@lssm-tech/lib.voice 1.7.29 → 3.0.2

package/dist/stt/segment-splitter.js

@@ -1,2 +0,0 @@
-// @bun
-class F{static DEFAULT_MAX_CHUNK_MS=300000;split(j,z=F.DEFAULT_MAX_CHUNK_MS){let q=j.durationMs??this.estimateDurationMs(j);if(q<=z)return[j];let w=[],E=j.data.length/Math.max(q,1),v=0;while(v<q){let A=Math.min(z,q-v),G=Math.floor(v*E),H=Math.floor((v+A)*E);w.push({data:j.data.slice(G,H),format:j.format,sampleRateHz:j.sampleRateHz,durationMs:A,channels:j.channels}),v+=A}return w}estimateDurationMs(j){let q=j.channels??1,w=j.data.length/(2*q);return Math.ceil(w/j.sampleRateHz*1000)}}export{F as SegmentSplitter};

package/dist/stt/subtitle-formatter.d.ts

@@ -1,19 +0,0 @@
-import type { TranscriptionSegment } from './types';
-/**
- * Format transcription segments as SRT or VTT subtitles.
- */
-export declare class SubtitleFormatter {
-    /**
-     * Convert segments to SRT format.
-     */
-    toSRT(segments: TranscriptionSegment[]): string;
-    /**
-     * Convert segments to WebVTT format.
-     */
-    toVTT(segments: TranscriptionSegment[]): string;
-    /** Format ms as SRT timestamp: HH:MM:SS,mmm */
-    private formatTimeSRT;
-    /** Format ms as VTT timestamp: HH:MM:SS.mmm */
-    private formatTimeVTT;
-    private pad;
-}

package/dist/stt/subtitle-formatter.js

@@ -1,12 +0,0 @@
-// @bun
-class A{toSRT(f){return f.map((j,w)=>{let k=this.formatTimeSRT(j.startMs),q=this.formatTimeSRT(j.endMs),x=j.speakerLabel?`[${j.speakerLabel}] `:"";return`${w+1}
-${k} --> ${q}
-${x}${j.text}`}).join(`
-
-`)}toVTT(f){return`WEBVTT
-
-`+f.map((k,q)=>{let x=this.formatTimeVTT(k.startMs),y=this.formatTimeVTT(k.endMs),z=k.speakerLabel?`<v ${k.speakerLabel}>`:"";return`${q+1}
-${x} --> ${y}
-${z}${k.text}`}).join(`
-
-`)}formatTimeSRT(f){let j=Math.floor(f/3600000),w=Math.floor(f%3600000/60000),k=Math.floor(f%60000/1000),q=f%1000;return`${this.pad(j,2)}:${this.pad(w,2)}:${this.pad(k,2)},${this.pad(q,3)}`}formatTimeVTT(f){let j=Math.floor(f/3600000),w=Math.floor(f%3600000/60000),k=Math.floor(f%60000/1000),q=f%1000;return`${this.pad(j,2)}:${this.pad(w,2)}:${this.pad(k,2)}.${this.pad(q,3)}`}pad(f,j){return f.toString().padStart(j,"0")}}export{A as SubtitleFormatter};

package/dist/stt/transcriber.d.ts

@@ -1,21 +0,0 @@
-import type { STTBrief, STTOptions, TranscriptionProject, TranscriptionSegment } from './types';
-/**
- * Main STT orchestrator.
- *
- * Pipeline:
- * 1. Split audio into processable chunks (if needed)
- * 2. Transcribe via STTProvider
- * 3. Map speaker IDs to labels (if diarization enabled)
- * 4. Format subtitles (if requested)
- */
-export declare class Transcriber {
-    private readonly stt;
-    private readonly segmentSplitter;
-    private readonly diarizationMapper;
-    private readonly subtitleFormatter;
-    constructor(options: STTOptions);
-    /** Transcribe audio to text */
-    transcribe(brief: STTBrief): Promise<TranscriptionProject>;
-    /** Stream transcription (real-time, if provider supports it) */
-    transcribeStream(audio: AsyncIterable<Uint8Array>, options?: Partial<STTBrief>): AsyncIterable<TranscriptionSegment>;
-}

package/dist/stt/transcriber.js

@@ -1,12 +0,0 @@
-// @bun
-class Z{map(q,A="Speaker"){let G=[],w=new Map;for(let E of q)if(E.speakerId&&!G.includes(E.speakerId))G.push(E.speakerId),w.set(E.speakerId,{segmentCount:0,totalSpeakingMs:0});let J=q.map((E)=>{if(!E.speakerId)return E;let L=G.indexOf(E.speakerId),N=`${A} ${L+1}`,R=w.get(E.speakerId);if(!R)return{...E,speakerLabel:N};return R.segmentCount+=1,R.totalSpeakingMs+=E.endMs-E.startMs,{...E,speakerLabel:N}}),K=G.map((E,L)=>{let N=w.get(E);return{id:E,label:`${A} ${L+1}`,segmentCount:N?.segmentCount??0,totalSpeakingMs:N?.totalSpeakingMs??0}});return{segments:J,speakers:K}}}class W{static DEFAULT_MAX_CHUNK_MS=300000;split(q,A=W.DEFAULT_MAX_CHUNK_MS){let G=q.durationMs??this.estimateDurationMs(q);if(G<=A)return[q];let w=[],J=q.data.length/Math.max(G,1),K=0;while(K<G){let E=Math.min(A,G-K),L=Math.floor(K*J),N=Math.floor((K+E)*J);w.push({data:q.data.slice(L,N),format:q.format,sampleRateHz:q.sampleRateHz,durationMs:E,channels:q.channels}),K+=E}return w}estimateDurationMs(q){let G=q.channels??1,w=q.data.length/(2*G);return Math.ceil(w/q.sampleRateHz*1000)}}class _{toSRT(q){return q.map((A,G)=>{let w=this.formatTimeSRT(A.startMs),J=this.formatTimeSRT(A.endMs),K=A.speakerLabel?`[${A.speakerLabel}] `:"";return`${G+1}
-${w} --> ${J}
-${K}${A.text}`}).join(`
-
-`)}toVTT(q){return`WEBVTT
-
-`+q.map((w,J)=>{let K=this.formatTimeVTT(w.startMs),E=this.formatTimeVTT(w.endMs),L=w.speakerLabel?`<v ${w.speakerLabel}>`:"";return`${J+1}
-${K} --> ${E}
-${L}${w.text}`}).join(`
-
-`)}formatTimeSRT(q){let A=Math.floor(q/3600000),G=Math.floor(q%3600000/60000),w=Math.floor(q%60000/1000),J=q%1000;return`${this.pad(A,2)}:${this.pad(G,2)}:${this.pad(w,2)},${this.pad(J,3)}`}formatTimeVTT(q){let A=Math.floor(q/3600000),G=Math.floor(q%3600000/60000),w=Math.floor(q%60000/1000),J=q%1000;return`${this.pad(A,2)}:${this.pad(G,2)}:${this.pad(w,2)}.${this.pad(J,3)}`}pad(q,A){return q.toString().padStart(A,"0")}}class j{stt;segmentSplitter=new W;diarizationMapper=new Z;subtitleFormatter=new _;constructor(q){this.stt=q.stt}async transcribe(q){let A=z(),G=this.segmentSplitter.split(q.audio),w=[],J="",K=0,E=0;for(let U of G){let Y=await this.stt.transcribe({audio:U,language:q.language,diarize:q.diarize,speakerCount:q.speakerCount,wordTimestamps:!0,vocabularyHints:q.vocabularyHints}),V=Y.segments.map((Q)=>({text:Q.text,startMs:Q.startMs+E,endMs:Q.endMs+E,speakerId:Q.speakerId,speakerName:Q.speakerName,confidence:Q.confidence}));w.push(...V),J+=(J?" ":"")+Y.text,K+=Y.durationMs,E+=U.durationMs??0}let L=w,N;if(q.diarize){let U=this.diarizationMapper.map(w);L=U.segments,N=U.speakers}let R={text:J,segments:L,language:q.language??"en",durationMs:K},X,$=q.subtitleFormat??"none";if($==="srt")X=this.subtitleFormatter.toSRT(L);else if($==="vtt")X=this.subtitleFormatter.toVTT(L);return{id:A,transcript:R,subtitles:X,speakers:N}}async*transcribeStream(q,A){if(!this.stt.transcribeStream)throw Error("Streaming transcription not supported by the current STT provider");let G=this.stt.transcribeStream(q,{language:A?.language,diarize:A?.diarize,speakerCount:A?.speakerCount,wordTimestamps:!0,vocabularyHints:A?.vocabularyHints});for await(let w of G)yield{text:w.text,startMs:w.startMs,endMs:w.endMs,speakerId:w.speakerId,speakerLabel:w.speakerName,confidence:w.confidence}}}function z(){let q=Date.now().toString(36),A=Math.random().toString(36).slice(2,8);return`stt_${q}_${A}`}export{j as Transcriber};

