@firetable/project-xiaochun 0.1.16 → 0.1.18

package/README-CN.md

@@ -152,12 +152,15 @@ await xc.speakAudio(arrayBuffer, { text: '你好', motion: true, lipsync: true }
 await xc.speakAudio(blob);                                   // Blob(mp3 / wav / ogg 等)
 await xc.speakAudio('https://cdn.example.com/voice.mp3');    // URL:默认由宿主页 fetch({ fetch: 'frame' } = 交给 iframe 去 fetch)
 await xc.speakAudio(pcm, { format: 'pcm16', sampleRate: 24000 });   // 无头原始 PCM 必须给 sampleRate(8000–96000)
+await xc.speakAudio(pcm, { format: 'pcm16', sampleRate: 16000, audible: false }); // 只做动作和口型, iframe 增益为 0
 
 const s = xc.speakAudioStream({ sampleRate: 24000 });        // 流式:比如 TTS 服务边合成边返回的 PCM 块
 s.write(int16Chunk); s.write(next); s.end(); await s.done;   // s.abort() 立即停止
 // 随时可取消:传 { signal: abortController.signal }
 ```
 
+* `audible` 默认 `true`。`false` 仍会解码、生成动作并驱动口型,但 iframe 增益为 0,声音可以由宿主页面自己播放。
+* 宿主音频的 `utterance start` 在思考动作和 EMAGE 首窗之后、声音时钟起步时才发。页面自己的播放器要等这个事件再出声,口型才对齐。
 * 失败时 reject:`bad_request`(无法解码 / 空音频 / URL 或 sampleRate 不合法)、`unsupported`、`failed`。
 * 浏览器仍要求宿主页先有用户手势才能出声,iframe 也要有 `allow="autoplay"`(SDK 已自动设置)。
 * 新的 `say()` / `speakAudio()` 会打断正在进行的那一次,被打断的 Promise 会 resolve。

package/README.md

@@ -152,12 +152,15 @@ await xc.speakAudio(arrayBuffer, { text: 'Hi', motion: true, lipsync: true }); /
 await xc.speakAudio(blob);                                   // Blob (mp3 / wav / ogg …)
 await xc.speakAudio('https://cdn.example.com/voice.mp3');    // URL: fetched by the host page by default ({ fetch: 'frame' } = fetched inside the iframe)
 await xc.speakAudio(pcm, { format: 'pcm16', sampleRate: 24000 });   // headerless PCM needs sampleRate (8000–96000)
+await xc.speakAudio(pcm, { format: 'pcm16', sampleRate: 16000, audible: false }); // motion + lip-sync only; iframe gain is 0
 
 const s = xc.speakAudioStream({ sampleRate: 24000 });        // streaming: e.g. PCM chunks from a TTS server
 s.write(int16Chunk); s.write(next); s.end(); await s.done;   // s.abort() stops immediately
 // Cancel any time: pass { signal: abortController.signal }
 ```
 
+* `audible` defaults to `true`. `false` still decodes the audio, generates motion, and drives lip-sync, but the iframe gain is 0 so the host page can be the only speaker.
+* For host audio, `utterance` `start` fires when the audio clock starts, after the thinking pose and the first EMAGE window. Hold the page's own player until that event if it must stay in sync.
 * Rejects with `bad_request` (undecodable / empty audio, bad URL or sampleRate), `unsupported`, or `failed`.
 * Browsers still require a user gesture on the host page before audio can play, and the iframe needs `allow="autoplay"` (the SDK sets it).
 * A newer `say()` / `speakAudio()` call preempts the one in progress, and the preempted promise resolves.

package/dist/chunks/{chunk-UWL65W7D.js → chunk-BN4H52OR.js}

@@ -1,4 +1,4 @@
-/*! @firetable/project-xiaochun v0.1.16 | MIT | https://github.com/FireTable/project-xiaochun */
+/*! @firetable/project-xiaochun v0.1.18 | MIT | https://github.com/FireTable/project-xiaochun */
 import {
   XC_CAMERA_RANGES,
   XC_DEFAULT_SRC,
@@ -11,7 +11,7 @@ import {
   normalizeXcUiOption,
   parseXcUiAutoHide,
   xcMessage
-} from "./chunk-RHQW4QZH.js";
+} from "./chunk-KL7L3MGQ.js";
 
 // src/gesture-box.ts
 var DEFAULT_RESIZE_LIMITS = Object.freeze({
@@ -1080,4 +1080,4 @@ export {
   defaultHeightCss,
   createXiaochun
 };
-//# sourceMappingURL=chunk-UWL65W7D.js.map
+//# sourceMappingURL=chunk-BN4H52OR.js.map