAM;aACvB,CAAqB,CAAC;QACzB,CAAC;QACD,MAAM,MAAM,GAAG,qBAAqB,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;QAChD,IAAI,MAAM,EAAE,CAAC;YACX,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,MAAM,KAAK,GAAG,YAAY,CAAC,KAAK,EAAE;YAChC,MAAM,EAAE,OAAO,CAAC,MAAM;SACvB,CAAqB,CAAC;QACvB,qBAAqB,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;QACxC,OAAO,KAAK,CAAC;IACf,CAAC,CAAC;IAEF,MAAM,gBAAgB,GACpB,OAAO,CAAC,KAAK,IAAI,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,WAAW,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;IACjE,MAAM,eAAe,GAAuB,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE;QACrE,IAAI,IAAI,CAAC,KAAK,IAAI,IAAI,IAAI,gBAAgB,EAAE,CAAC;YAC3C,OAAO,gBAAgB,CAAC;QAC1B,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC;QACzB,IAAI,CAAC,KAAK,EAAE,CAAC;YACX,MAAM,IAAI,KAAK,CAAC,yCAAyC,CAAC,CAAC;QAC7D,CAAC;QACD,OAAO,WAAW,CAAC,KAAK,CAAC,CAAC;IAC5B,CAAC,CAAC,CAAC;IAEH,MAAM,IAAI,GAAG,sBAAsB,CAAC;QAClC,eAAe;QACf,KAAK,EAAE,OAAO,CAAC,KAAK;KACrB,CAAC,CAAC;IAEH,IAAI,IAAI,CAAC,IAAI,KAAK,SAAS,EAAE,CAAC;QAC5B,OAAO,MAAM,UAAU,CAAC;YACtB,OAAO;YACP,eAAe,EAAE,IAAI,CAAC,eAAe;YACrC,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,mBAAmB;SACtD,CAAC,CAAC;IACL,CAAC;IAED,IAAI,IAAI,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;QAC3B,4FAA4F;QAC5F,MAAM,YAAY,GAAG,OAAO,CAAC,KAAK,CAAC,SAAS,CAC1C,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,eAAe,KAAK,SAAS,CACvC,CAAC;QACF,IAAI,YAAY,KAAK,CAAC,CAAC,EAAE,CAAC;YACxB,MAAM,IAAI,sBAAsB,CAC9B,SAAS,YAAY,iCAAiC,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,IAAI,CAAC,QAAQ,CAAC,OAAO,gMAAgM,CACzS,CAAC;QACJ,CAAC;QACD,OAAO,MAAM,SAAS,CAAC;YACrB,OAAO;YACP,QAAQ,EAAE,IAAI,CAAC,QAAQ;YACvB,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,mBAAmB;SACtD,CAAC,CAAC;IACL,CAAC;IAED,yEAAyE;IACzE,MAAM,EAAE,SAAS,EAAE,GAAG,MAAM,MAAM,CAAC,0BAA0B,CAAC,CAAC;IAC/D,MAAM,QAAQ,GAAG,MAAM,SAAS,CAAC;QAC/B,eAAe;QACf,KAAK,EAAE,OAAO,CAAC,KAAK;QACpB,oBAAoB,EAAE,IAAI,CAAC,oBAAoB;QAC/C,uBAAuB,EAAE,OAAO,CAAC,eAAe;QAChD,MAAM,EAAE,OAAO,CAAC,MAAM;QACtB,KAAK,EAAE,OAAO,CAAC,KAAK,IAAI,cAAc;QACtC,cAAc,EAAE,OAAO,CAAC,cAAc,IAAI,uBAAuB;QACjE,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,mBAAmB;QACrD,UAAU,EAAE,OAAO,CAAC,UAAU;QAC9B,WAAW,EAAE,OAAO,CAAC,WAAW;QAChC,OAAO,EAAE,OAAO,CAAC,OAAO