package/dist/stt/types.d.ts

@@ -1,44 +0,0 @@
-import type { AudioData, STTProvider, VoiceOptions } from '../types';
-export interface STTBrief {
-    audio: AudioData;
-    language?: string;
-    diarize?: boolean;
-    speakerCount?: number;
-    vocabularyHints?: string[];
-    /** Output subtitle format */
-    subtitleFormat?: 'srt' | 'vtt' | 'none';
-}
-export interface TranscriptionProject {
-    id: string;
-    transcript: TranscriptionResult;
-    subtitles?: string;
-    speakers?: SpeakerMap[];
-}
-export interface TranscriptionResult {
-    text: string;
-    segments: TranscriptionSegment[];
-    language: string;
-    durationMs: number;
-    wordTimings?: {
-        word: string;
-        startMs: number;
-        endMs: number;
-    }[];
-}
-export interface TranscriptionSegment {
-    text: string;
-    startMs: number;
-    endMs: number;
-    speakerId?: string;
-    speakerLabel?: string;
-    confidence?: number;
-}
-export interface SpeakerMap {
-    id: string;
-    label: string;
-    segmentCount: number;
-    totalSpeakingMs: number;
-}
-export interface STTOptions extends VoiceOptions {
-    stt: STTProvider;
-}

package/dist/stt/types.js

@@ -1 +0,0 @@
-// @bun

package/dist/sync/duration-negotiator.d.ts

@@ -1,37 +0,0 @@
-import type { VoiceTimingMap } from '../types';
-interface NegotiationResult {
-    /** Updated timing map with negotiated durations */
-    timingMap: VoiceTimingMap;
-    /** Per-scene negotiation details */
-    adjustments: SceneAdjustment[];
-}
-interface SceneAdjustment {
-    sceneId: string;
-    originalSceneDurationInFrames: number;
-    voiceDurationInFrames: number;
-    action: 'no_change' | 'extend_scene' | 'pad_silence' | 'suggest_rate_change';
-    suggestedRate?: number;
-    finalSceneDurationInFrames: number;
-}
-/**
- * Negotiate duration between voice audio and scene durations.
- *
- * One-pass duration balancing:
- * - Voice fits scene -> no change
- * - Voice > 110% of scene -> suggest rate increase (cap 1.3x), extend scene
- * - Voice < 70% of scene -> suggest rate decrease (floor 0.8x), pad silence
- */
-export declare class DurationNegotiator {
-    private static readonly UPPER_THRESHOLD;
-    private static readonly LOWER_THRESHOLD;
-    private static readonly MAX_RATE;
-    private static readonly MIN_RATE;
-    /**
-     * Negotiate voice-vs-scene durations.
-     *
-     * @param timingMap - Voice timing map with per-segment durations
-     * @param sceneDurations - Map of sceneId -> original scene duration in frames
-     */
-    negotiate(timingMap: VoiceTimingMap, sceneDurations: Map<string, number>): NegotiationResult;
-}
-export {};

package/dist/sync/duration-negotiator.js

@@ -1,2 +0,0 @@
-// @bun
-class q{static UPPER_THRESHOLD=1.1;static LOWER_THRESHOLD=0.7;static MAX_RATE=1.3;static MIN_RATE=0.8;negotiate(z,A){let v=[],B=z.segments.map((b)=>{let k=A.get(b.sceneId);if(k===void 0)return v.push({sceneId:b.sceneId,originalSceneDurationInFrames:b.recommendedSceneDurationInFrames,voiceDurationInFrames:b.durationInFrames,action:"no_change",finalSceneDurationInFrames:b.recommendedSceneDurationInFrames}),b;let x=b.durationInFrames/k;if(x>q.UPPER_THRESHOLD){let y=Math.min(x,q.MAX_RATE);return v.push({sceneId:b.sceneId,originalSceneDurationInFrames:k,voiceDurationInFrames:b.durationInFrames,action:x>q.MAX_RATE?"extend_scene":"suggest_rate_change",suggestedRate:y,finalSceneDurationInFrames:b.recommendedSceneDurationInFrames}),b}if(x<q.LOWER_THRESHOLD){let y=Math.max(x,q.MIN_RATE);return v.push({sceneId:b.sceneId,originalSceneDurationInFrames:k,voiceDurationInFrames:b.durationInFrames,action:"pad_silence",suggestedRate:y,finalSceneDurationInFrames:k}),{...b,recommendedSceneDurationInFrames:k}}return v.push({sceneId:b.sceneId,originalSceneDurationInFrames:k,voiceDurationInFrames:b.durationInFrames,action:"no_change",finalSceneDurationInFrames:b.recommendedSceneDurationInFrames}),b});return{timingMap:{...z,segments:B},adjustments:v}}}export{q as DurationNegotiator};

