@speech-sdk/core 0.7.0 → 0.8.0

package/dist/conversation/stitch.d.ts

@@ -1,6 +1,8 @@
+import { type AudioOutput } from "../audio-output.js";
+import type { SpeechMetadata } from "../metadata.js";
+import type { PronunciationsInput } from "../pronunciations/types.js";
 import type { ResolvedModel, Voice } from "../speech-provider.js";
-import type { ResolvedSTTModel } from "../speech-to-text-provider.js";
-import type { TimestampMode, WordTimestamp } from "../timestamps.js";
+import type { ConversationWordTimestamp } from "../timestamps.js";
 import type { ConversationTurn } from "./types.js";
 interface StitchInput<V extends Voice = Voice> {
     readonly abortSignal?: AbortSignal;
@@ -9,14 +11,14 @@ interface StitchInput<V extends Voice = Voice> {
     readonly headers?: Record<string, string>;
     readonly maxConcurrency: number;
     readonly maxRetries: number;
-    readonly normalizeVolume: boolean;
+    readonly output?: AudioOutput;
+    readonly pronunciations?: PronunciationsInput;
     readonly resolvedPerTurn: readonly ResolvedModel<V>[];
     readonly stitchOptionsPerTurn: readonly {
         providerOptions: Record<string, unknown>;
         mediaType: string;
     }[];
-    readonly timestampProvider?: ResolvedSTTModel;
-    readonly timestamps: TimestampMode;
+    readonly timestamps: boolean;
     readonly topLevelProviderOptions?: Record<string, unknown>;
     readonly turns: readonly ConversationTurn<V>[];
     readonly volumeDbfs?: number;
@@ -29,8 +31,9 @@ interface StitchOutput {
         readonly latencyMs: number;
         readonly audioDurationMs?: number;
     };
+    readonly metadataPerTurn: readonly SpeechMetadata[];
     readonly providerMetadataPerTurn: readonly (Record<string, unknown> | undefined)[];
-    readonly timestamps?: readonly WordTimestamp[];
+    readonly timestamps?: readonly ConversationWordTimestamp[];
     readonly warnings: readonly string[];
 }
 export declare function runStitch<V extends Voice>(input: StitchInput<V>): Promise<StitchOutput>;

package/dist/conversation/stitch.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"stitch.d.ts","sourceRoot":"","sources":["../../src/conversation/stitch.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,aAAa,EAAE,KAAK,EAAE,MAAM,uBAAuB,CAAC;AAClE,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,+BAA+B,CAAC;AACtE,OAAO,KAAK,EAAE,aAAa,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAOrE,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,YAAY,CAAC;AAEnD,UAAU,WAAW,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK;IAC3C,QAAQ,CAAC,WAAW,CAAC,EAAE,WAAW,CAAC;IACnC,QAAQ,CAAC,MAAM,CAAC,EAAE,MAAM,CAAC;IACzB,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAC1C,QAAQ,CAAC,cAAc,EAAE,MAAM,CAAC;IAChC,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,eAAe,EAAE,OAAO,CAAC;IAClC,QAAQ,CAAC,eAAe,EAAE,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC;IACtD,QAAQ,CAAC,oBAAoB,EAAE,SAAS;QACtC,eAAe,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QACzC,SAAS,EAAE,MAAM,CAAC;KACnB,EAAE,CAAC;IACJ,QAAQ,CAAC,iBAAiB,CAAC,EAAE,gBAAgB,CAAC;IAC9C,QAAQ,CAAC,UAAU,EAAE,aAAa,CAAC;IACnC,QAAQ,CAAC,uBAAuB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC3D,QAAQ,CAAC,KAAK,EAAE,SAAS,gBAAgB,CAAC,CAAC,CAAC,EAAE,CAAC;IAC/C,QAAQ,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC;CAC9B;AAED,UAAU,YAAY;IACpB,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;IAC3B,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,QAAQ,CAAC,QAAQ,EAAE;QACjB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;QAC5B,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;QAC3B,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC;KACnC,CAAC;IACF,QAAQ,CAAC,uBAAuB,EAAE,SAAS,CACvC,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GACvB,SAAS,CACZ,EAAE,CAAC;IACJ,QAAQ,CAAC,UAAU,CAAC,EAAE,SAAS,aAAa,EAAE,CAAC;IAC/C,QAAQ,CAAC,QAAQ,EAAE,SAAS,MAAM,EAAE,CAAC;CACtC;AA+BD,wBAAsB,SAAS,CAAC,CAAC,SAAS,KAAK,EAC7C,KAAK,EAAE,WAAW,CAAC,CAAC,CAAC,GACpB,OAAO,CAAC,YAAY,CAAC,CA2HvB"}
+{"version":3,"file":"stitch.d.ts","sourceRoot":"","sources":["../../src/conversation/stitch.ts"],"names":[],"mappings":"AACA,OAAO,EACL,KAAK,WAAW,EAEjB,MAAM,oBAAoB,CAAC;AAI5B,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,gBAAgB,CAAC;AACrD,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,4BAA4B,CAAC;AACtE,OAAO,KAAK,EAAE,aAAa,EAAE,KAAK,EAAE,MAAM,uBAAuB,CAAC;AAClE,OAAO,KAAK,EAAE,yBAAyB,EAAE,MAAM,kBAAkB,CAAC;AAGlE,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,YAAY,CAAC;AAEnD,UAAU,WAAW,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK;IAC3C,QAAQ,CAAC,WAAW,CAAC,EAAE,WAAW,CAAC;IACnC,QAAQ,CAAC,MAAM,CAAC,EAAE,MAAM,CAAC;IACzB,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAC1C,QAAQ,CAAC,cAAc,EAAE,MAAM,CAAC;IAChC,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,MAAM,CAAC,EAAE,WAAW,CAAC;IAC9B,QAAQ,CAAC,cAAc,CAAC,EAAE,mBAAmB,CAAC;IAC9C,QAAQ,CAAC,eAAe,EAAE,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC;IACtD,QAAQ,CAAC,oBAAoB,EAAE,SAAS;QACtC,eAAe,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QACzC,SAAS,EAAE,MAAM,CAAC;KACnB,EAAE,CAAC;IACJ,QAAQ,CAAC,UAAU,EAAE,OAAO,CAAC;IAC7B,QAAQ,CAAC,uBAAuB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC3D,QAAQ,CAAC,KAAK,EAAE,SAAS,gBAAgB,CAAC,CAAC,CAAC,EAAE,CAAC;IAC/C,QAAQ,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC;CAC9B;AAED,UAAU,YAAY;IACpB,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;IAC3B,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,QAAQ,CAAC,QAAQ,EAAE;QACjB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;QAC5B,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;QAC3B,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC;KACnC,CAAC;IACF,QAAQ,CAAC,eAAe,EAAE,SAAS,cAAc,EAAE,CAAC;IACpD,QAAQ,CAAC,uBAAuB,EAAE,SAAS,CACvC,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GACvB,SAAS,CACZ,EAAE,CAAC;IACJ,QAAQ,CAAC,UAAU,CAAC,EAAE,SAAS,yBAAyB,EAAE,CAAC;IAC3D,QAAQ,CAAC,QAAQ,EAAE,SAAS,MAAM,EAAE,CAAC;CACtC;AA4BD,wBAAsB,SAAS,CAAC,CAAC,SAAS,KAAK,EAC7C,KAAK,EAAE,WAAW,CAAC,CAAC,CAAC,GACpB,OAAO,CAAC,YAAY,CAAC,CA8IvB"}

