Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

fix(demo): give a silent chapter its own audio track so the film still joins

nirholase2d15724e4c9
diff --git a/scripts/lib/demo-stage.mjs b/scripts/lib/demo-stage.mjs
index f29ba643b..747bd818d 100644
--- a/scripts/lib/demo-stage.mjs
+++ b/scripts/lib/demo-stage.mjs
@@ -651,7 +651,15 @@ export class Narrator {
* a mix, which keeps it sample accurate over an hour-long tour.
*/
buildTrack(out, totalMs) {
- if (!this.clips.length) return null;
+ /* A chapter with a voice track and a chapter without one cannot be joined
+ by a stream copy, so a chapter that somehow spoke no line still gets a
+ silent track of its own length rather than no audio stream at all. */
+ if (!this.enabled) return null;
+ if (!this.clips.length) {
+ ffmpeg(['-f', 'lavfi', '-t', (totalMs / 1000).toFixed(3),
+ '-i', 'anullsrc=r=44100:cl=mono', '-c:a', 'pcm_s16le', out]);
+ return out;
+ }
const work = path.join(this.dir, `.track-${createHash('sha1').update(out).digest('hex').slice(0, 8)}`);
mkdirSync(work, { recursive: true });
const parts = [];