package/dist/sync/index.d.ts

@@ -1,3 +0,0 @@
-export { DurationNegotiator } from './duration-negotiator';
-export { SceneAdapter } from './scene-adapter';
-export { TimingCalculator } from './timing-calculator';

package/dist/sync/index.js

@@ -1,2 +0,0 @@
-// @bun
-class O{static DEFAULT_WPM=150;estimateSeconds(A,B){let G=B??O.DEFAULT_WPM,H=A.split(/\s+/).filter(Boolean).length;return Math.ceil(H/G*60)}estimateMs(A,B){let G=B??O.DEFAULT_WPM,H=A.split(/\s+/).filter(Boolean).length;return Math.ceil(H/G*60*1000)}estimateWordCount(A,B){let G=B??O.DEFAULT_WPM;return Math.round(A/60*G)}}class L{static UPPER_THRESHOLD=1.1;static LOWER_THRESHOLD=0.7;static MAX_RATE=1.3;static MIN_RATE=0.8;negotiate(A,B){let G=[],H=A.segments.map((z)=>{let q=B.get(z.sceneId);if(q===void 0)return G.push({sceneId:z.sceneId,originalSceneDurationInFrames:z.recommendedSceneDurationInFrames,voiceDurationInFrames:z.durationInFrames,action:"no_change",finalSceneDurationInFrames:z.recommendedSceneDurationInFrames}),z;let J=z.durationInFrames/q;if(J>L.UPPER_THRESHOLD){let K=Math.min(J,L.MAX_RATE);return G.push({sceneId:z.sceneId,originalSceneDurationInFrames:q,voiceDurationInFrames:z.durationInFrames,action:J>L.MAX_RATE?"extend_scene":"suggest_rate_change",suggestedRate:K,finalSceneDurationInFrames:z.recommendedSceneDurationInFrames}),z}if(J<L.LOWER_THRESHOLD){let K=Math.max(J,L.MIN_RATE);return G.push({sceneId:z.sceneId,originalSceneDurationInFrames:q,voiceDurationInFrames:z.durationInFrames,action:"pad_silence",suggestedRate:K,finalSceneDurationInFrames:q}),{...z,recommendedSceneDurationInFrames:q}}return G.push({sceneId:z.sceneId,originalSceneDurationInFrames:q,voiceDurationInFrames:z.durationInFrames,action:"no_change",finalSceneDurationInFrames:z.recommendedSceneDurationInFrames}),z});return{timingMap:{...A,segments:H},adjustments:G}}}class U{durationEstimator=new O;adapt(A){let B=A.scenes.filter((q)=>q.narrationText&&q.narrationText.trim().length>0),G=B.map((q,J)=>{let K=q.narrationText??"";return{sceneId:q.id,text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:this.inferContentType(J,B.length)}}),H=G.map((q)=>q.text).join(" "),z=G.reduce((q,J)=>q+J.estimatedDurationSeconds,0);return{fullText:H,segments:G,estimatedDurationSeconds:z}}inferContentType(A,B){if(A===0)return"intro";if(A===B-1)return"cta";if(A===1&&B>3)return"problem";if(A===B-2&&B>3)return"metric";return"solution"}}class X{calculate(A,B,G=1.15){let H=A.map((q)=>{let J=Math.ceil(q.durationMs/1000*B),K=Math.ceil(J*G),Y=q.wordTimings?.map((Q)=>({word:Q.word,startMs:Q.startMs,endMs:Q.endMs}));return{sceneId:q.sceneId,durationMs:q.durationMs,durationInFrames:J,recommendedSceneDurationInFrames:K,wordTimings:Y}});return{totalDurationMs:A.reduce((q,J)=>q+J.durationMs,0),segments:H,fps:B}}recalculateForFps(A,B){let G=A.segments.map((H)=>{let z=Math.ceil(H.durationMs/1000*B),q=H.recommendedSceneDurationInFrames/Math.max(H.durationInFrames,1);return{...H,durationInFrames:z,recommendedSceneDurationInFrames:Math.ceil(z*q)}});return{...A,segments:G,fps:B}}}export{X as TimingCalculator,U as SceneAdapter,L as DurationNegotiator};

package/dist/sync/scene-adapter.d.ts

@@ -1,29 +0,0 @@
-import type { TTSScript } from '../tts/types';
-interface ScenePlanScene {
-    id: string;
-    compositionId: string;
-    durationInFrames: number;
-    narrationText?: string;
-}
-interface ScenePlan {
-    scenes: ScenePlanScene[];
-    estimatedDurationSeconds: number;
-}
-/**
- * Adapt a video-gen ScenePlan into a TTSScript.
- *
- * Bridges the video-gen domain to the voice TTS domain,
- * preserving scene IDs for timing map correlation.
- */
-export declare class SceneAdapter {
-    private readonly durationEstimator;
-    /**
-     * Convert a ScenePlan into a TTSScript.
-     *
-     * Filters out scenes without narration text.
-     * Assigns contentType based on scene position (first = intro, last = cta, middle = solution).
-     */
-    adapt(scenePlan: ScenePlan): TTSScript;
-    private inferContentType;
-}
-export {};

package/dist/sync/scene-adapter.js

@@ -1,2 +0,0 @@
-// @bun
-class G{static DEFAULT_WPM=150;estimateSeconds(q,j){let z=j??G.DEFAULT_WPM,H=q.split(/\s+/).filter(Boolean).length;return Math.ceil(H/z*60)}estimateMs(q,j){let z=j??G.DEFAULT_WPM,H=q.split(/\s+/).filter(Boolean).length;return Math.ceil(H/z*60*1000)}estimateWordCount(q,j){let z=j??G.DEFAULT_WPM;return Math.round(q/60*z)}}class L{durationEstimator=new G;adapt(q){let j=q.scenes.filter((B)=>B.narrationText&&B.narrationText.trim().length>0),z=j.map((B,I)=>{let J=B.narrationText??"";return{sceneId:B.id,text:J,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(J),contentType:this.inferContentType(I,j.length)}}),H=z.map((B)=>B.text).join(" "),K=z.reduce((B,I)=>B+I.estimatedDurationSeconds,0);return{fullText:H,segments:z,estimatedDurationSeconds:K}}inferContentType(q,j){if(q===0)return"intro";if(q===j-1)return"cta";if(q===1&&j>3)return"problem";if(q===j-2&&j>3)return"metric";return"solution"}}export{L as SceneAdapter};