package/dist/conversation/stitch.js

@@ -1,11 +1,12 @@
+import { decodeAudioToPcm16 } from "../audio-decode.js";
+import { applyOptionalOutputConversion, } from "../audio-output.js";
+import { withTurnIndex } from "../errors.js";
 import { generateSpeech } from "../generate-speech.js";
 import { debug } from "../logger.js";
-import { concatPcmToWav, dbfsToInt16Rms, decodeToPcm16, normalizeRms, } from "./pcm-concat.js";
+import { concatPcmToWav, dbfsToInt16Rms, normalizeRms } from "./pcm-concat.js";
+import { fillTurnTimestampsProportional } from "./proportional-fill.js";
 const TARGET_SAMPLE_RATE = 24_000;
-/**
- * Run `worker(items[i], i)` for each item, capping in-flight executions at
- * `concurrency`. Preserves input ordering in the returned array.
- */
+const WHITESPACE_RE = /\s+/;
 async function mapWithConcurrency(items, concurrency, worker) {
     const results = new Array(items.length);
     let next = 0;
@@ -31,85 +32,105 @@ export async function runStitch(input) {
             ...turn.providerOptions,
             ...stitchOpts.providerOptions,
         };
-        const result = await generateSpeech({
-            model: resolved,
-            text: turn.text,
-            voice: turn.voice,
-            apiKey: input.apiKey,
-            providerOptions: mergedProviderOptions,
-            maxRetries: input.maxRetries,
-            abortSignal: input.abortSignal,
-            headers: input.headers,
-            timestamps: input.timestamps,
-            timestampProvider: input.timestampProvider,
-        });
-        // Prefer the mediaType from getStitchOptions over the response
-        // content-type: providers' response headers often omit the sample
-        // rate (e.g. Hume sends `audio/pcm` for what is actually 48 kHz),
-        // and getStitchOptions is the authoritative declaration of what
-        // the provider returns for the requested format.
-        const segment = decodeToPcm16(result.audio.uint8Array, stitchOpts.mediaType);
+        let result;
+        try {
+            result = await generateSpeech({
+                model: resolved,
+                text: turn.text,
+                voice: turn.voice,
+                apiKey: input.apiKey,
+                providerOptions: mergedProviderOptions,
+                maxRetries: input.maxRetries,
+                abortSignal: input.abortSignal,
+                headers: input.headers,
+                timestamps: input.timestamps,
+                pronunciations: input.pronunciations,
+            });
+        }
+        catch (err) {
+            throw withTurnIndex(err, i);
+        }
+        // Hume and others omit sample rate from content-type; prefer getStitchOptions.
+        const segment = await decodeAudioToPcm16(result.audio.uint8Array, stitchOpts.mediaType);
         return { result, segment };
     });
     const segments = perTurn.map((p) => p.segment);
-    const leveledSegments = input.normalizeVolume
-        ? normalizeRms(segments, input.volumeDbfs == null ? undefined : dbfsToInt16Rms(input.volumeDbfs))
-        : segments;
+    const leveledSegments = normalizeRms(segments, input.volumeDbfs == null ? undefined : dbfsToInt16Rms(input.volumeDbfs));
     const audio = await concatPcmToWav(leveledSegments, {
         gapMs: input.gapMs,
         targetSampleRate: TARGET_SAMPLE_RATE,
     });
+    const { audio: finalAudio, mediaType } = await applyOptionalOutputConversion({
+        audio,
+        mediaType: "audio/wav",
+        output: input.output,
+    });
     const totalSamples = perTurn.reduce((n, p) => n +
         Math.round((p.segment.pcm.length / p.segment.sampleRate) * TARGET_SAMPLE_RATE), 0) +
         (perTurn.length - 1) *
             Math.round((input.gapMs / 1000) * TARGET_SAMPLE_RATE);
     const audioDurationMs = Math.round((totalSamples / TARGET_SAMPLE_RATE) * 1000);
     const warnings = perTurn.flatMap((p) => p.result.warnings ?? []);
+    const metadataPerTurn = perTurn.map((p) => p.result.metadata);
     const providerMetadataPerTurn = perTurn.map((p) => p.result.providerMetadata);
-    // Compose per-turn word timestamps into a single flat list, offset by the
-    // cumulative duration of prior turns + (gapMs * number of preceding gaps).
-    // Uses each segment's *source* duration (pcm.length / sampleRate) rather
-    // than the resampled target, because the offsets must match the audio the
-    // per-turn STT/native path actually saw — resampling is a constant-duration
-    // transform but rounding differences can drift by a sample or two.
+    // Use source duration (pre-resample) so offsets match what the per-turn STT/native saw.
     const gapSeconds = input.gapMs / 1000;
     const turnDurations = perTurn.map((p) => p.segment.pcm.length / p.segment.sampleRate);
-    const allTurnsHaveTimestamps = input.timestamps !== "off" &&
-        perTurn.every((p) => p.result.timestamps !== undefined);
+    const fillWarnings = [];
     let timestamps;
-    if (allTurnsHaveTimestamps) {
+    if (input.timestamps) {
         timestamps = [];
         let offsetSec = 0;
+        const filledTurns = [];
         for (let i = 0; i < perTurn.length; i++) {
-            const turnTimestamps = perTurn[i]?.result.timestamps ?? [];
-            for (const w of turnTimestamps) {
-                timestamps.push({
-                    text: w.text,
-                    start: w.start + offsetSec,
-                    end: w.end + offsetSec,
+            const turnTimestamps = perTurn[i]?.result.timestamps;
+            if (turnTimestamps && turnTimestamps.length > 0) {
+                for (const w of turnTimestamps) {
+                    timestamps.push({
+                        text: w.text,
+                        start: w.start + offsetSec,
+                        end: w.end + offsetSec,
+                        turnIndex: i,
+                    });
+                }
+            }
+            else {
+                const turnText = input.turns[i]?.text ?? "";
+                const tokens = turnText
+                    .split(WHITESPACE_RE)
+                    .filter((t) => t.length > 0);
+                const turnSec = turnDurations[i] ?? 0;
+                const filled = fillTurnTimestampsProportional({
+                    turnIndex: i,
+                    tokenCount: tokens.length,
+                    startSec: offsetSec,
+                    endSec: offsetSec + turnSec,
+                    texts: tokens,
                 });
+                timestamps.push(...filled);
+                filledTurns.push(i);
             }
             offsetSec += (turnDurations[i] ?? 0) + gapSeconds;
         }
-        debug(`stitch: composed ${timestamps.length} word timestamps across ${perTurn.length} turn(s).`);
-    }
-    else if (input.timestamps !== "off") {
-        const missing = perTurn
-            .map((p, i) => (p.result.timestamps === undefined ? i : -1))
-            .filter((i) => i !== -1);
-        debug(`stitch: returning no timestamps — ${missing.length}/${perTurn.length} turn(s) had no alignment data (turns: ${missing.join(", ")}). Use timestamps: "on" and/or mark provider models as native/derived to get full coverage.`);
+        if (filledTurns.length > 0) {
+            fillWarnings.push(`speech-sdk: stitch path filled timestamps for turn(s) [${filledTurns.join(",")}] proportionally — provider returned no per-word alignment for those turns.`);
+        }
+        debug(`stitch: composed ${timestamps.length} word timestamps across ${perTurn.length} turn(s); ${filledTurns.length} turn(s) filled proportionally.`);
     }
     return {
-        audio,
-        mediaType: "audio/wav",
+        audio: finalAudio,
+        mediaType,
         metadata: {
             inputChars: input.turns.reduce((n, t) => n + t.text.length, 0),
             latencyMs: Math.round(performance.now() - start),
             audioDurationMs,
         },
+        metadataPerTurn,
         providerMetadataPerTurn,
         timestamps,
-        warnings,
+        warnings: warnings.length > 0 || fillWarnings.length > 0
+            ? [...warnings, ...fillWarnings]
+            : warnings,
     };
 }
 //# sourceMappingURL=stitch.js.map