;QACxB,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,KAAK;KACxC,CAAC,CAAC;IAEH,IAAI,QAAQ,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAChC,MAAM,IAAI,sBAAsB,EAAE,CAAC;IACrC,CAAC;IAED,MAAM,QAAQ,GAAmB;QAC/B,SAAS,EAAE,QAAQ,CAAC,QAAQ,CAAC,SAAS;QACtC,UAAU,EAAE,QAAQ,CAAC,QAAQ,CAAC,UAAU;QACxC,GAAG,CAAC,QAAQ,CAAC,QAAQ,CAAC,eAAe,IAAI,IAAI,IAAI;YAC/C,eAAe,EAAE,QAAQ,CAAC,QAAQ,CAAC,eAAe;SACnD,CAAC;KACH,CAAC;IAEF,OAAO;QACL,KAAK,EAAE,IAAI,yBAAyB,CAAC;YACnC,IAAI,EAAE,QAAQ,CAAC,KAAK;YACpB,SAAS,EAAE,QAAQ,CAAC,SAAS;SAC9B,CAAC;QACF,QAAQ;QACR,gBAAgB,EAAE,EAAE,KAAK,EAAE,QAAQ,CAAC,uBAAuB,EAAE;QAC7D,QAAQ,EAAE,QAAQ,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,QAAQ,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,SAAS;QAC3E,UAAU,EAAE,QAAQ,CAAC,UAAU;KAChC,CAAC;AACJ,CAAC;AAED,KAAK,UAAU,UAAU,CAAkB,IAI1C;IACC,MAAM,EAAE,OAAO,EAAE,eAAe,EAAE,UAAU,EAAE,GAAG,IAAI,CAAC;IACtD,MAAM,KAAK,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;IAEhC,MAAM,QAAQ,GAAG,eAAe,CAAC,CAAC,CAAC;SAChC,QAA4C,CAAC;IAEhD,MAAM,iBAAiB,GAAG,OAAO,CAAC,UAAU,IAAI,KAAK,CAAC;IAEtD,6EAA6E;IAC7E,yEAAyE;IACzE,0DAA0D;IAC1D,MAAM,YAAY,GAAG,eAAe,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC;IAChD,MAAM,YAAY,GAAG,eAAe,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,KAAK,YAAY,CAAC,CAAC;IAE9E,0EAA0E;IAC1E,MAAM,SAAS,GAAG,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE;QAC3C,IAAI,OAAO,CAAC,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;YAChC,MAAM,IAAI,KAAK,CACb,wFAAwF,CAAC,uBAAuB,CACjH,CAAC;QACJ,CAAC;QACD,OAAO;YACL,GAAG,CAAC,YAAY,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,eAAe,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,CAAC;YAC9D,KAAK,EAAE,CAAC,CAAC,KAAK;YACd,IAAI,EAAE,CAAC,CAAC,IAAI;YACZ,GAAG,CAAC,CAAC,CAAC,eAAe,IAAI,EAAE,eAAe,EAAE,CAAC,CAAC,eAAe,EAAE,CAAC;SACjE,CAAC;IACJ,CAAC,CAAC,CAAC;IAEH,MAAM,MAAM,GAAG,MAAM,MAAM,CACzB,GAAG,EAAE,CACH,QAAQ,CAAC,oBAAoB,CAAC;QAC5B,GAAG,CAAC,YAAY,IAAI,EAAE,OAAO,EAAE,YAAY,EAAE,CAAC;QAC9C,KAAK,EAAE,SAAS;QAChB,KAAK,EAAE,OAAO,CAAC,KAAK,IAAI,cAAc;QACtC,UAAU,EAAE,OAAO,CAAC,UAAU;QAC9B,eAAe,EAAE,OAAO,CAAC,eAAe;QACxC,WAAW,EAAE,OAAO,CAAC,WAAW;QAChC,OAAO,EAAE,OAAO,CAAC,OAAO;QACxB,iBAAiB;KAClB,CAAC,EACJ;QACE,OAAO,EAAE,UAAU;QACnB,MAAM,EAAE,OAAO,CAAC,WAAW;QAC3B,WAAW,EAAE,CAAC,EAAE,KAAK,EAAE