package/dist/sync/timing-calculator.d.ts

@@ -1,21 +0,0 @@
-import type { SynthesizedSegment } from '../tts/types';
-import type { VoiceTimingMap } from '../types';
-/**
- * Calculate timing maps from synthesized segments.
- *
- * Converts SynthesizedSegment[] into a VoiceTimingMap with frame calculations.
- */
-export declare class TimingCalculator {
-    /**
-     * Build a timing map from synthesized segments.
-     *
-     * @param segments - Synthesized audio segments with duration info
-     * @param fps - Frames per second for frame calculations
-     * @param breathingRoomFactor - Factor to add breathing room (default 1.15)
-     */
-    calculate(segments: SynthesizedSegment[], fps: number, breathingRoomFactor?: number): VoiceTimingMap;
-    /**
-     * Recalculate timing map for a different FPS.
-     */
-    recalculateForFps(timingMap: VoiceTimingMap, newFps: number): VoiceTimingMap;
-}

package/dist/sync/timing-calculator.js

@@ -1,2 +0,0 @@
-// @bun
-class z{calculate(h,j,l=1.15){let f=h.map((b)=>{let k=Math.ceil(b.durationMs/1000*j),x=Math.ceil(k*l),y=b.wordTimings?.map((v)=>({word:v.word,startMs:v.startMs,endMs:v.endMs}));return{sceneId:b.sceneId,durationMs:b.durationMs,durationInFrames:k,recommendedSceneDurationInFrames:x,wordTimings:y}});return{totalDurationMs:h.reduce((b,k)=>b+k.durationMs,0),segments:f,fps:j}}recalculateForFps(h,j){let l=h.segments.map((f)=>{let q=Math.ceil(f.durationMs/1000*j),b=f.recommendedSceneDurationInFrames/Math.max(f.durationInFrames,1);return{...f,durationInFrames:q,recommendedSceneDurationInFrames:Math.ceil(q*b)}});return{...h,segments:l,fps:j}}}export{z as TimingCalculator};

package/dist/tts/audio-assembler.d.ts

@@ -1,19 +0,0 @@
-import type { AudioData, VoicePacingDirective } from '../types';
-import type { SynthesizedSegment } from './types';
-/**
- * Assemble synthesized segments into a single audio track.
- *
- * Inserts silence between segments based on pacing directives.
- */
-export declare class AudioAssembler {
-    private readonly concatenator;
-    private readonly silenceGenerator;
-    /**
-     * Assemble segments with silence padding.
-     *
-     * @param segments - Synthesized audio segments
-     * @param directives - Pacing directives for silence timing
-     * @param defaultPauseMs - Default pause between segments (default 500ms)
-     */
-    assemble(segments: SynthesizedSegment[], directives: VoicePacingDirective[], defaultPauseMs?: number): AudioData;
-}

package/dist/tts/audio-assembler.js

@@ -1,2 +0,0 @@
-// @bun
-class V{concatenate(j){if(j.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[x]=j;if(!x)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(j.length===1)return{...x};let{format:N,sampleRateHz:E}=x,Q=x.channels??1;for(let k of j){if(k.format!==N)throw Error(`Format mismatch: expected ${N}, got ${k.format}`);if(k.sampleRateHz!==E)throw Error(`Sample rate mismatch: expected ${E}, got ${k.sampleRateHz}`)}let J=j.reduce((k,O)=>k+O.data.length,0),K=new Uint8Array(J),q=0;for(let k of j)K.set(k.data,q),q+=k.data.length;let U=j.reduce((k,O)=>k+(O.durationMs??0),0);return{data:K,format:N,sampleRateHz:E,durationMs:U,channels:Q}}}class W{generate(j,x="wav",N=44100,E=1){let Q=Math.ceil(N*j/1000),J=2,K=Q*2*E;return{data:new Uint8Array(K),format:x,sampleRateHz:N,durationMs:j,channels:E}}}class Y{concatenator=new V;silenceGenerator=new W;assemble(j,x,N=500){if(j.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[E]=j;if(!E)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let Q=new Map(x.map((q)=>[q.sceneId,q])),J=E.audio,K=[];for(let q=0;q<j.length;q++){let U=j[q];if(!U)continue;let k=Q.get(U.sceneId),O=k?.leadingSilenceMs??0;if(O>0)K.push(this.silenceGenerator.generate(O,J.format,J.sampleRateHz,J.channels??1));K.push(U.audio);let X=k?.trailingSilenceMs??(q<j.length-1?N:0);if(X>0)K.push(this.silenceGenerator.generate(X,J.format,J.sampleRateHz,J.channels??1))}return this.concatenator.concatenate(K)}}export{Y as AudioAssembler};

package/dist/tts/emphasis-planner.d.ts

@@ -1,31 +0,0 @@
-import type { ModelSelectionContext, ModelSelector } from '@contractspec/lib.ai-providers/selector-types';
-import type { LLMProvider, VoicePacingDirective } from '../types';
-import type { TTSScriptSegment } from './types';
-export interface EmphasisPlannerOptions {
-    llm?: LLMProvider;
-    model?: string;
-    modelSelector?: ModelSelector;
-    selectionContext?: ModelSelectionContext;
-}
-/**
- * Plan emphasis and tone per segment.
- *
- * With LLM: requests fine-grained tone/emphasis analysis.
- * Without LLM: falls back to PaceAnalyzer's content-type mapping.
- */
-export declare class EmphasisPlanner {
-    private readonly llm?;
-    private readonly model?;
-    private readonly modelSelector?;
-    private readonly selectionContext?;
-    private readonly paceAnalyzer;
-    constructor(options?: EmphasisPlannerOptions);
-    /**
-     * Plan emphasis and tone directives for segments.
-     *
-     * Falls back to deterministic mapping if LLM is unavailable.
-     */
-    plan(segments: TTSScriptSegment[], baseRate?: number): Promise<VoicePacingDirective[]>;
-    private resolveModel;
-    private planWithLlm;
-}

package/dist/tts/emphasis-planner.js

@@ -1,3 +0,0 @@
-// @bun
-var H={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class F{analyze(j,q=1){return j.map((B)=>{let w=H[B.contentType];return{sceneId:B.sceneId,rate:w.rate*q,emphasis:w.emphasis,tone:w.tone,leadingSilenceMs:w.leadingSilenceMs,trailingSilenceMs:w.trailingSilenceMs}})}getDefaults(j){return{...H[j]}}}class J{llm;model;modelSelector;selectionContext;paceAnalyzer;constructor(j){this.llm=j?.llm,this.model=j?.model,this.modelSelector=j?.modelSelector,this.selectionContext=j?.selectionContext,this.paceAnalyzer=new F}async plan(j,q=1){if(!this.llm)return this.paceAnalyzer.analyze(j,q);try{return await this.planWithLlm(j,q)}catch{return this.paceAnalyzer.analyze(j,q)}}async resolveModel(){if(this.model)return this.model;if(this.modelSelector){let j=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(j)).modelId}return}async planWithLlm(j,q){if(!this.llm)return this.paceAnalyzer.analyze(j,q);let B=await this.resolveModel(),D=(await this.llm.chat([{role:"system",content:[{type:"text",text:["You are a voice director planning emphasis and pacing for TTS narration.","For each segment, return a JSON array of directives.","Each directive has: sceneId, rate (0.7-1.3), emphasis (reduced|normal|strong),","tone (neutral|urgent|excited|calm|authoritative), leadingSilenceMs, trailingSilenceMs.","Return ONLY a JSON array, no other text."].join(`
-`)}]},{role:"user",content:[{type:"text",text:JSON.stringify(j.map((v)=>({sceneId:v.sceneId,text:v.text,contentType:v.contentType})))}]}],{model:B,temperature:0.3,responseFormat:"json"})).message.content.find((v)=>v.type==="text");if(!D||D.type!=="text")return this.paceAnalyzer.analyze(j,q);return JSON.parse(D.text).map((v)=>({...v,rate:v.rate*q}))}}export{J as EmphasisPlanner};