package/dist/conversation/stitch.js.map

@@ -1 +1 @@
-{"version":3,"file":"stitch.js","sourceRoot":"","sources":["../../src/conversation/stitch.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AACvD,OAAO,EAAE,KAAK,EAAE,MAAM,cAAc,CAAC;AAIrC,OAAO,EACL,cAAc,EACd,cAAc,EACd,aAAa,EACb,YAAY,GACb,MAAM,iBAAiB,CAAC;AAuCzB,MAAM,kBAAkB,GAAG,MAAM,CAAC;AAElC;;;GAGG;AACH,KAAK,UAAU,kBAAkB,CAC/B,KAAmB,EACnB,WAAmB,EACnB,MAA8C;IAE9C,MAAM,OAAO,GAAQ,IAAI,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IAC7C,IAAI,IAAI,GAAG,CAAC,CAAC;IACb,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,CACxB,EAAE,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,WAAW,EAAE,CAAC,CAAC,EAAE,KAAK,CAAC,MAAM,CAAC,EAAE,EAC5D,KAAK,IAAI,EAAE;QACT,OAAO,IAAI,EAAE,CAAC;YACZ,MAAM,CAAC,GAAG,IAAI,EAAE,CAAC;YACjB,IAAI,CAAC,IAAI,KAAK,CAAC,MAAM,EAAE,CAAC;gBACtB,OAAO;YACT,CAAC;YACD,OAAO,CAAC,CAAC,CAAC,GAAG,MAAM,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;QACzC,CAAC;IACH,CAAC,CACF,CAAC;IACF,MAAM,OAAO,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC;IAC3B,OAAO,OAAO,CAAC;AACjB,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,SAAS,CAC7B,KAAqB;IAErB,MAAM,KAAK,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;IAEhC,MAAM,OAAO,GAAG,MAAM,kBAAkB,CACtC,KAAK,CAAC,KAAK,EACX,KAAK,CAAC,cAAc,EACpB,KAAK,EAAE,IAAI,EAAE,CAAC,EAAE,EAAE;QAChB,MAAM,QAAQ,GAAG,KAAK,CAAC,eAAe,CAAC,CAAC,CAAC,CAAC;QAC1C,MAAM,UAAU,GAAG,KAAK,CAAC,oBAAoB,CAAC,CAAC,CAAC,CAAC;QACjD,MAAM,qBAAqB,GAAG;YAC5B,GAAG,KAAK,CAAC,uBAAuB;YAChC,GAAG,IAAI,CAAC,eAAe;YACvB,GAAG,UAAU,CAAC,eAAe;SAC9B,CAAC;QACF,MAAM,MAAM,GAAG,MAAM,cAAc,CAAC;YAClC,KAAK,EAAE,QAAQ;YACf,IAAI,EAAE,IAAI,CAAC,IAAI;YACf,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,MAAM,EAAE,KAAK,CAAC,MAAM;YACpB,eAAe,EAAE,qBAAqB;YACtC,UAAU,EAAE,KAAK,CAAC,UAAU;YAC5B,WAAW,EAAE,KAAK,CAAC,WAAW;YAC9B,OAAO,EAAE,KAAK,CAAC,OAAO;YACtB,UAAU,EAAE,KAAK,CAAC,UAAU;YAC5B,iBAAiB,EAAE,KAAK,CAAC,iBAAiB;SAC3C,CAAC,CAAC;QACH,+DAA+D;QAC/D,kEAAkE;QAClE,kEAAkE;QAClE,gEAAgE;QAChE,iDAAiD;QACjD,MAAM,OAAO,GAAG,aAAa,CAC3B,MAAM,CAAC,KAAK,CAAC,UAAU,EACvB,UAAU,CAAC,SAAS,CACrB,CAAC;QACF,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,CAAC;IAC7B,CAAC,CACF,CAAC;IAEF,MAAM,QAAQ,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC;IAC/C,MAAM,eAAe,GAAG,KAAK,CAAC,eAAe;QAC3C,CAAC,CAAC,YAAY,CACV,QAAQ,EACR,KAAK,CAAC,UAAU,IAAI,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,KAAK,CAAC,UAAU,CAAC,CACxE;QACH,CAAC,CAAC,QAAQ,CAAC;IAEb,MAAM,KAAK,GAAG,MAAM,cAAc,CAAC,eAAe,EAAE;QAClD,KAAK,EAAE,KAAK,CAAC,KAAK;QAClB,gBAAgB,EAAE,kBAAkB;KACrC,CAAC,CAAC;IAEH,MAAM,YAAY,GAChB,OAAO,CAAC,MAAM,CACZ,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CACP,CAAC;QACD,IAAI,CAAC,KAAK,CACR,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,CAAC,OAAO,CAAC,UAAU,CAAC,GAAG,kBAAkB,CACnE,EACH,CAAC,CACF;QACD,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC;YAClB,IAAI,CAAC,KAAK,CAAC,CAAC,KAAK,CAAC,KAAK,GAAG,IAAI,CAAC,GAAG,kBAAkB,CAAC,CAAC;IAC1D,MAAM,eAAe,GAAG,IAAI,CAAC,KAAK,CAChC,CAAC,YAAY,GAAG,kBAAkB,CAAC,GAAG,IAAI,CAC3C,CAAC;IAEF,MAAM,QAAQ,GAAG,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,QAAQ,IAAI,EAAE,CAAC,CAAC;IACjE,MAAM,uBAAuB,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,gBAAgB,CAAC,CAAC;IAE9E,0EAA0E;IAC1E,2EAA2E;IAC3E,yEAAyE;IACzE,0EAA0E;IAC1E,4EAA4E;IAC5E,mEAAmE;IACnE,MAAM,UAAU,GAAG,KAAK,CAAC,KAAK,GAAG,IAAI,CAAC;IACtC,MAAM,aAAa,GAAG,OAAO,CAAC,GAAG,CAC/B,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,CAAC,OAAO,CAAC,UAAU,CACnD,CAAC;IACF,MAAM,sBAAsB,GAC1B,KAAK,CAAC,UAAU,KAAK,KAAK;QAC1B,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,UAAU,KAAK,SAAS,CAAC,CAAC;IAE1D,IAAI,UAAuC,CAAC;IAC5C,IAAI,sBAAsB,EAAE,CAAC;QAC3B,UAAU,GAAG,EAAE,CAAC;QAChB,IAAI,SAAS,GAAG,CAAC,CAAC;QAClB,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;YACxC,MAAM,cAAc,GAAG,OAAO,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC,UAAU,IAAI,EAAE,CAAC;YAC3D,KAAK,MAAM,CAAC,IAAI,cAAc,EAAE,CAAC;gBAC/B,UAAU,CAAC,IAAI,CAAC;oBACd,IAAI,EAAE,CAAC,CAAC,IAAI;oBACZ,KAAK,EAAE,CAAC,CAAC,KAAK,GAAG,SAAS;oBAC1B,GAAG,EAAE,CAAC,CAAC,GAAG,GAAG,SAAS;iBACvB,CAAC,CAAC;YACL,CAAC;YACD,SAAS,IAAI,CAAC,aAAa,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,UAAU,CAAC;QACpD,CAAC;QACD,KAAK,CACH,oBAAoB,UAAU,CAAC,MAAM,2BAA2B,OAAO,CAAC,MAAM,WAAW,CAC1F,CAAC;IACJ,CAAC;SAAM,IAAI,KAAK,CAAC,UAAU,KAAK,KAAK,EAAE,CAAC;QACtC,MAAM,OAAO,GAAG,OAAO;aACpB,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,UAAU,KAAK,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;aAC3D,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAC3B,KAAK,CACH,qCAAqC,OAAO,CAAC,MAAM,IAAI,OAAO,CAAC,MAAM,0CAA0C,OAAO,CAAC,IAAI,CAAC,IAAI,CAAC,6FAA6F,CAC/N,CAAC;IACJ,CAAC;IAED,OAAO;QACL,KAAK;QACL,SAAS,EAAE,WAAW;QACtB,QAAQ,EAAE;YACR,UAAU,EAAE,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC;YAC9D,SAAS,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC;YAChD,eAAe;SAChB;QACD,uBAAuB;QACvB,UAAU;QACV,QAAQ;KACT,CAAC;AACJ,CAAC"}
