@streamoji/aitwin 0.4.8 → 0.5.0

package/dist/index.d.ts

@@ -167,8 +167,24 @@ export declare type AvatarTtsSpeakOptions = {
     ttsEngineId?: TtsEngineId;
 };
 
+/** Pair transitions for playback, including sil→first and last→sil bookends. */
+export declare function buildPlaybackTransitions(queue: VisemeQueueItem[]): RawVisemeTransition[];
+
+/** Collapsed + compacted viseme series used for lipsync playback. */
+export declare function buildPlaybackVisemeSeries(queue: VisemeQueueItem[]): VisemeQueueItem[];
+
+export declare function buildRawVisemeTransitions(queue: VisemeQueueItem[]): RawVisemeTransition[];
+
 export declare function buildVoiceWebSocketUrl(options: RealtimeVoiceOptions): string;
 
+/**
+ * Drop brief middle visemes in {aa,E,I} and {DD,SS,kk,CH} when either adjacent
+ * gap is under VISEME_SKIP_GAP_MS. First and last entries in the series are kept.
+ */
+export declare function compactSkippableVisemes(series: VisemeQueueItem[]): VisemeQueueItem[];
+
+export declare const CONSONANT_SKIP_GROUP: Set<string>;
+
 export declare function createAvatarTtsLipsyncController(onStatus: (status: AvatarTtsLipsyncStatus) => void, apiBase?: string, options?: AvatarTtsLipsyncControllerOptions): AvatarTtsLipsyncController;
 
 export declare function createRealtimeVoiceLipsyncController(onStatus: (status: AvatarTtsLipsyncStatus) => void, initialOptions?: RealtimeVoiceOptions): RealtimeVoiceLipsyncController;
@@ -250,6 +266,9 @@ export declare function getAiTwinSourceUrl(faceId: string, cdnBase?: string, ext
 /** Public CDN thumbnail for custom-faces/{faceId}/ (same id as pipeline avatarId). */
 export declare function getAiTwinThumbnailUrl(faceId: string, cdnBase?: string): string;
 
+/** Visemes removed by compaction (for debug). */
+export declare function listSkippedVisemes(queue: VisemeQueueItem[]): string[];
+
 export declare const MIC_SAMPLE_RATE = 16000;
 
 /** Google TTS streams Oculus viseme symbols; normalize casing and unknown ids. */
@@ -262,6 +281,13 @@ export declare const OCULUS_VISEME_IDS: readonly ["aa", "CH", "DD", "E", "FF", "
 
 export declare const PLAYBACK_SAMPLE_RATE = 16000;
 
+declare type RawVisemeTransition = {
+    from: string;
+    to: string;
+    fromVtime: number;
+    toVtime: number;
+};
+
 export declare type RealtimeLipsyncDebugSnapshot = {
     event: "metadata" | "audio_chunk" | "lipsync" | "bot_started" | "bot_stopped" | "user_started" | "interrupt" | "reset";
     spokenText: string;
@@ -312,6 +338,9 @@ export declare function resolveDirectFaceId(avatarId?: string, faceId?: string):
 /** Pick the active viseme at playback offset (latest vtime wins on overlap). */
 export declare function resolveVisemeAtTime(queue: VisemeQueueItem[], elapsedMs: number): string;
 
+/** Latest non-sil viseme in the playback series at or before elapsedMs. */
+export declare function resolveVisemeFromPlaybackSeries(series: VisemeQueueItem[], elapsedMs: number): string;
+
 /** Word active at playback offset (ms). */
 export declare function resolveWordAtTime(words: WordQueueItem[], timeMs: number): WordQueueItem | null;
 
@@ -350,6 +379,9 @@ export declare type TwinAssetUrls = {
 
 export declare const VISEME_IDS: readonly ["aa", "E", "I", "O", "U", "PP", "FF", "DD", "SS", "TH", "CH", "RR", "kk", "nn"];
 
+/** Minimum gap (ms) between consecutive vtimes to keep a skippable viseme. */
+export declare const VISEME_SKIP_GAP_MS = 35;
+
 /** Preset Cartesia voice (Katie) for testing without a custom clone. */
 export declare const VISEME_TEST_CARTESIA_VOICE_ID = "f786b574-daa5-4673-aa0c-cbe3e8534c02";
 
@@ -381,6 +413,8 @@ export declare type VoiceConnectOptions = {
     wsUrl?: string;
 };
 
+export declare const VOWEL_SKIP_GROUP: Set<string>;
+
 export declare interface WordQueueItem {
     word: string;
     wtime: number;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@streamoji/aitwin",
-  "version": "0.4.8",
+  "version": "0.5.0",
   "description": "Embeddable React AI twin face with TTS lipsync",
   "type": "module",
   "main": "./dist/aitwin.cjs",