package/dist/tts/index.d.ts

@@ -1,6 +0,0 @@
-export { AudioAssembler } from './audio-assembler';
-export { EmphasisPlanner } from './emphasis-planner';
-export { PaceAnalyzer } from './pace-analyzer';
-export { SegmentSynthesizer } from './segment-synthesizer';
-export type { PacingConfig, SynthesizedSegment, TTSBrief, TTSOptions, TTSProject, TTSScript, TTSScriptSegment, TTSVoiceConfig, VideoTTSBrief, } from './types';
-export { VoiceSynthesizer } from './voice-synthesizer';

package/dist/tts/index.js

@@ -1,3 +0,0 @@
-// @bun
-class H{concatenate(V){if(V.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[q]=V;if(!q)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(V.length===1)return{...q};let{format:J,sampleRateHz:K}=q,U=q.channels??1;for(let Z of V){if(Z.format!==J)throw Error(`Format mismatch: expected ${J}, got ${Z.format}`);if(Z.sampleRateHz!==K)throw Error(`Sample rate mismatch: expected ${K}, got ${Z.sampleRateHz}`)}let B=V.reduce((Z,$)=>Z+$.data.length,0),Q=new Uint8Array(B),X=0;for(let Z of V)Q.set(Z.data,X),X+=Z.data.length;let j=V.reduce((Z,$)=>Z+($.durationMs??0),0);return{data:Q,format:J,sampleRateHz:K,durationMs:j,channels:U}}}class O{static DEFAULT_WPM=150;estimateSeconds(V,q){let J=q??O.DEFAULT_WPM,K=V.split(/\s+/).filter(Boolean).length;return Math.ceil(K/J*60)}estimateMs(V,q){let J=q??O.DEFAULT_WPM,K=V.split(/\s+/).filter(Boolean).length;return Math.ceil(K/J*60*1000)}estimateWordCount(V,q){let J=q??O.DEFAULT_WPM;return Math.round(V/60*J)}}class _{generate(V,q="wav",J=44100,K=1){let U=Math.ceil(J*V/1000),B=2,Q=U*2*K;return{data:new Uint8Array(Q),format:q,sampleRateHz:J,durationMs:V,channels:K}}}class T{concatenator=new H;silenceGenerator=new _;assemble(V,q,J=500){if(V.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[K]=V;if(!K)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let U=new Map(q.map((X)=>[X.sceneId,X])),B=K.audio,Q=[];for(let X=0;X<V.length;X++){let j=V[X];if(!j)continue;let Z=U.get(j.sceneId),$=Z?.leadingSilenceMs??0;if($>0)Q.push(this.silenceGenerator.generate($,B.format,B.sampleRateHz,B.channels??1));Q.push(j.audio);let k=Z?.trailingSilenceMs??(X<V.length-1?J:0);if(k>0)Q.push(this.silenceGenerator.generate(k,B.format,B.sampleRateHz,B.channels??1))}return this.concatenator.concatenate(Q)}}var G={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class W{analyze(V,q=1){return V.map((J)=>{let K=G[J.contentType];return{sceneId:J.sceneId,rate:K.rate*q,emphasis:K.emphasis,tone:K.tone,leadingSilenceMs:K.leadingSilenceMs,trailingSilenceMs:K.trailingSilenceMs}})}getDefaults(V){return{...G[V]}}}class Y{llm;model;modelSelector;selectionContext;paceAnalyzer;constructor(V){this.llm=V?.llm,this.model=V?.model,this.modelSelector=V?.modelSelector,this.selectionContext=V?.selectionContext,this.paceAnalyzer=new W}async plan(V,q=1){if(!this.llm)return this.paceAnalyzer.analyze(V,q);try{return await this.planWithLlm(V,q)}catch{return this.paceAnalyzer.analyze(V,q)}}async resolveModel(){if(this.model)return this.model;if(this.modelSelector){let V=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(V)).modelId}return}async planWithLlm(V,q){if(!this.llm)return this.paceAnalyzer.analyze(V,q);let J=await this.resolveModel(),U=(await this.llm.chat([{role:"system",content:[{type:"text",text:["You are a voice director planning emphasis and pacing for TTS narration.","For each segment, return a JSON array of directives.","Each directive has: sceneId, rate (0.7-1.3), emphasis (reduced|normal|strong),","tone (neutral|urgent|excited|calm|authoritative), leadingSilenceMs, trailingSilenceMs.","Return ONLY a JSON array, no other text."].join(`
-`)}]},{role:"user",content:[{type:"text",text:JSON.stringify(V.map((Q)=>({sceneId:Q.sceneId,text:Q.text,contentType:Q.contentType})))}]}],{model:J,temperature:0.3,responseFormat:"json"})).message.content.find((Q)=>Q.type==="text");if(!U||U.type!=="text")return this.paceAnalyzer.analyze(V,q);return JSON.parse(U.text).map((Q)=>({...Q,rate:Q.rate*q}))}}class L{tts;constructor(V){this.tts=V}async synthesizeAll(V,q,J){let K=new Map(J.map((B)=>[B.sceneId,B]));return await Promise.all(V.map((B)=>this.synthesizeOne(B,q,K.get(B.sceneId))))}async synthesizeOne(V,q,J){let K=await this.tts.synthesize({text:V.text,voiceId:q.voiceId,language:q.language,style:q.style,stability:q.stability,rate:J?.rate,emphasis:J?.emphasis});return{sceneId:V.sceneId,audio:K.audio,durationMs:K.audio.durationMs??0,wordTimings:K.wordTimings?.map((U)=>({word:U.word,startMs:U.startMs,endMs:U.endMs}))}}}class N{segmentSynthesizer;emphasisPlanner;audioAssembler=new T;durationEstimator=new O;paceAnalyzer=new W;options;constructor(V){this.options=V,this.segmentSynthesizer=new L(V.tts),this.emphasisPlanner=new Y({llm:V.llm,model:V.model})}async synthesize(V){let q=this.buildScript(V);return this.executePipeline(q,V.voice,V.pacing)}async synthesizeForVideo(V){let q=this.buildScriptFromScenePlan(V);return this.executePipeline(q,V.voice,V.pacing,V.fps)}async executePipeline(V,q,J,K){let U=F(),B=J?.baseRate??1,Q=await this.emphasisPlanner.plan(V.segments,B),X=await this.segmentSynthesizer.synthesizeAll(V.segments,q,Q),j=J?.segmentPauseMs??500,Z=this.audioAssembler.assemble(X,Q,j),$=K??this.options.fps??30,k=J?.breathingRoomFactor??1.15,C=this.buildTimingMap(X,$,k);return{id:U,script:V,pacingDirectives:Q,segments:X,assembledAudio:Z,timingMap:C}}buildScript(V){let q=[],J=`${V.content.title}. ${V.content.summary}`;if(q.push({sceneId:"intro",text:J,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(J),contentType:"intro"}),V.content.problems.length>0){let B=V.content.problems.join(". ");q.push({sceneId:"problems",text:B,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(B),contentType:"problem"})}if(V.content.solutions.length>0){let B=V.content.solutions.join(". ");q.push({sceneId:"solutions",text:B,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(B),contentType:"solution"})}if(V.content.metrics&&V.content.metrics.length>0){let B=V.content.metrics.join(". ");q.push({sceneId:"metrics",text:B,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(B),contentType:"metric"})}if(V.content.callToAction)q.push({sceneId:"cta",text:V.content.callToAction,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(V.content.callToAction),contentType:"cta"});let K=q.map((B)=>B.text).join(" "),U=q.reduce((B,Q)=>B+Q.estimatedDurationSeconds,0);return{fullText:K,segments:q,estimatedDurationSeconds:U}}buildScriptFromScenePlan(V){let q=V.scenePlan.scenes.filter((U)=>U.narrationText).map((U)=>{let B=U.narrationText??"";return{sceneId:U.id,text:B,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(B),contentType:"intro"}}),J=q.map((U)=>U.text).join(" "),K=q.reduce((U,B)=>U+B.estimatedDurationSeconds,0);return{fullText:J,segments:q,estimatedDurationSeconds:K}}buildTimingMap(V,q,J){let K=V.map((B)=>{let Q=Math.ceil(B.durationMs/1000*q);return{sceneId:B.sceneId,durationMs:B.durationMs,durationInFrames:Q,recommendedSceneDurationInFrames:Math.ceil(Q*J),wordTimings:B.wordTimings?.map((X)=>({word:X.word,startMs:X.startMs,endMs:X.endMs}))}});return{totalDurationMs:V.reduce((B,Q)=>B+Q.durationMs,0),segments:K,fps:q}}}function F(){let V=Date.now().toString(36),q=Math.random().toString(36).slice(2,8);return`tts_${V}_${q}`}export{N as VoiceSynthesizer,L as SegmentSynthesizer,W as PaceAnalyzer,Y as EmphasisPlanner,T as AudioAssembler};