+{"version":3,"file":"stitch.js","sourceRoot":"","sources":["../../src/conversation/stitch.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,MAAM,oBAAoB,CAAC;AACxD,OAAO,EAEL,6BAA6B,GAC9B,MAAM,oBAAoB,CAAC;AAC5B,OAAO,EAAE,aAAa,EAAE,MAAM,cAAc,CAAC;AAC7C,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AACvD,OAAO,EAAE,KAAK,EAAE,MAAM,cAAc,CAAC;AAKrC,OAAO,EAAE,cAAc,EAAE,cAAc,EAAE,YAAY,EAAE,MAAM,iBAAiB,CAAC;AAC/E,OAAO,EAAE,8BAA8B,EAAE,MAAM,wBAAwB,CAAC;AAwCxE,MAAM,kBAAkB,GAAG,MAAM,CAAC;AAClC,MAAM,aAAa,GAAG,KAAK,CAAC;AAE5B,KAAK,UAAU,kBAAkB,CAC/B,KAAmB,EACnB,WAAmB,EACnB,MAA8C;IAE9C,MAAM,OAAO,GAAQ,IAAI,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IAC7C,IAAI,IAAI,GAAG,CAAC,CAAC;IACb,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,CACxB,EAAE,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,WAAW,EAAE,CAAC,CAAC,EAAE,KAAK,CAAC,MAAM,CAAC,EAAE,EAC5D,KAAK,IAAI,EAAE;QACT,OAAO,IAAI,EAAE,CAAC;YACZ,MAAM,CAAC,GAAG,IAAI,EAAE,CAAC;YACjB,IAAI,CAAC,IAAI,KAAK,CAAC,MAAM,EAAE,CAAC;gBACtB,OAAO;YACT,CAAC;YACD,OAAO,CAAC,CAAC,CAAC,GAAG,MAAM,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;QACzC,CAAC;IACH,CAAC,CACF,CAAC;IACF,MAAM,OAAO,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC;IAC3B,OAAO,OAAO,CAAC;AACjB,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,SAAS,CAC7B,KAAqB;IAErB,MAAM,KAAK,GAAG,WAAW,CAAC,GAAG,EAAE,CAAC;IAEhC,MAAM,OAAO,GAAG,MAAM,kBAAkB,CACtC,KAAK,CAAC,KAAK,EACX,KAAK,CAAC,cAAc,EACpB,KAAK,EAAE,IAAI,EAAE,CAAC,EAAE,EAAE;QAChB,MAAM,QAAQ,GAAG,KAAK,CAAC,eAAe,CAAC,CAAC,CAAC,CAAC;QAC1C,MAAM,UAAU,GAAG,KAAK,CAAC,oBAAoB,CAAC,CAAC,CAAC,CAAC;QACjD,MAAM,qBAAqB,GAAG;YAC5B,GAAG,KAAK,CAAC,uBAAuB;YAChC,GAAG,IAAI,CAAC,eAAe;YACvB,GAAG,UAAU,CAAC,eAAe;SAC9B,CAAC;QACF,IAAI,MAAkD,CAAC;QACvD,IAAI,CAAC;YACH,MAAM,GAAG,MAAM,cAAc,CAAC;gBAC5B,KAAK,EAAE,QAAQ;gBACf,IAAI,EAAE,IAAI,CAAC,IAAI;gBACf,KAAK,EAAE,IAAI,CAAC,KAAK;gBACjB,MAAM,EAAE,KAAK,CAAC,MAAM;gBACpB,eAAe,EAAE,qBAAqB;gBACtC,UAAU,EAAE,KAAK,CAAC,UAAU;gBAC5B,WAAW,EAAE,KAAK,CAAC,WAAW;gBAC9B,OAAO,EAAE,KAAK,CAAC,OAAO;gBACtB,UAAU,EAAE,KAAK,CAAC,UAAU;gBAC5B,cAAc,EAAE,KAAK,CAAC,cAAc;aACrC,CAAC,CAAC;QACL,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,aAAa,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC;QAC9B,CAAC;QACD,+EAA+E;QAC/E,MAAM,OAAO,GAAG,MAAM,kBAAkB,CACtC,MAAM,CAAC,KAAK,CAAC,UAAU,EACvB,UAAU,CAAC,SAAS,CACrB,CAAC;QACF,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,CAAC;IAC7B,CAAC,CACF,CAAC;IAEF,MAAM,QAAQ,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC;IAC/C,MAAM,eAAe,GAAG,YAAY,CAClC,QAAQ,EACR,KAAK,CAAC,UAAU,IAAI,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,cAAc,CAAC,KAAK,CAAC,UAAU,CAAC,CACxE,CAAC;IAEF,MAAM,KAAK,GAAG,MAAM,cAAc,CAAC,eAAe,EAAE;QAClD,KAAK,EAAE,KAAK,CAAC,KAAK;QAClB,gBAAgB,EAAE,kBAAkB;KACrC,CAAC,CAAC;IAEH,MAAM,EAAE,KAAK,EAAE,UAAU,EAAE,SAAS,EAAE,GAAG,MAAM,6BAA6B,CAAC;QAC3E,KAAK;QACL,SAAS,EAAE,WAAW;QACtB,MAAM,EAAE,KAAK,CAAC,MAAM;KACrB,CAAC,CAAC;IAEH,MAAM,YAAY,GAChB,OAAO,CAAC,MAAM,CACZ,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CACP,CAAC;QACD,IAAI,CAAC,KAAK,CACR,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,CAAC,OAAO,CAAC,UAAU,CAAC,GAAG,kBAAkB,CACnE,EACH,CAAC,CACF;QACD,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC;YAClB,IAAI,CAAC,KAAK,CAAC,CAAC,KAAK,CAAC,KAAK,GAAG,IAAI,CAAC,GAAG,kBAAkB,CAAC,CAAC;IAC1D,MAAM,eAAe,GAAG,IAAI,CAAC,KAAK,CAChC,CAAC,YAAY,GAAG,kBAAkB,CAAC,GAAG,IAAI,CAC3C,CAAC;IAEF,MAAM,QAAQ,GAAG,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,QAAQ,IAAI,EAAE,CAAC,CAAC;IACjE,MAAM,eAAe,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;IAC9D,MAAM,uBAAuB,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,gBAAgB,CAAC,CAAC;IAE9E,wFAAwF;IACxF,MAAM,UAAU,GAAG,KAAK,CAAC,KAAK,GAAG,IAAI,CAAC;IACtC,MAAM,aAAa,GAAG,OAAO,CAAC,GAAG,CAC/B,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,CAAC,OAAO,CAAC,UAAU,CACnD,CAAC;IACF,MAAM,YAAY,GAAa,EAAE,CAAC;IAClC,IAAI,UAAmD,CAAC;IACxD,IAAI,KAAK,CAAC,UAAU,EAAE,CAAC;QACrB,UAAU,GAAG,EAAE,CAAC;QAChB,IAAI,SAAS,GAAG,CAAC,CAAC;QAClB,MAAM,WAAW,GAAa,EAAE,CAAC;QACjC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;YACxC,MAAM,cAAc,GAAG,OAAO,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC,UAAU,CAAC;YACrD,IAAI,cAAc,IAAI,cAAc,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;gBAChD,KAAK,MAAM,CAAC,IAAI,cAAc,EAAE,CAAC;oBAC/B,UAAU,CAAC,IAAI,CAAC;wBACd,IAAI,EAAE,CAAC,CAAC,IAAI;wBACZ,KAAK,EAAE,CAAC,CAAC,KAAK,GAAG,SAAS;wBAC1B,GAAG,EAAE,CAAC,CAAC,GAAG,GAAG,SAAS;wBACtB,SAAS,EAAE,CAAC;qBACb,CAAC,CAAC;gBACL,CAAC;YACH,CAAC;iBAAM,CAAC;gBACN,MAAM,QAAQ,GAAG,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC;gBAC5C,MAAM,MAAM,GAAG,QAAQ;qBACpB,KAAK,CAAC,aAAa,CAAC;qBACpB,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;gBAC/B,MAAM,OAAO,GAAG,aAAa,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;gBACtC,MAAM,MAAM,GAAG,8BAA8B,CAAC;oBAC5C,SAAS,EAAE,CAAC;oBACZ,UAAU,EAAE,MAAM,CAAC,MAAM;oBACzB,QAAQ,EAAE,SAAS;oBACnB,MAAM,EAAE,SAAS,GAAG,OAAO;oBAC3B,KAAK,EAAE,MAAM;iBACd,CAAC,CAAC;gBACH,UAAU,CAAC,IAAI,CAAC,GAAG,MAAM,CAAC,CAAC;gBAC3B,WAAW,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;YACtB,CAAC;YACD,SAAS,IAAI,CAAC,aAAa,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,UAAU,CAAC;QACpD,CAAC;QACD,IAAI,WAAW,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAC3B,YAAY,CAAC,IAAI,CACf,0DAA0D,WAAW,CAAC,IAAI,CAAC,GAAG,CAAC,6EAA6E,CAC7J,CAAC;QACJ,CAAC;QACD,KAAK,CACH,oBAAoB,UAAU,CAAC,MAAM,2BAA2B,OAAO,CAAC,MAAM,aAAa,WAAW,CAAC,MAAM,iCAAiC,CAC/I,CAAC;IACJ,CAAC;IAED,OAAO;QACL,KAAK,EAAE,UAAU;QACjB,SAAS;QACT,QAAQ,EAAE;YACR,UAAU,EAAE,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC;YAC9D,SAAS,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,GAAG,EAAE,GAAG,KAAK,CAAC;YAChD,eAAe;SAChB;QACD,eAAe;QACf,uBAAuB;QACvB,UAAU;QACV,QAAQ,EACN,QAAQ,CAAC,MAAM,GAAG,CAAC,IAAI,YAAY,CAAC,MAAM,GAAG,CAAC;YAC5C,CAAC,CAAC,CAAC,GAAG,QAAQ,EAAE,GAAG,YAAY,CAAC;YAChC,CAAC,CAAC,QAAQ;KACf,CAAC;AACJ,CAAC"}