,EAAE,EAAE;YACzB,IAAI,KAAK,YAAY,QAAQ,IAAI,CAAC,mBAAmB,CAAC,KAAK,CAAC,EAAE,CAAC;gBAC7D,OAAO,KAAK,CAAC;YACf,CAAC;YACD,OAAO,IAAI,CAAC;QACd,CAAC;KACF,CACF,CAAC;IAEF,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC,CAAC;IAExD,IAAI,MAAM,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC9B,MAAM,IAAI,sBAAsB,EAAE,CAAC;IACrC,CAAC;IAED,MAAM,KAAK,GAAG,IAAI,yBAAyB,CAAC;QAC1C,IAAI,EAAE,MAAM,CAAC,KAAK;QAClB,SAAS,EAAE,MAAM,CAAC,SAAS;KAC5B,CAAC,CAAC;IACH,MAAM,eAAe,GAAG,MAAM,oBAAoB,CAChD,KAAK,CAAC,UAAU,EAChB,MAAM,CAAC,SAAS,CACjB,CAAC;IAEF,MAAM,UAAU,GACd,iBAAiB,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,UAAU,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAC5D,MAAM,QAAQ,GAAG,MAAM,CAAC,QAAQ,CAAC;IAEjC,MAAM,UAAU,GAAG,OAAO,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IAExE,MAAM,QAAQ,GAAmB;QAC/B,SAAS;QACT,UAAU;QACV,GAAG,CAAC,eAAe,IAAI,IAAI,IAAI,EAAE,eAAe,EAAE,CAAC;KACpD,CAAC;IAEF,OAAO;QACL,KAAK;QACL,QAAQ;QACR,GAAG,CAAC,MAAM,CAAC,gBAAgB,KAAK,SAAS,IAAI;YAC3C,gBAAgB,EAAE,MAAM,CAAC,gBAAgB;SAC1C,CAAC;QACF,UAAU;QACV,QAAQ,EAAE,QAAQ,IAAI,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,SAAS;KACjE,CAAC;AACJ,CAAC;AAED,KAAK,UAAU,SAAS,CAAkB,IAIzC;IACC,MAAM,EAAE,OAAO,EAAE,QAAQ,EAAE,UAAU,EAAE,GAAG,IAAI,CAAC;IAC/C,MAAM,KAAK,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;IAEhC,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,gBAAgB,EAAE,CAAC;QACxC,MAAM,IAAI,KAAK,CACb,yBAAyB,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,oDAAoD,CACtH,CAAC;IACJ,CAAC;IAED,MAAM,gBAAgB,GAAG,QAAQ,CAAC,QAAQ,CAAC,gBAAgB,CAAC,IAAI,CAC9D,QAAQ,CAAC,QAAQ,CAClB,CAAC;IAEF,4HAA4H;IAC5H,MAAM,UAAU,GAAG,QAAQ,CAAC,QAAQ,CAAC,gBAAgB,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;IAC1E,MAAM,QAAQ,GAAa,EAAE,CAAC;IAC9B,IAAI,CAAC,UAAU,EAAE,CAAC;QAChB,QAAQ,CAAC,IAAI,CACX,GAAG,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,qHAAqH,CACjK,CAAC;IACJ,CAAC;IAED,qFAAqF;IACrF,MAAM,uBAAuB,GAAG,UAAU;QACxC,CAAC,CAAC,EAAE,GAAG,OAAO,CAAC,eAAe,EAAE,GAAG,UAAU,CAAC,eAAe,EAAE;QAC/D,CAAC,CAAC,OAAO,CAAC,eAAe,CAAC;IAE5B,MAAM,iBAAiB,GAAG,OAAO,CAAC,UAAU,IAAI,KAAK,CAAC;IACtD,MAAM,2BAA2B,GAAG,6BAA6B,CAAC,QAAQ,CAAC,CAAC;IAC5E,MAAM,mBAAmB,GAAG,iBAAiB,IAAI,2BAA2B,CAAC;IAE7E,MAAM,UAAU,GAAG,GAAG,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,EAAE,CAAC;IACjE,IAAI,CAAC,iBAAiB,EAAE,CAAC;QACvB,KAAK,CAAC,GAAG,UAAU,sDAAsD,CAAC,CAAC;IAC7E,CAAC;SAAM,IAAI,mBAAmB,EAAE,CAAC;QAC/B,KAAK,CACH,GAAG,UAAU,uEAAuE,CACrF,CAAC;IACJ,CAAC;SAAM,CAAC;QACN,KAAK,CACH,GAAG,UAAU,2IAA2I,CACzJ,CAAC;IACJ,CAAC;IAED,MAAM,MAAM,GAAG,MAAM,MAAM,CACzB,GAAG,EAAE,CACH,gBAAgB,CAAC;QACf,OAAO,EAAE,QAAQ,CAAC,OAAO;QACzB,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC;QACnE,eAAe,EAAE,uBAAuB;QACxC,WAAW,EAAE,OAAO,CAAC,WAAW;QAChC,OAAO,EAAE,OAAO,CAAC,OAAO;QACxB,iBAAiB,EAAE,mBAAmB;KACvC,CAAC,EACJ;QACE,OAAO,EAAE,UAAU;QACnB,MAAM,EAAE,OAAO,CAAC,WAAW;QAC3B,WAAW,EAAE,CAAC,EAAE,KAAK,EAAE,EAAE,EAAE;YACzB,IAAI,KAAK,YAAY,QAAQ,IAAI,CAAC,mBAAmB,CAAC,KAAK,CAAC,EAAE,CAAC;gBAC7D,OAAO,KAAK,CAAC;YACf,CAAC;YACD,OAAO,IAAI,CAAC;QACd,CAAC;KACF,CACF,CAAC;IAEF,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC,CAAC;IAExD,IAAI,MAAM,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC9B,MAAM,IAAI,sBAAsB,EAAE,CAAC;IACrC,CAAC;IAED,IAAI,UAAU,GAAwB,MAAM,CAAC,KAAK,CAAC;IACnD,+EAA+E;IAC/E,IAAI,eAAe,GAAG,UAAU,EAAE,SAAS,IAAI,MAAM,CAAC,SAAS,CAAC;IAEhE,IAAI,UAAU,EAAE,CAAC;QACf,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,MAAM,CAAC,oBAAoB,CAAC,CAAC;QAC5D,UAAU,GAAG,MAAM,YAAY,CAAC;YAC9B,KAAK,EAAE,MAAM,CAAC,KAAK;YACnB,SAAS,EAAE,UAAU,CAAC,SAAS;YAC/B,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,CAAC,EAAE;SACtC,CAAC,CAAC;QACH,eAAe,GAAG,WAAW,CAAC;IAChC,CAAC;IAED,MAAM,KAAK,GAAG,IAAI,yBAAyB,CAAC;QAC1C,IAAI,EAAE,UAAU;QAChB,SAAS,EAAE,eAAe;KAC3B,CAAC,CAAC;IAEH,MAAM,gBAAgB,GAAG,MAAM,oBAAoB,CACjD,KAAK,CAAC,UAAU,EAChB,eAAe,CAChB,CAAC;IACF,MAAM,eAAe,GAAG,gBAAgB,IAAI,MAAM,CAAC,eAAe,CAAC;IAEnE,MAAM,EAAE,UAAU,EAAE,QAAQ,EAAE,mBAAmB,EAAE,GACjD,MAAM,+BAA+B,CAAC;QACpC,iBAAiB;QACjB,gBAAgB,EAAE,MAAM,CAAC,UAAU;QACnC,KAAK,EAAE,KAAK,CAAC,UAAU;QACvB,SAAS,EAAE,eAAe;QAC1B,QAAQ,EAAE,GAAG,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,EAAE;QACvD,QAAQ;QACR,WAAW,EAAE,OAAO,CAAC,WAAW;QAChC,KAAK,EAAE,OAAO,CAAC,KAAK;KACrB,CAAC,CAAC;IAEL,MAAM,UAAU,GAAG,OAAO,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IAExE,MAAM,QAAQ,GAAmB;QAC/B,SAAS;QACT,UAAU;QACV,GAAG,CAAC,eAAe,IAAI,IAAI,IAAI,EAAE,eAAe,EAAE,CAAC;KACpD,CAAC;IAEF,MAAM,cAAc,GAClB,QAAQ,CAAC,MAAM,GAAG,CAAC,IAAI,mBAAmB,CAAC,MAAM,GAAG,CAAC;QACnD,CAAC,CAAC,CAAC,GAAG,QAAQ,EAAE,GAAG,mBAAmB,