package/dist/tts/pace-analyzer.d.ts

@@ -1,30 +0,0 @@
-import type { VoicePacingDirective } from '../types';
-import type { TTSScriptSegment } from './types';
-type ContentType = TTSScriptSegment['contentType'];
-type Tone = VoicePacingDirective['tone'];
-type Emphasis = VoicePacingDirective['emphasis'];
-interface PacingDefaults {
-    rate: number;
-    emphasis: Emphasis;
-    tone: Tone;
-    leadingSilenceMs: number;
-    trailingSilenceMs: number;
-}
-/**
- * Analyze content segments and produce pacing directives.
- *
- * Uses a deterministic content-type -> pacing mapping as fallback.
- * When an LLM is provided, can enhance with fine-grained analysis.
- */
-export declare class PaceAnalyzer {
-    /**
-     * Generate pacing directives for a list of script segments.
-     *
-     * @param segments - The TTS script segments to analyze
-     * @param baseRate - Base rate multiplier applied to all directives
-     */
-    analyze(segments: TTSScriptSegment[], baseRate?: number): VoicePacingDirective[];
-    /** Get the default pacing for a content type */
-    getDefaults(contentType: ContentType): PacingDefaults;
-}
-export {};

package/dist/tts/pace-analyzer.js

@@ -1,2 +0,0 @@
-// @bun
-var v={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class x{analyze(k,w=1){return k.map((q)=>{let j=v[q.contentType];return{sceneId:q.sceneId,rate:j.rate*w,emphasis:j.emphasis,tone:j.tone,leadingSilenceMs:j.leadingSilenceMs,trailingSilenceMs:j.trailingSilenceMs}})}getDefaults(k){return{...v[k]}}}export{x as PaceAnalyzer};

package/dist/tts/segment-synthesizer.d.ts

@@ -1,21 +0,0 @@
-import type { TTSProvider, VoicePacingDirective } from '../types';
-import type { SynthesizedSegment, TTSScriptSegment, TTSVoiceConfig } from './types';
-/**
- * Synthesize individual script segments via TTSProvider.
- *
- * Applies pacing directives (rate, emphasis) to each synthesis call.
- * Runs segments in parallel via Promise.all().
- */
-export declare class SegmentSynthesizer {
-    private readonly tts;
-    constructor(tts: TTSProvider);
-    /**
-     * Synthesize all segments in parallel.
-     *
-     * @param segments - Script segments to synthesize
-     * @param voice - Voice configuration
-     * @param directives - Pacing directives per segment (matched by sceneId)
-     */
-    synthesizeAll(segments: TTSScriptSegment[], voice: TTSVoiceConfig, directives: VoicePacingDirective[]): Promise<SynthesizedSegment[]>;
-    private synthesizeOne;
-}

package/dist/tts/segment-synthesizer.js

@@ -1,2 +0,0 @@
-// @bun
-class C{tts;constructor(j){this.tts=j}async synthesizeAll(j,k,A){let q=new Map(A.map((x)=>[x.sceneId,x]));return await Promise.all(j.map((x)=>this.synthesizeOne(x,k,q.get(x.sceneId))))}async synthesizeOne(j,k,A){let q=await this.tts.synthesize({text:j.text,voiceId:k.voiceId,language:k.language,style:k.style,stability:k.stability,rate:A?.rate,emphasis:A?.emphasis});return{sceneId:j.sceneId,audio:q.audio,durationMs:q.audio.durationMs??0,wordTimings:q.wordTimings?.map((B)=>({word:B.word,startMs:B.startMs,endMs:B.endMs}))}}}export{C as SegmentSynthesizer};