package/dist/conversation/types.d.ts

@@ -1,55 +1,25 @@
+import type { AudioOutput } from "../audio-output.js";
+import type { PronunciationsFor } from "../pronunciations/types.js";
 import type { ResolvedModel, Voice } from "../speech-provider.js";
-import type { ResolvedSTTModel } from "../speech-to-text-provider.js";
-import type { TimestampMode } from "../timestamps.js";
 export interface ConversationTurn<V extends Voice = Voice> {
     readonly model?: string | ResolvedModel<V>;
     readonly providerOptions?: Record<string, unknown>;
     readonly text: string;
     readonly voice: V;
 }
-export interface GenerateConversationOptions<V extends Voice = Voice> {
+export interface GenerateConversationOptions<V extends Voice = Voice, M extends string | ResolvedModel<V> | undefined = string | ResolvedModel<V> | undefined> {
     readonly abortSignal?: AbortSignal;
     readonly apiKey?: string;
     readonly gapMs?: number;
     readonly headers?: Record<string, string>;
     readonly maxConcurrency?: number;
     readonly maxRetries?: number;
-    readonly model?: string | ResolvedModel<V>;
-    /**
-     * RMS-normalize the output audio to an absolute target level (see
-     * `volumeDbfs` for the level itself, default -20 dBFS — the broadcast /
-     * podcast voice standard). Every call to generateConversation produces
-     * output at the same loudness regardless of which providers or content
-     * are used, so two separate conversations can be played back-to-back
-     * without the listener adjusting volume. Roughly two O(N) passes over
-     * the int16 PCM samples — cheap. Pass `false` to skip the step entirely
-     * (~zero work) and keep the raw provider levels. Applied on both the
-     * stitch and native dialogue paths, provided the chosen provider
-     * exposes a decodable PCM/WAV mode via `getStitchOptions`. Default: true.
-     */
-    readonly normalizeVolume?: boolean;
+    readonly model?: M;
+    readonly output?: AudioOutput;
+    readonly pronunciations?: PronunciationsFor<M>;
     readonly providerOptions?: Record<string, unknown>;
-    /**
-     * Override the STT provider used for the derived-timestamps path. Construct
-     * via a factory (e.g. `createOpenAISTT({ apiKey })("whisper-1")`). Only
-     * consulted when the TTS provider can't supply timestamps natively. Defaults
-     * to OpenAI Whisper read from `OPENAI_API_KEY`.
-     */
-    readonly timestampProvider?: ResolvedSTTModel;
-    /**
-     * Controls whether the returned `SpeechResult` includes word-level
-     * timestamps. Default `"auto"`. On the stitch path each turn's timestamps
-     * are offset by cumulative duration + gap and concatenated flat; on the
-     * native path the mixed audio yields a flat list without speaker labels.
-     */
-    readonly timestamps?: TimestampMode;
+    readonly timestamps?: boolean;
     readonly turns: readonly ConversationTurn<V>[];
-    /**
-     * Target loudness in dBFS for `normalizeVolume`. Must be ≤ 0 (0 dBFS is
-     * the int16 ceiling). Lower values are quieter — -20 leaves ~20 dB of
-     * peak headroom so typical TTS speech doesn't clip after gain. Ignored
-     * when `normalizeVolume` is `false`. Default: -20.
-     */
     readonly volumeDbfs?: number;
 }
 //# sourceMappingURL=types.d.ts.map

package/dist/conversation/types.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/conversation/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,KAAK,EAAE,MAAM,uBAAuB,CAAC;AAClE,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,+BAA+B,CAAC;AACtE,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAEtD,MAAM,WAAW,gBAAgB,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK;IACvD,QAAQ,CAAC,KAAK,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC,CAAC,CAAC,CAAC;IAC3C,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACnD,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,KAAK,EAAE,CAAC,CAAC;CACnB;AAED,MAAM,WAAW,2BAA2B,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK;IAClE,QAAQ,CAAC,WAAW,CAAC,EAAE,WAAW,CAAC;IACnC,QAAQ,CAAC,MAAM,CAAC,EAAE,MAAM,CAAC;IACzB,QAAQ,CAAC,KAAK,CAAC,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAC1C,QAAQ,CAAC,cAAc,CAAC,EAAE,MAAM,CAAC;IACjC,QAAQ,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,KAAK,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC,CAAC,CAAC,CAAC;IAC3C;;;;;;;;;;;OAWG;IACH,QAAQ,CAAC,eAAe,CAAC,EAAE,OAAO,CAAC;IACnC,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACnD;;;;;OAKG;IACH,QAAQ,CAAC,iBAAiB,CAAC,EAAE,gBAAgB,CAAC;IAC9C;;;;;OAKG;IACH,QAAQ,CAAC,UAAU,CAAC,EAAE,aAAa,CAAC;IACpC,QAAQ,CAAC,KAAK,EAAE,SAAS,gBAAgB,CAAC,CAAC,CAAC,EAAE,CAAC;IAC/C;;;;;OAKG;IACH,QAAQ,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC;CAC9B"}
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/conversation/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,oBAAoB,CAAC;AACtD,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,4BAA4B,CAAC;AACpE,OAAO,KAAK,EAAE,aAAa,EAAE,KAAK,EAAE,MAAM,uBAAuB,CAAC;AAElE,MAAM,WAAW,gBAAgB,CAAC,CAAC,SAAS,KAAK,GAAG,KAAK;IACvD,QAAQ,CAAC,KAAK,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC,CAAC,CAAC,CAAC;IAC3C,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACnD,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,KAAK,EAAE,CAAC,CAAC;CACnB;AAED,MAAM,WAAW,2BAA2B,CAC1C,CAAC,SAAS,KAAK,GAAG,KAAK,EACvB,CAAC,SAAS,MAAM,GAAG,aAAa,CAAC,CAAC,CAAC,GAAG,SAAS,GAC3C,MAAM,GACN,aAAa,CAAC,CAAC,CAAC,GAChB,SAAS;IAEb,QAAQ,CAAC,WAAW,CAAC,EAAE,WAAW,CAAC;IACnC,QAAQ,CAAC,MAAM,CAAC,EAAE,MAAM,CAAC;IACzB,QAAQ,CAAC,KAAK,CAAC,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAC1C,QAAQ,CAAC,cAAc,CAAC,EAAE,MAAM,CAAC;IACjC,QAAQ,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IACnB,QAAQ,CAAC,MAAM,CAAC,EAAE,WAAW,CAAC;IAC9B,QAAQ,CAAC,cAAc,CAAC,EAAE,iBAAiB,CAAC,CAAC,CAAC,CAAC;IAC/C,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACnD,QAAQ,CAAC,UAAU,CAAC,EAAE,OAAO,CAAC;IAC9B,QAAQ,CAAC,KAAK,EAAE,SAAS,gBAAgB,CAAC,CAAC,CAAC,EAAE,CAAC;IAE/C,QAAQ,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC;CAC9B"}

package/dist/conversation/validate.d.ts

@@ -1,19 +1,4 @@
 import type { ConversationTurn, GenerateConversationOptions } from "./types.js";
-/**
- * Stable key for a voice so we can count unique voices across turns within
- * one call. String voices and URL voices use their value; binary
- * `Uint8Array` audio voices use object-reference identity (two distinct
- * buffers with the same length/endpoints would otherwise collide).
- */
-export declare function voiceKey(voice: ConversationTurn["voice"], refIds: WeakMap<object, number>, refCounter: {
-    next: number;
-}): string;
-/** Build a fresh ref-id context for a single conversation. */
-export declare function newVoiceKeyContext(): {
-    refIds: WeakMap<object, number>;
-    refCounter: {
-        next: number;
-    };
-};
+export declare function newVoiceKeyer(): (voice: ConversationTurn["voice"]) => string;
 export declare function validateConversationInput(options: GenerateConversationOptions): void;
 //# sourceMappingURL=validate.d.ts.map