CAAC;QACvC,CAAC,CAAC,SAAS,CAAC;IAEhB,OAAO;QACL,KAAK;QACL,QAAQ;QACR,gBAAgB,EAAE,MAAM,CAAC,gBAAgB;QACzC,QAAQ,EAAE,cAAc;QACxB,UAAU;KACX,CAAC;AACJ,CAAC;AAED,KAAK,UAAU,+BAA+B,CAAkB,IAS/D;IAIC,IAAI,CAAC,IAAI,CAAC,iBAAiB,EAAE,CAAC;QAC5B,OAAO,EAAE,UAAU,EAAE,SAAS,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC;IACjD,CAAC;IAED,iEAAiE;IACjE,IAAI,cAAwC,CAAC;IAC7C,IAAI,IAAI,CAAC,gBAAgB,IAAI,IAAI,CAAC,gBAAgB,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC9D,cAAc,GAAG,IAAI,CAAC,gBAAgB,CAAC;IACzC,CAAC;SAAM,CAAC;QACN,MAAM,QAAQ,GACZ,IAAI,CAAC,QAAQ,CAAC,WAAW,IAAI,CAAC,MAAM,qBAAqB,EAAE,CAAC,CAAC;QAC/D,cAAc,GAAG,MAAM,sBAAsB,CAAC;YAC5C,QAAQ,EAAE,IAAI,CAAC,QAAQ;YACvB,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,SAAS,EAAE,IAAI,CAAC,SAAS;YACzB,iBAAiB,EAAE,QAAQ;YAC3B,WAAW,EAAE,IAAI,CAAC,WAAW;SAC9B,CAAC,CAAC;IACL,CAAC;IAED,MAAM,EAAE,aAAa,EAAE,GAAG,MAAM,MAAM,CAAC,8BAA8B,CAAC,CAAC;IACvE,MAAM,EAAE,iBAAiB,EAAE,GAAG,MAAM,MAAM,CACxC,qCAAqC,CACtC,CAAC;IACF,MAAM,EAAE,mBAAmB,EAAE,GAAG,MAAM,MAAM,CAC1C,wCAAwC,CACzC,CAAC;IAEF,IAAI,WAAW,GACb,EAAE,CAAC;IACL,IAAI,CAAC;QACH,MAAM,OAAO,GAAG,aAAa,CAAC,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,SAAS,CAAC,CAAC;QAC1D,MAAM,IAAI,GAAG,iBAAiB,CAAC,OAAO,CAAC,GAAG,EAAE;YAC1C,UAAU,EAAE,OAAO,CAAC,UAAU;YAC9B,aAAa,EAAE,GAAG;SACnB,CAAC,CAAC;QACH,WAAW,GAAG,IAAI,CAAC;IACrB,CAAC;IAAC,MAAM,CAAC;QACP,qFAAqF;QACrF,gEAAgE;IAClE,CAAC;IAED,MAAM,MAAM,GAAG,mBAAmB,CAAC;QACjC,UAAU,EAAE,cAAc;QAC1B,SAAS,EAAE,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC;QACxC,WAAW;KACZ,CAAC,CAAC;IAEH,OAAO;QACL,UAAU,EAAE,MAAM,CAAC,UAAU;QAC7B,QAAQ,EAAE,MAAM,CAAC,QAAQ;KAC1B,CAAC;AACJ,CAAC"}

package/dist/generate-speech.d.ts

@@ -1,4 +1,4 @@
-import type { ResolvedModel, Voice } from "./speech-provider.js";
+import { type ResolvedModel, type Voice } from "./speech-provider.js";
 import type { SpeechResult } from "./speech-result.js";
 export declare function generateSpeech<V extends Voice = Voice>(options: {
     model: string | ResolvedModel<V>;
@@ -9,15 +9,7 @@ export declare function generateSpeech<V extends Voice = Voice>(options: {
     maxRetries?: number;
     abortSignal?: AbortSignal;
     headers?: Record<string, string>;
-    /**
-     * RMS-normalize the returned audio to this dBFS level. Must be ≤ 0.