package/dist/tts/types.d.ts

@@ -1,76 +0,0 @@
-import type { ContentBrief } from '@contractspec/lib.content-gen/types';
-import type { AudioData, TTSProvider, VoiceOptions, VoicePacingDirective, VoiceTimingMap } from '../types';
-export interface TTSBrief {
-    content: ContentBrief;
-    voice: TTSVoiceConfig;
-    pacing?: PacingConfig;
-    targetDurationSeconds?: number;
-    locale?: string;
-}
-export interface TTSVoiceConfig {
-    voiceId: string;
-    language?: string;
-    style?: number;
-    stability?: number;
-}
-export interface PacingConfig {
-    /** Base speaking rate multiplier. Default 1.0 */
-    baseRate?: number;
-    strategy: 'uniform' | 'dynamic' | 'scene-matched';
-    /** Pause between segments in ms. Default 500 */
-    segmentPauseMs?: number;
-    /** Breathing room factor for scene duration. Default 1.15 */
-    breathingRoomFactor?: number;
-}
-export interface TTSProject {
-    id: string;
-    script: TTSScript;
-    pacingDirectives: VoicePacingDirective[];
-    segments?: SynthesizedSegment[];
-    assembledAudio?: AudioData;
-    timingMap?: VoiceTimingMap;
-}
-export interface TTSScript {
-    fullText: string;
-    segments: TTSScriptSegment[];
-    estimatedDurationSeconds: number;
-}
-export interface TTSScriptSegment {
-    sceneId: string;
-    text: string;
-    estimatedDurationSeconds: number;
-    contentType: 'intro' | 'problem' | 'solution' | 'metric' | 'cta' | 'transition';
-}
-export interface SynthesizedSegment {
-    sceneId: string;
-    audio: AudioData;
-    durationMs: number;
-    wordTimings?: {
-        word: string;
-        startMs: number;
-        endMs: number;
-    }[];
-}
-export interface TTSOptions extends VoiceOptions {
-    tts: TTSProvider;
-    defaultVoiceId?: string;
-    fps?: number;
-    defaultPacing?: PacingConfig;
-}
-/** Video-aware brief: takes a ScenePlan instead of standalone content */
-export interface VideoTTSBrief {
-    content: ContentBrief;
-    scenePlan: {
-        scenes: {
-            id: string;
-            compositionId: string;
-            durationInFrames: number;
-            narrationText?: string;
-        }[];
-        estimatedDurationSeconds: number;
-    };
-    voice: TTSVoiceConfig;
-    pacing?: PacingConfig;
-    fps: number;
-    locale?: string;
-}

package/dist/tts/types.js

@@ -1 +0,0 @@
-// @bun

package/dist/tts/voice-synthesizer.d.ts

@@ -1,28 +0,0 @@
-import type { TTSBrief, TTSOptions, TTSProject, VideoTTSBrief } from './types';
-/**
- * Main TTS orchestrator.
- *
- * Pipeline:
- * 1. Build script from content brief (or scene plan)
- * 2. Analyze pacing per segment
- * 3. Plan emphasis/tone
- * 4. Synthesize each segment via TTSProvider
- * 5. Assemble into final audio + timing map
- */
-export declare class VoiceSynthesizer {
-    private readonly segmentSynthesizer;
-    private readonly emphasisPlanner;
-    private readonly audioAssembler;
-    private readonly durationEstimator;
-    private readonly paceAnalyzer;
-    private readonly options;
-    constructor(options: TTSOptions);
-    /** Standalone TTS from content brief */
-    synthesize(brief: TTSBrief): Promise<TTSProject>;
-    /** Scene-aware TTS for video-gen */
-    synthesizeForVideo(brief: VideoTTSBrief): Promise<TTSProject>;
-    private executePipeline;
-    private buildScript;
-    private buildScriptFromScenePlan;
-    private buildTimingMap;
-}