package/dist/conversation/validate.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"validate.d.ts","sourceRoot":"","sources":["../../src/conversation/validate.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,gBAAgB,EAAE,2BAA2B,EAAE,MAAM,YAAY,CAAC;AAEhF;;;;;GAKG;AACH,wBAAgB,QAAQ,CACtB,KAAK,EAAE,gBAAgB,CAAC,OAAO,CAAC,EAChC,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,MAAM,CAAC,EAC/B,UAAU,EAAE;IAAE,IAAI,EAAE,MAAM,CAAA;CAAE,GAC3B,MAAM,CAiBR;AAED,8DAA8D;AAC9D,wBAAgB,kBAAkB,IAAI;IACpC,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAChC,UAAU,EAAE;QAAE,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;CAC9B,CAEA;AAED,wBAAgB,yBAAyB,CACvC,OAAO,EAAE,2BAA2B,GACnC,IAAI,CAkBN"}
+{"version":3,"file":"validate.d.ts","sourceRoot":"","sources":["../../src/conversation/validate.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,gBAAgB,EAAE,2BAA2B,EAAE,MAAM,YAAY,CAAC;AAGhF,wBAAgB,aAAa,IAAI,CAAC,KAAK,EAAE,gBAAgB,CAAC,OAAO,CAAC,KAAK,MAAM,CAoB5E;AAED,wBAAgB,yBAAyB,CACvC,OAAO,EAAE,2BAA2B,GACnC,IAAI,CA2BN"}

package/dist/conversation/validate.js

@@ -1,43 +1,43 @@
 import { ConversationInputError } from "./errors.js";
-/**
- * Stable key for a voice so we can count unique voices across turns within
- * one call. String voices and URL voices use their value; binary
- * `Uint8Array` audio voices use object-reference identity (two distinct
- * buffers with the same length/endpoints would otherwise collide).
- */
-export function voiceKey(voice, refIds, refCounter) {
-    if (typeof voice === "string") {
-        return `s:${voice}`;
-    }
-    if ("url" in voice) {
-        return `u:${voice.url}`;
-    }
-    if ("audio" in voice && typeof voice.audio === "string") {
-        return `a:${voice.audio}`;
-    }
-    // Object voice (Uint8Array audio, or any other shape) — key by reference.
-    let id = refIds.get(voice);
-    if (id === undefined) {
-        id = refCounter.next++;
-        refIds.set(voice, id);
-    }
-    return `o:${id}`;
-}
-/** Build a fresh ref-id context for a single conversation. */
-export function newVoiceKeyContext() {
-    return { refIds: new WeakMap(), refCounter: { next: 0 } };
+// Object voices key by reference — distinct buffers with identical content must not collide.
+export function newVoiceKeyer() {
+    const refIds = new WeakMap();
+    let nextId = 0;
+    return (voice) => {
+        if (typeof voice === "string") {
+            return `s:${voice}`;
+        }
+        if ("url" in voice) {
+            return `u:${voice.url}`;
+        }
+        if ("audio" in voice && typeof voice.audio === "string") {
+            return `a:${voice.audio}`;
+        }
+        let id = refIds.get(voice);
+        if (id === undefined) {
+            id = nextId++;
+            refIds.set(voice, id);
+        }
+        return `o:${id}`;
+    };
 }
 export function validateConversationInput(options) {
     if (options.turns.length === 0) {
         throw new ConversationInputError("generateConversation requires at least one turn.");
     }
+    // Model placement must be all-or-nothing — partial mix hides which model actually ran where.
+    const hasTopLevel = options.model != null;
     for (let i = 0; i < options.turns.length; i++) {
         const turn = options.turns[i];
         if (turn.text.trim().length === 0) {
             throw new ConversationInputError(`turns[${i}].text must not be empty.`);
         }
-        if (options.model == null && turn.model == null) {
-            throw new ConversationInputError(`turns[${i}]: model must be set, either at top-level or on the turn.`);
+        const hasTurnModel = turn.model != null;
+        if (hasTopLevel && hasTurnModel) {
+            throw new ConversationInputError(`turns[${i}].model is set, but options.model is also set. Set the model either at the top level for all turns, or on every turn — not both.`);
+        }
+        if (!(hasTopLevel || hasTurnModel)) {
+            throw new ConversationInputError(`turns[${i}].model is required because options.model is not set. Either set options.model for all turns, or set model on every turn.`);
         }
     }
 }

package/dist/conversation/validate.js.map

@@ -1 +1 @@
-{"version":3,"file":"validate.js","sourceRoot":"","sources":["../../src/conversation/validate.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,sBAAsB,EAAE,MAAM,aAAa,CAAC;AAGrD;;;;;GAKG;AACH,MAAM,UAAU,QAAQ,CACtB,KAAgC,EAChC,MAA+B,EAC/B,UAA4B;IAE5B,IAAI,OAAO,KAAK,KAAK,QAAQ,EAAE,CAAC;QAC9B,OAAO,KAAK,KAAK,EAAE,CAAC;IACtB,CAAC;IACD,IAAI,KAAK,IAAI,KAAK,EAAE,CAAC;QACnB,OAAO,KAAK,KAAK,CAAC,GAAG,EAAE,CAAC;IAC1B,CAAC;IACD,IAAI,OAAO,IAAI,KAAK,IAAI,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;QACxD,OAAO,KAAK,KAAK,CAAC,KAAK,EAAE,CAAC;IAC5B,CAAC;IACD,0EAA0E;IAC1E,IAAI,EAAE,GAAG,MAAM,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;IAC3B,IAAI,EAAE,KAAK,SAAS,EAAE,CAAC;QACrB,EAAE,GAAG,UAAU,CAAC,IAAI,EAAE,CAAC;QACvB,MAAM,CAAC,GAAG,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;IACxB,CAAC;IACD,OAAO,KAAK,EAAE,EAAE,CAAC;AACnB,CAAC;AAED,8DAA8D;AAC9D,MAAM,UAAU,kBAAkB;IAIhC,OAAO,EAAE,MAAM,EAAE,IAAI,OAAO,EAAE,EAAE,UAAU,EAAE,EAAE,IAAI,EAAE,CAAC,EAAE,EAAE,CAAC;AAC5D,CAAC;AAED,MAAM,UAAU,yBAAyB,CACvC,OAAoC;IAEpC,IAAI,OAAO,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC/B,MAAM,IAAI,sBAAsB,CAC9B,kDAAkD,CACnD,CAAC;IACJ,CAAC;IAED,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,OAAO,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QAC9C,MAAM,IAAI,GAAG,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAC9B,IAAI,IAAI,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAClC,MAAM,IAAI,sBAAsB,CAAC,SAAS,CAAC,2BAA2B,CAAC,CAAC;QAC1E,CAAC;QACD,IAAI,OAAO,CAAC,KAAK,IAAI,IAAI,IAAI,IAAI,CAAC,KAAK,IAAI,IAAI,EAAE,CAAC;YAChD,MAAM,IAAI,sBAAsB,CAC9B,SAAS,CAAC,2DAA2D,CACtE,CAAC;QACJ,CAAC;IACH,CAAC;AACH,CAAC"}
+{"version":3,"file":"validate.js","sourceRoot":"","sources":["../../src/conversation/validate.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,sBAAsB,EAAE,MAAM,aAAa,CAAC;AAGrD,6FAA6F;AAC7F,MAAM,UAAU,aAAa;IAC3B,MAAM,MAAM,GAAG,IAAI,OAAO,EAAkB,CAAC;IAC7C,IAAI,MAAM,GAAG,CAAC,CAAC;IACf,OAAO,CAAC,KAAK,EAAE,EAAE;QACf,IAAI,OAAO,KAAK,KAAK,QAAQ,EAAE,CAAC;YAC9B,OAAO,KAAK,KAAK,EAAE,CAAC;QACtB,CAAC;QACD,IAAI,KAAK,IAAI,KAAK,EAAE,CAAC;YACnB,OAAO,KAAK,KAAK,CAAC,GAAG,EAAE,CAAC;QAC1B,CAAC;QACD,IAAI,OAAO,IAAI,KAAK,IAAI,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;YACxD,OAAO,KAAK,KAAK,CAAC,KAAK,EAAE,CAAC;QAC5B,CAAC;QACD,IAAI,EAAE,GAAG,MAAM,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;QAC3B,IAAI,EAAE,KAAK,SAAS,EAAE,CAAC;YACrB,EAAE,GAAG,MAAM,EAAE,CAAC;YACd,MAAM,CAAC,GAAG,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;QACxB,CAAC;QACD,OAAO,KAAK,EAAE,EAAE,CAAC;IACnB,CAAC,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,yBAAyB,CACvC,OAAoC;IAEpC,IAAI,OAAO,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC/B,MAAM,IAAI,sBAAsB,CAC9B,kDAAkD,CACnD,CAAC;IACJ,CAAC;IAED,6FAA6F;IAC7F,MAAM,WAAW,GAAG,OAAO,CAAC,KAAK,IAAI,IAAI,CAAC;IAE1C,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,OAAO,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QAC9C,MAAM,IAAI,GAAG,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAC9B,IAAI,IAAI,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAClC,MAAM,IAAI,sBAAsB,CAAC,SAAS,CAAC,2BAA2B,CAAC,CAAC;QAC1E,CAAC;QACD,MAAM,YAAY,GAAG,IAAI,CAAC,KAAK,IAAI,IAAI,CAAC;QACxC,IAAI,WAAW,IAAI,YAAY,EAAE,CAAC;YAChC,MAAM,IAAI,sBAAsB,CAC9B,SAAS,CAAC,kIAAkI,CAC7I,CAAC;QACJ,CAAC;QACD,IAAI,CAAC,CAAC,WAAW,IAAI,YAAY,CAAC,EAAE,CAAC;YACnC,MAAM,IAAI,sBAAsB,CAC9B,SAAS,CAAC,2HAA2H,CACtI,CAAC;QACJ,CAAC;IACH,CAAC;AACH,CAAC"}