-     * When set, generateSpeech requests the provider's decodable PCM/WAV
-     * output mode (via `getStitchOptions`), normalizes the samples to the
-     * target loudness, and re-encodes the result as 16-bit mono WAV — so
-     * the response `mediaType` will be `audio/wav` regardless of the
-     * provider's native default. Throws `VolumeAdjustmentUnsupportedError`
-     * if the provider doesn't expose a decodable output mode.
-     */
     volumeDbfs?: number;
+    timestamps?: boolean;
 }): Promise<SpeechResult>;
 //# sourceMappingURL=generate-speech.d.ts.map

package/dist/generate-speech.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"generate-speech.d.ts","sourceRoot":"","sources":["../src/generate-speech.ts"],"names":[],"mappings":"AAUA,OAAO,KAAK,EAAE,aAAa,EAAE,KAAK,EAAE,MAAM,sBAAsB,CAAC;AACjE,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAGvD,wBAAsB,cAAc,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK,EAAE,OAAO,EAAE;IACrE,KAAK,EAAE,MAAM,GAAG,aAAa,CAAC,CAAC,CAAC,CAAC;IACjC,IAAI,EAAE,MAAM,CAAC;IACb,KAAK,EAAE,CAAC,CAAC;IACT,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC1C,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,WAAW,CAAC,EAAE,WAAW,CAAC;IAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IACjC;;;;;;;;OAQG;IACH,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB,GAAG,OAAO,CAAC,YAAY,CAAC,CAwHxB"}
+{"version":3,"file":"generate-speech.d.ts","sourceRoot":"","sources":["../src/generate-speech.ts"],"names":[],"mappings":"AAcA,OAAO,EAGL,KAAK,aAAa,EAClB,KAAK,KAAK,EACX,MAAM,sBAAsB,CAAC;AAC9B,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAKvD,wBAAsB,cAAc,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK,EAAE,OAAO,EAAE;IACrE,KAAK,EAAE,MAAM,GAAG,aAAa,CAAC,CAAC,CAAC,CAAC;IACjC,IAAI,EAAE,MAAM,CAAC;IACb,KAAK,EAAE,CAAC,CAAC;IACT,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC1C,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,WAAW,CAAC,EAAE,WAAW,CAAC;IAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAEjC,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,UAAU,CAAC,EAAE,OAAO,CAAC;CACtB,GAAG,OAAO,CAAC,YAAY,CAAC,CAoJxB"}

package/dist/generate-speech.js

@@ -1,61 +1,76 @@
 import pRetry from "p-retry";
 import { computeAudioDuration } from "./audio-duration.js";
 import { detectAudioTags, stripAudioTags } from "./audio-tags.js";
+import { getDefaultSTTFallback } from "./default-stt-fallback.js";
+import { deriveTimestampsViaSTT } from "./derive-timestamps.js";
 import { ApiError, NoSpeechGeneratedError, VolumeAdjustmentUnsupportedError, } from "./errors.js";
+import { debug } from "./logger.js";
+import { isRetriableApiError } from "./provider-utils.js";
 import { resolveModel } from "./resolve-provider.js";
+import { isSpeechGatewayModel, modelDeclaresNativeTimestamps, } from "./speech-provider.js";
 import { DefaultGeneratedAudioFile } from "./speech-result.js";
 export async function generateSpeech(options) {
-    const { model, voice, abortSignal, headers, volumeDbfs } = options;