package/dist/tts/voice-synthesizer.js

@@ -1,3 +0,0 @@
-// @bun
-class Y{concatenate(q){if(q.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[J]=q;if(!J)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(q.length===1)return{...J};let{format:Q,sampleRateHz:U}=J,Z=J.channels??1;for(let V of q){if(V.format!==Q)throw Error(`Format mismatch: expected ${Q}, got ${V.format}`);if(V.sampleRateHz!==U)throw Error(`Sample rate mismatch: expected ${U}, got ${V.sampleRateHz}`)}let K=q.reduce((V,W)=>V+W.data.length,0),X=new Uint8Array(K),$=0;for(let V of q)X.set(V.data,$),$+=V.data.length;let j=q.reduce((V,W)=>V+(W.durationMs??0),0);return{data:X,format:Q,sampleRateHz:U,durationMs:j,channels:Z}}}class k{static DEFAULT_WPM=150;estimateSeconds(q,J){let Q=J??k.DEFAULT_WPM,U=q.split(/\s+/).filter(Boolean).length;return Math.ceil(U/Q*60)}estimateMs(q,J){let Q=J??k.DEFAULT_WPM,U=q.split(/\s+/).filter(Boolean).length;return Math.ceil(U/Q*60*1000)}estimateWordCount(q,J){let Q=J??k.DEFAULT_WPM;return Math.round(q/60*Q)}}class L{generate(q,J="wav",Q=44100,U=1){let Z=Math.ceil(Q*q/1000),K=2,X=Z*2*U;return{data:new Uint8Array(X),format:J,sampleRateHz:Q,durationMs:q,channels:U}}}class H{concatenator=new Y;silenceGenerator=new L;assemble(q,J,Q=500){if(q.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[U]=q;if(!U)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let Z=new Map(J.map(($)=>[$.sceneId,$])),K=U.audio,X=[];for(let $=0;$<q.length;$++){let j=q[$];if(!j)continue;let V=Z.get(j.sceneId),W=V?.leadingSilenceMs??0;if(W>0)X.push(this.silenceGenerator.generate(W,K.format,K.sampleRateHz,K.channels??1));X.push(j.audio);let O=V?.trailingSilenceMs??($<q.length-1?Q:0);if(O>0)X.push(this.silenceGenerator.generate(O,K.format,K.sampleRateHz,K.channels??1))}return this.concatenator.concatenate(X)}}var N={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class B{analyze(q,J=1){return q.map((Q)=>{let U=N[Q.contentType];return{sceneId:Q.sceneId,rate:U.rate*J,emphasis:U.emphasis,tone:U.tone,leadingSilenceMs:U.leadingSilenceMs,trailingSilenceMs:U.trailingSilenceMs}})}getDefaults(q){return{...N[q]}}}class _{llm;model;modelSelector;selectionContext;paceAnalyzer;constructor(q){this.llm=q?.llm,this.model=q?.model,this.modelSelector=q?.modelSelector,this.selectionContext=q?.selectionContext,this.paceAnalyzer=new B}async plan(q,J=1){if(!this.llm)return this.paceAnalyzer.analyze(q,J);try{return await this.planWithLlm(q,J)}catch{return this.paceAnalyzer.analyze(q,J)}}async resolveModel(){if(this.model)return this.model;if(this.modelSelector){let q=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(q)).modelId}return}async planWithLlm(q,J){if(!this.llm)return this.paceAnalyzer.analyze(q,J);let Q=await this.resolveModel(),Z=(await this.llm.chat([{role:"system",content:[{type:"text",text:["You are a voice director planning emphasis and pacing for TTS narration.","For each segment, return a JSON array of directives.","Each directive has: sceneId, rate (0.7-1.3), emphasis (reduced|normal|strong),","tone (neutral|urgent|excited|calm|authoritative), leadingSilenceMs, trailingSilenceMs.","Return ONLY a JSON array, no other text."].join(`
-`)}]},{role:"user",content:[{type:"text",text:JSON.stringify(q.map((X)=>({sceneId:X.sceneId,text:X.text,contentType:X.contentType})))}]}],{model:Q,temperature:0.3,responseFormat:"json"})).message.content.find((X)=>X.type==="text");if(!Z||Z.type!=="text")return this.paceAnalyzer.analyze(q,J);return JSON.parse(Z.text).map((X)=>({...X,rate:X.rate*J}))}}class G{tts;constructor(q){this.tts=q}async synthesizeAll(q,J,Q){let U=new Map(Q.map((K)=>[K.sceneId,K]));return await Promise.all(q.map((K)=>this.synthesizeOne(K,J,U.get(K.sceneId))))}async synthesizeOne(q,J,Q){let U=await this.tts.synthesize({text:q.text,voiceId:J.voiceId,language:J.language,style:J.style,stability:J.stability,rate:Q?.rate,emphasis:Q?.emphasis});return{sceneId:q.sceneId,audio:U.audio,durationMs:U.audio.durationMs??0,wordTimings:U.wordTimings?.map((Z)=>({word:Z.word,startMs:Z.startMs,endMs:Z.endMs}))}}}class I{segmentSynthesizer;emphasisPlanner;audioAssembler=new H;durationEstimator=new k;paceAnalyzer=new B;options;constructor(q){this.options=q,this.segmentSynthesizer=new G(q.tts),this.emphasisPlanner=new _({llm:q.llm,model:q.model})}async synthesize(q){let J=this.buildScript(q);return this.executePipeline(J,q.voice,q.pacing)}async synthesizeForVideo(q){let J=this.buildScriptFromScenePlan(q);return this.executePipeline(J,q.voice,q.pacing,q.fps)}async executePipeline(q,J,Q,U){let Z=C(),K=Q?.baseRate??1,X=await this.emphasisPlanner.plan(q.segments,K),$=await this.segmentSynthesizer.synthesizeAll(q.segments,J,X),j=Q?.segmentPauseMs??500,V=this.audioAssembler.assemble($,X,j),W=U??this.options.fps??30,O=Q?.breathingRoomFactor??1.15,F=this.buildTimingMap($,W,O);return{id:Z,script:q,pacingDirectives:X,segments:$,assembledAudio:V,timingMap:F}}buildScript(q){let J=[],Q=`${q.content.title}. ${q.content.summary}`;if(J.push({sceneId:"intro",text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"intro"}),q.content.problems.length>0){let K=q.content.problems.join(". ");J.push({sceneId:"problems",text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:"problem"})}if(q.content.solutions.length>0){let K=q.content.solutions.join(". ");J.push({sceneId:"solutions",text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:"solution"})}if(q.content.metrics&&q.content.metrics.length>0){let K=q.content.metrics.join(". ");J.push({sceneId:"metrics",text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:"metric"})}if(q.content.callToAction)J.push({sceneId:"cta",text:q.content.callToAction,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(q.content.callToAction),contentType:"cta"});let U=J.map((K)=>K.text).join(" "),Z=J.reduce((K,X)=>K+X.estimatedDurationSeconds,0);return{fullText:U,segments:J,estimatedDurationSeconds:Z}}buildScriptFromScenePlan(q){let J=q.scenePlan.scenes.filter((Z)=>Z.narrationText).map((Z)=>{let K=Z.narrationText??"";return{sceneId:Z.id,text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:"intro"}}),Q=J.map((Z)=>Z.text).join(" "),U=J.reduce((Z,K)=>Z+K.estimatedDurationSeconds,0);return{fullText:Q,segments:J,estimatedDurationSeconds:U}}buildTimingMap(q,J,Q){let U=q.map((K)=>{let X=Math.ceil(K.durationMs/1000*J);return{sceneId:K.sceneId,durationMs:K.durationMs,durationInFrames:X,recommendedSceneDurationInFrames:Math.ceil(X*Q),wordTimings:K.wordTimings?.map(($)=>({word:$.word,startMs:$.startMs,endMs:$.endMs}))}});return{totalDurationMs:q.reduce((K,X)=>K+X.durationMs,0),segments:U,fps:J}}}function C(){let q=Date.now().toString(36),J=Math.random().toString(36).slice(2,8);return`tts_${q}_${J}`}export{I as VoiceSynthesizer};

package/dist/types.d.ts

@@ -1,18 +0,0 @@
-import type { ContentBrief } from '@contractspec/lib.content-gen/types';
-import type { LLMProvider } from '@contractspec/lib.contracts-integrations/integrations/providers/llm';
-import type { AudioData, AudioFormat, ConversationalEvent, ConversationalProvider, ConversationalSessionSummary, STTProvider, TTSProvider, Voice, WordTiming } from '@contractspec/lib.contracts-integrations/integrations/providers/voice';
-import type { VoicePacingDirective, VoiceTimingMap } from '@contractspec/lib.contracts-integrations/integrations/providers/voice-video-sync';
-export type { AudioData, AudioFormat, ContentBrief, ConversationalEvent, ConversationalProvider, ConversationalSessionSummary, LLMProvider, STTProvider, TTSProvider, Voice, VoicePacingDirective, VoiceTimingMap, WordTiming, };
-/** Shared options base for all voice generators */
-export interface VoiceOptions {
-    llm?: LLMProvider;
-    model?: string;
-    temperature?: number;
-    locale?: string;
-    /** Transport mode for voice providers. */
-    transport?: 'rest' | 'mcp' | 'sdk';
-    /** Auth method for voice providers. */
-    authMethod?: 'api-key' | 'oauth2' | 'bearer';
-    /** Custom auth headers for voice providers. */
-    authHeaders?: Record<string, string>;
-}

package/dist/types.js

@@ -1 +0,0 @@
-// @bun

package/dist/voice.feature.d.ts

@@ -1 +0,0 @@
-export declare const VoiceFeature: import("@contractspec/lib.contracts-spec").FeatureModuleSpec;

package/dist/voice.feature.js

@@ -1,2 +0,0 @@
-// @bun
-import{defineFeature as g}from"@contractspec/lib.contracts-spec/features";var j=g({meta:{key:"libs.voice",version:"1.0.0",title:"Voice",description:"Voice capabilities: TTS, STT, and conversational AI",domain:"voice",owners:["@contractspec-core"],tags:["package","libs","voice"],stability:"experimental"}});export{j as VoiceFeature};