package/dist/default-stt-fallback.d.ts

@@ -0,0 +1,3 @@
+import type { ResolvedSTTModel } from "./speech-to-text-provider.js";
+export declare function getDefaultSTTFallback(): Promise<ResolvedSTTModel>;
+//# sourceMappingURL=default-stt-fallback.d.ts.map

package/dist/default-stt-fallback.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"default-stt-fallback.d.ts","sourceRoot":"","sources":["../src/default-stt-fallback.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,8BAA8B,CAAC;AAKrE,wBAAsB,qBAAqB,IAAI,OAAO,CAAC,gBAAgB,CAAC,CAOvE"}

package/dist/default-stt-fallback.js

@@ -0,0 +1,11 @@
+let cached;
+// Dynamic import keeps the OpenAI provider out of bundles for callers who never trigger the STT fallback.
+export async function getDefaultSTTFallback() {
+    if (cached) {
+        return cached;
+    }
+    const { createOpenAI } = await import("./providers/openai/index.js");
+    cached = createOpenAI().stt("whisper-1");
+    return cached;
+}
+//# sourceMappingURL=default-stt-fallback.js.map

package/dist/default-stt-fallback.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"default-stt-fallback.js","sourceRoot":"","sources":["../src/default-stt-fallback.ts"],"names":[],"mappings":"AAEA,IAAI,MAAoC,CAAC;AAEzC,0GAA0G;AAC1G,MAAM,CAAC,KAAK,UAAU,qBAAqB;IACzC,IAAI,MAAM,EAAE,CAAC;QACX,OAAO,MAAM,CAAC;IAChB,CAAC;IACD,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,MAAM,CAAC,6BAA6B,CAAC,CAAC;IACrE,MAAM,GAAG,YAAY,EAAE,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC;IACzC,OAAO,MAAM,CAAC;AAChB,CAAC"}

package/dist/derive-timestamps.d.ts

@@ -1,14 +1,10 @@
 import type { ResolvedSTTModel } from "./speech-to-text-provider.js";
 import type { WordTimestamp } from "./timestamps.js";
-/**
- * Pipes synthesized audio through an STT provider to produce word-level
- * timestamps. Shared between `generateSpeech()` and conversation paths.
- */
 export declare function deriveTimestampsViaSTT(args: {
     ttsModel: string;
     audio: Uint8Array;
     mediaType: string;
-    timestampProvider: ResolvedSTTModel | undefined;
+    timestampFallback: ResolvedSTTModel;
     abortSignal: AbortSignal | undefined;
 }): Promise<readonly WordTimestamp[]>;
 //# sourceMappingURL=derive-timestamps.d.ts.map

package/dist/derive-timestamps.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"derive-timestamps.d.ts","sourceRoot":"","sources":["../src/derive-timestamps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,8BAA8B,CAAC;AAErE,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAYrD;;;GAGG;AACH,wBAAsB,sBAAsB,CAAC,IAAI,EAAE;IACjD,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,EAAE,UAAU,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,iBAAiB,EAAE,gBAAgB,GAAG,SAAS,CAAC;IAChD,WAAW,EAAE,WAAW,GAAG,SAAS,CAAC;CACtC,GAAG,OAAO,CAAC,SAAS,aAAa,EAAE,CAAC,CAqBpC"}
+{"version":3,"file":"derive-timestamps.d.ts","sourceRoot":"","sources":["../src/derive-timestamps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,8BAA8B,CAAC;AACrE,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAErD,wBAAsB,sBAAsB,CAAC,IAAI,EAAE;IACjD,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,EAAE,UAAU,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,iBAAiB,EAAE,gBAAgB,CAAC;IACpC,WAAW,EAAE,WAAW,GAAG,SAAS,CAAC;CACtC,GAAG,OAAO,CAAC,SAAS,aAAa,EAAE,CAAC,CAqBpC"}

package/dist/derive-timestamps.js

@@ -1,20 +1,6 @@
 import { MissingApiKeyError, TimestampKeyMissingError } from "./errors.js";
-import { OpenAISpeechToTextProvider } from "./stt-providers/openai/index.js";
-/**
- * Default STT model used on the derived-timestamps path when the caller
- * hasn't supplied a `timestampProvider` override. Reads `OPENAI_API_KEY`
- * from the environment via the provider's own key resolution.
- */
-function defaultTimestampProvider() {
-    const provider = new OpenAISpeechToTextProvider();
-    return { provider, modelId: provider.defaultModel };
-}
-/**
- * Pipes synthesized audio through an STT provider to produce word-level
- * timestamps. Shared between `generateSpeech()` and conversation paths.
- */
 export async function deriveTimestampsViaSTT(args) {
-    const sttModel = args.timestampProvider ?? defaultTimestampProvider();
+    const sttModel = args.timestampFallback;
     try {
         const { timestamps } = await sttModel.provider.transcribe({
             modelId: sttModel.modelId,

package/dist/derive-timestamps.js.map

@@ -1 +1 @@
-{"version":3,"file":"derive-timestamps.js","sourceRoot":"","sources":["../src/derive-timestamps.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,wBAAwB,EAAE,MAAM,aAAa,CAAC;AAE3E,OAAO,EAAE,0BAA0B,EAAE,MAAM,iCAAiC,CAAC;AAG7E;;;;GAIG;AACH,SAAS,wBAAwB;IAC/B,MAAM,QAAQ,GAAG,IAAI,0BAA0B,EAAE,CAAC;IAClD,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,QAAQ,CAAC,YAAY,EAAE,CAAC;AACtD,CAAC;AAED;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,sBAAsB,CAAC,IAM5C;IACC,MAAM,QAAQ,GAAG,IAAI,CAAC,iBAAiB,IAAI,wBAAwB,EAAE,CAAC;IAEtE,IAAI,CAAC;QACH,MAAM,EAAE,UAAU,EAAE,GAAG,MAAM,QAAQ,CAAC,QAAQ,CAAC,UAAU,CAAC;YACxD,OAAO,EAAE,QAAQ,CAAC,OAAO;YACzB,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,SAAS,EAAE,IAAI,CAAC,SAAS;YACzB,WAAW,EAAE,IAAI,CAAC,WAAW;SAC9B,CAAC,CAAC;QACH,OAAO,UAAU,CAAC;IACpB,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,IAAI,GAAG,YAAY,kBAAkB,EAAE,CAAC;YACtC,MAAM,IAAI,wBAAwB,CAAC;gBACjC,QAAQ,EAAE,IAAI,CAAC,QAAQ;gBACvB,WAAW,EAAE,GAAG,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,EAAE;gBAC1D,MAAM,EAAE,GAAG,CAAC,MAAM;aACnB,CAAC,CAAC;QACL,CAAC;QACD,MAAM,GAAG,CAAC;IACZ,CAAC;AACH,CAAC"}
+{"version":3,"file":"derive-timestamps.js","sourceRoot":"","sources":["../src/derive-timestamps.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,wBAAwB,EAAE,MAAM,aAAa,CAAC;AAI3E,MAAM,CAAC,KAAK,UAAU,sBAAsB,CAAC,IAM5C;IACC,MAAM,QAAQ,GAAG,IAAI,CAAC,iBAAiB,CAAC;IAExC,IAAI,CAAC;QACH,MAAM,EAAE,UAAU,EAAE,GAAG,MAAM,QAAQ,CAAC,QAAQ,CAAC,UAAU,CAAC;YACxD,OAAO,EAAE,QAAQ,CAAC,OAAO;YACzB,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,SAAS,EAAE,IAAI,CAAC,SAAS;YACzB,WAAW,EAAE,IAAI,CAAC,WAAW;SAC9B,CAAC,CAAC;QACH,OAAO,UAAU,CAAC;IACpB,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,IAAI,GAAG,YAAY,kBAAkB,EAAE,CAAC;YACtC,MAAM,IAAI,wBAAwB,CAAC;gBACjC,QAAQ,EAAE,IAAI,CAAC,QAAQ;gBACvB,WAAW,EAAE,GAAG,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,EAAE;gBAC1D,MAAM,EAAE,GAAG,CAAC,MAAM;aACnB,CAAC,CAAC;QACL,CAAC;QACD,MAAM,GAAG,CAAC;IACZ,CAAC;AACH,CAAC"}