+    const { model, voice, abortSignal, headers, volumeDbfs, timestamps = false, } = options;
     const maxRetries = options.maxRetries ?? 2;
     const resolved = resolveModel(model, { apiKey: options.apiKey });
     const modelIdentifier = `${resolved.provider.id}/${resolved.modelId}`;
+    const isGateway = isSpeechGatewayModel(resolved);
     let providerOptions = options.providerOptions;
-    if (volumeDbfs != null) {
+    if (volumeDbfs != null && !isGateway) {
         const stitchOpts = resolved.provider.getStitchOptions?.(resolved.modelId);
         if (!stitchOpts) {
             throw new VolumeAdjustmentUnsupportedError(modelIdentifier);
         }
-        // Stitch-mode options are applied last so they win over user-supplied
-        // providerOptions — otherwise a caller could silently break the decoder
-        // by e.g. passing `response_format: "mp3"` alongside `volumeDbfs`.
+        // Stitch options must win — caller-supplied response_format would break the decoder.
         providerOptions = {
             ...options.providerOptions,
             ...stitchOpts.providerOptions,
         };
     }
-    let processedText;
-    let warnings;
-    if (resolved.provider.processAudioTags) {
-        ({ text: processedText, warnings } = resolved.provider.processAudioTags(options.text, resolved.modelId));
-    }
-    else {
-        const tags = detectAudioTags(options.text);
-        if (tags.length > 0) {
-            ({ text: processedText, warnings } = stripAudioTags(options.text, modelIdentifier));
-        }
-        else {
-            processedText = options.text;
-            warnings = [];
-        }
-    }
+    const { text: processedText, warnings } = preprocessText(resolved, options.text, modelIdentifier);
     if (processedText.trim().length === 0) {
         throw new NoSpeechGeneratedError(warnings.length > 0
             ? `Text is empty after removing unsupported audio tags for ${modelIdentifier}.`
             : "Text must not be empty.");
     }
+    const hasNativeTimestamps = modelDeclaresNativeTimestamps(resolved);
+    const shouldRequestNative = timestamps && (hasNativeTimestamps || isGateway);
+    const effectiveFallback = !timestamps || shouldRequestNative
+        ? undefined
+        : (resolved.fallbackSTT ?? (await getDefaultSTTFallback()));
+    logTimestampDecision({
+        modelIdentifier,
+        enabled: timestamps,
+        hasNative: hasNativeTimestamps,
+        willRequestNative: shouldRequestNative,
+        effectiveFallback,
+    });
     const startTime = performance.now();
-    const result = await pRetry(() => resolved.provider.generate({
-        modelId: resolved.modelId,
-        text: processedText,
-        voice,
-        providerOptions,
-        abortSignal,
-        headers,
-    }), {
+    const result = await pRetry(() => isSpeechGatewayModel(resolved)
+        ? resolved.provider.generate({
+            modelId: resolved.modelId,
+            text: processedText,
+            // Gateway inline mode only accepts string voice IDs.
