Sync the picture to the sound people hear, not the sound queued

The diagnostics came back clean — aheadBy steady at 3.3 s, nothing
dropped, nothing re-pegged — so the remaining lip-sync error was a
constant, and the cause is the clock: media.seconds counts audio handed
to LiveKit, whose audio source holds up to a second (its default queue)
before anyone hears it. Frames released against that ran a queue-length
early, every time.

The clock now subtracts what is still queued, which the source reports
directly, so the compensation follows the real buffer instead of a
guessed constant.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

src/core/player.ts

@@ -383,10 +383,10 @@ export class GuildPlayer extends EventEmitter<GuildPlayerEvents> {
       width: video.width,
       height: video.height,
       fps: video.fps,
-      // Frames are released against the sound that has actually played, which is
+      // Frames are released against the sound people have actually heard, not
-      // the only clock both sides share: the audio path runs through revoice's
+      // the sound handed to LiveKit: its audio source buffers up to a second,
-      // own ffmpeg and buffer, so it always trails the raw frames.
+      // and gating on the queued position showed every frame that much early.
-      audioClock: () => this.media?.seconds ?? 0,
+      audioClock: () => this.heardSeconds(),
     });
     try {
       await publisher.start(room, video.stream, (reason) => {
@@ -402,6 +402,14 @@ export class GuildPlayer extends EventEmitter<GuildPlayerEvents> {
     }
   }
 
+  /** Playback position as heard, i.e. minus whatever is still queued in LiveKit. */
+  private heardSeconds(): number {
+    const media = this.media;
+    if (!media) return 0;
+    const queuedMs = media.source?.queuedDuration ?? 0;
+    return Math.max(0, media.seconds - queuedMs / 1000);
+  }
+
   private stopScreenShare(): void {
     const publisher = this.videoPublisher;
     this.videoPublisher = null;

src/core/revoice.ts

@@ -9,6 +9,12 @@ const require = createRequire(import.meta.url);
 
 export interface MediaPlayerLike {
   readonly seconds: number;
+  /**
+   * LiveKit's audio source. `seconds` counts what has been handed to it, and it
+   * holds up to a second before anyone hears it — `queuedDuration` (in ms) is
+   * that gap, which matters when lining a picture up with the sound.
+   */
+  source?: { readonly queuedDuration: number } | null;
   readonly duration: number;
   codecData?: { duration?: string } | null;
   paused: boolean;