package/dist/encoders/mp3.d.ts

@@ -0,0 +1,6 @@
+export declare function encodePcm16ToMp3(args: {
+    readonly pcm: Uint8Array;
+    readonly sampleRate: number;
+    readonly bitrateKbps: number;
+}): Promise<Uint8Array>;
+//# sourceMappingURL=mp3.d.ts.map

package/dist/encoders/mp3.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"mp3.d.ts","sourceRoot":"","sources":["../../src/encoders/mp3.ts"],"names":[],"mappings":"AA4BA,wBAAsB,gBAAgB,CAAC,IAAI,EAAE;IAC3C,QAAQ,CAAC,GAAG,EAAE,UAAU,CAAC;IACzB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;CAC9B,GAAG,OAAO,CAAC,UAAU,CAAC,CAwCtB"}

package/dist/encoders/mp3.js

@@ -0,0 +1,54 @@
+import { AudioSample, AudioSampleSource, BufferTarget, canEncodeAudio, Mp3OutputFormat, Output, } from "mediabunny";
+const SUPPORTED_SAMPLE_RATES = new Set([
+    16_000, 22_050, 24_000, 32_000, 44_100, 48_000,
+]);
+let mp3EncoderRegistered;
+async function ensureMp3Encoder() {
+    if (!mp3EncoderRegistered) {
+        mp3EncoderRegistered = (async () => {
+            if (await canEncodeAudio("mp3")) {
+                return;
+            }
+            const { registerMp3Encoder } = await import("@mediabunny/mp3-encoder");
+            registerMp3Encoder();
+        })();
+    }
+    await mp3EncoderRegistered;
+}
+export async function encodePcm16ToMp3(args) {
+    const { pcm, sampleRate, bitrateKbps } = args;
+    if (pcm.byteLength === 0) {
+        throw new Error("encodePcm16ToMp3: input pcm is empty");
+    }
+    if (!SUPPORTED_SAMPLE_RATES.has(sampleRate)) {
+        throw new Error(`encodePcm16ToMp3: unsupported sample rate ${sampleRate}`);
+    }
+    await ensureMp3Encoder();
+    const output = new Output({
+        format: new Mp3OutputFormat(),
+        target: new BufferTarget(),
+    });
+    const source = new AudioSampleSource({
+        codec: "mp3",
+        bitrate: bitrateKbps * 1000,
+    });
+    output.addAudioTrack(source);
+    await output.start();
+    const numberOfFrames = Math.floor(pcm.byteLength / 2);
+    const sample = new AudioSample({
+        data: new Int16Array(pcm.buffer, pcm.byteOffset, numberOfFrames),
+        format: "s16",
+        numberOfChannels: 1,
+        sampleRate,
+        timestamp: 0,
+    });
+    await source.add(sample);
+    source.close();
+    await output.finalize();
+    const buffer = output.target.buffer;
+    if (!buffer) {
+        throw new Error("encodePcm16ToMp3: Mp3OutputFormat produced no buffer");
+    }
+    return new Uint8Array(buffer);
+}
+//# sourceMappingURL=mp3.js.map

package/dist/encoders/mp3.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"mp3.js","sourceRoot":"","sources":["../../src/encoders/mp3.ts"],"names":[],"mappings":"AAAA,OAAO,EACL,WAAW,EACX,iBAAiB,EACjB,YAAY,EACZ,cAAc,EACd,eAAe,EACf,MAAM,GACP,MAAM,YAAY,CAAC;AAEpB,MAAM,sBAAsB,GAAG,IAAI,GAAG,CAAC;IACrC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM;CAC/C,CAAC,CAAC;AAEH,IAAI,oBAA+C,CAAC;AAEpD,KAAK,UAAU,gBAAgB;IAC7B,IAAI,CAAC,oBAAoB,EAAE,CAAC;QAC1B,oBAAoB,GAAG,CAAC,KAAK,IAAI,EAAE;YACjC,IAAI,MAAM,cAAc,CAAC,KAAK,CAAC,EAAE,CAAC;gBAChC,OAAO;YACT,CAAC;YACD,MAAM,EAAE,kBAAkB,EAAE,GAAG,MAAM,MAAM,CAAC,yBAAyB,CAAC,CAAC;YACvE,kBAAkB,EAAE,CAAC;QACvB,CAAC,CAAC,EAAE,CAAC;IACP,CAAC;IACD,MAAM,oBAAoB,CAAC;AAC7B,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,gBAAgB,CAAC,IAItC;IACC,MAAM,EAAE,GAAG,EAAE,UAAU,EAAE,WAAW,EAAE,GAAG,IAAI,CAAC;IAE9C,IAAI,GAAG,CAAC,UAAU,KAAK,CAAC,EAAE,CAAC;QACzB,MAAM,IAAI,KAAK,CAAC,sCAAsC,CAAC,CAAC;IAC1D,CAAC;IACD,IAAI,CAAC,sBAAsB,CAAC,GAAG,CAAC,UAAU,CAAC,EAAE,CAAC;QAC5C,MAAM,IAAI,KAAK,CAAC,6CAA6C,UAAU,EAAE,CAAC,CAAC;IAC7E,CAAC;IAED,MAAM,gBAAgB,EAAE,CAAC;IAEzB,MAAM,MAAM,GAAG,IAAI,MAAM,CAAC;QACxB,MAAM,EAAE,IAAI,eAAe,EAAE;QAC7B,MAAM,EAAE,IAAI,YAAY,EAAE;KAC3B,CAAC,CAAC;IACH,MAAM,MAAM,GAAG,IAAI,iBAAiB,CAAC;QACnC,KAAK,EAAE,KAAK;QACZ,OAAO,EAAE,WAAW,GAAG,IAAI;KAC5B,CAAC,CAAC;IACH,MAAM,CAAC,aAAa,CAAC,MAAM,CAAC,CAAC;IAC7B,MAAM,MAAM,CAAC,KAAK,EAAE,CAAC;IAErB,MAAM,cAAc,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,UAAU,GAAG,CAAC,CAAC,CAAC;IACtD,MAAM,MAAM,GAAG,IAAI,WAAW,CAAC;QAC7B,IAAI,EAAE,IAAI,UAAU,CAAC,GAAG,CAAC,MAAM,EAAE,GAAG,CAAC,UAAU,EAAE,cAAc,CAAC;QAChE,MAAM,EAAE,KAAK;QACb,gBAAgB,EAAE,CAAC;QACnB,UAAU;QACV,SAAS,EAAE,CAAC;KACb,CAAC,CAAC;IACH,MAAM,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;IACzB,MAAM,CAAC,KAAK,EAAE,CAAC;IAEf,MAAM,MAAM,CAAC,QAAQ,EAAE,CAAC;IACxB,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC;IACpC,IAAI,CAAC,MAAM,EAAE,CAAC;QACZ,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;IAC1E,CAAC;IACD,OAAO,IAAI,UAAU,CAAC,MAAM,CAAC,CAAC;AAChC,CAAC"}