+            voice: voice,
+            providerOptions,
+            abortSignal,
+            headers,
+            includeTimestamps: shouldRequestNative,
+            volumeDbfs,
+        })
+        : resolved.provider.generate({
+            modelId: resolved.modelId,
+            text: processedText,
+            voice,
+            providerOptions,
+            abortSignal,
+            headers,
+            includeTimestamps: shouldRequestNative,
+        }), {
         retries: maxRetries,
         signal: abortSignal,
         shouldRetry: ({ error }) => {
-            if (error instanceof ApiError && error.statusCode < 500) {
+            if (error instanceof ApiError && !isRetriableApiError(error)) {
                 return false;
             }
             return true;
@@ -68,7 +83,7 @@ export async function generateSpeech(options) {
     }
     let outputBytes = audioData;
     let outputMediaType = result.mediaType;
-    if (volumeDbfs != null) {
+    if (volumeDbfs != null && !isGateway) {
         const { adjustVolume } = await import("./volume-adjust.js");
         outputBytes = await adjustVolume({
             audio: audioData,
@@ -83,18 +98,81 @@ export async function generateSpeech(options) {
     });
     const audioDurationMs = (await computeAudioDuration(audio.uint8Array, outputMediaType)) ??
         result.audioDurationMs;
+    const resolvedTimestamps = await resolveTimestamps({
+        timestamps,
+        modelIdentifier,
+        resolved,
+        resultTimestamps: result.timestamps,
+        audio: audio.uint8Array,
+        mediaType: outputMediaType,
+        abortSignal,
+    });
     const metadata = {
         latencyMs,
         inputChars: processedText.length,
-        provider: resolved.provider.id,
-        model: resolved.modelId,
         ...(audioDurationMs != null && { audioDurationMs }),
     };
     return {
         audio,
         metadata,
         providerMetadata: result.providerMetadata,
-        warnings: warnings.length > 0 ? warnings : undefined,
+        warnings: mergeWarnings(warnings, result.warnings),
+        timestamps: resolvedTimestamps,
     };
 }
+function mergeWarnings(preprocessingWarnings, providerWarnings) {
+    const merged = [...preprocessingWarnings, ...(providerWarnings ?? [])];
+    return merged.length > 0 ? merged : undefined;
+}
+async function resolveTimestamps(args) {
+    if (!args.timestamps) {
+        return;
+    }
+    if (args.resultTimestamps?.length) {
+        debug(`${args.modelIdentifier}: returned ${args.resultTimestamps.length} native word timestamps.`);
+        return args.resultTimestamps;
+    }
+    if (isSpeechGatewayModel(args.resolved)) {
+        return;
+    }
+    const fallback = args.resolved.fallbackSTT ?? (await getDefaultSTTFallback());
+    const timestamps = await deriveTimestampsViaSTT({
+        ttsModel: args.modelIdentifier,
+        audio: args.audio,
+        mediaType: args.mediaType,
+        timestampFallback: fallback,
+        abortSignal: args.abortSignal,
+    });
+    debug(`${args.modelIdentifier}: derived ${timestamps.length} word timestamps via STT fallback.`);
+    return timestamps;
+}
+function preprocessText(resolved, rawText, modelIdentifier) {
+    // Gateway server handles audio-tag normalization itself — pass raw text through.
+    if (isSpeechGatewayModel(resolved)) {
+        return { text: rawText, warnings: [] };
+    }
+    if (resolved.provider.processAudioTags) {
+        return resolved.provider.processAudioTags(rawText, resolved.modelId);
+    }
+    const tags = detectAudioTags(rawText);
+    if (tags.length > 0) {
+        return stripAudioTags(rawText, modelIdentifier);
+    }
+    return { text: rawText, warnings: [] };
+}
+function logTimestampDecision(args) {
+    const { modelIdentifier, enabled, willRequestNative } = args;
+    if (!enabled) {
+        debug(`${modelIdentifier}: timestamps: false — skipping alignment.`);
+        return;
+    }
+    if (willRequestNative) {
+        debug(`${modelIdentifier}: timestamps: true — requesting native alignment from the provider.`);
+        return;
+    }
+    const target = args.effectiveFallback
+        ? `${args.effectiveFallback.provider.id}/${args.effectiveFallback.modelId}`
+        : "unconfigured STT fallback";
+    debug(`${modelIdentifier}: timestamps: true but no native alignment available — will pipe synthesized audio through ${target} for word timestamps (adds a round-trip).`);
+}
 //# sourceMappingURL=generate-speech.js.map