ptzreplay

server/src/services/ptzCameraService/index.js

@@ -25,7 +25,9 @@ const DEFAULT_ONVIF_PORT = 8000;
 const DEFAULT_PROFILE_TOKEN = '003';
 const DEFAULT_TURN_DURATION_MS = 5 * 60 * 1000;
 const DOCK_GRACE_MS = 60 * 1000;
-const DEFAULT_REPLAY_ENABLED = false;
+// PTZ is a normal replay source now, so capture should be on unless the feature
+// explicitly disables replay for the camera.
+const DEFAULT_REPLAY_ENABLED = true;
 const SNAPSHOT_DIR = process.env.ROVER_SNAPSHOT_DIR || '/var/lib/rover-snapshots';
 const SNAPSHOT_POLL_MS = 300;
 const SNAPSHOT_STREAM_INTERVAL_MS = 2000;
@@ -899,21 +901,36 @@ module.exports = {
   getReplaySource: () => enabled && isReplayEnabled()
     ? { type: 'ptz', id: PTZ_CAMERA_ID, label: cameraConfig.name || 'PTZ Camera' }
     : null,
-  getReplayWorkerSource: () => {
+  getReplayWorkerSources: () => {
     /*
-      PTZ replay capture is optional because the server ffmpeg build must be
-      able to produce browser/Discord-friendly replay segments. The camera live
-      feed can remain raw for MediaMTX while replay capture is left off until
-      the actual server has a working encoder or a copy-only PTZ replay path is
-      intentionally designed.
+      PTZ replay uses two internal workers from the same MediaMTX path. The
+      selectable replay source stays "ptz:ptz-camera", while the segment engine
+      records video and audio separately so replayBuilder can mix PTZ microphone
+      audio the same way it already mixes rover audio.
     */
-    if (!enabled || !isReplayEnabled()) return null;
-    return {
-      id: PTZ_CAMERA_ID,
-      sourceType: 'ptz',
-      kind: 'video',
-      label: cameraConfig.name || 'PTZ Camera',
-      inputUrl: `srt://127.0.0.1:9000?streamid=read:${encodeURIComponent(PTZ_STREAM_PATH)}`,
-    };
+    if (!enabled || !isReplayEnabled()) return [];
+    const inputUrl = `srt://127.0.0.1:9000?streamid=read:${encodeURIComponent(PTZ_STREAM_PATH)}`;
+    const label = cameraConfig.name || 'PTZ Camera';
+    return [
+      {
+        id: PTZ_CAMERA_ID,
+        sourceType: 'ptz',
+        kind: 'video',
+        label,
+        inputUrl,
+      },
+      {
+        id: `${PTZ_CAMERA_ID}-audio`,
+        sourceType: 'ptz',
+        sourceId: PTZ_CAMERA_ID,
+        kind: 'audio',
+        label: `${label} audio`,
+        inputUrl,
+      },
+    ];
+  },
+  getReplayWorkerSource: () => {
+    const [videoSource] = module.exports.getReplayWorkerSources();
+    return videoSource || null;
   },
 };

server/src/services/replayEngineV2/index.js

@@ -49,7 +49,7 @@ const replayBuilder = createReplayBuilder({
   ensureDir,
   renderSidebarVideo: sidebarRenderer.renderSidebarVideo,
   getVideoEntriesForSource: segmentStore.getVideoEntriesForSource,
-  getAudioEntriesForRover: segmentStore.getAudioEntriesForRover,
+  getAudioEntriesForSource: segmentStore.getAudioEntriesForSource,
   overlapping: segmentStore.overlapping,
 });
 registerReplaySocketHooks({ tryTriggerReplay, validateSources, getDefaultWebSources });

server/src/services/replayEngineV2/replayBuilder.js

@@ -51,7 +51,7 @@ function buildChatEventsForWindow(startMs, endMs, limit = 22, preWindowCount = 1
   return [...beforeWindow, ...inWindow].sort((a, b) => a.ts - b.ts);
 }
 
-function createReplayBuilder({ execFileAsync, fsp, ensureDir, renderSidebarVideo, getVideoEntriesForSource, getAudioEntriesForRover, overlapping }) {
+function createReplayBuilder({ execFileAsync, fsp, ensureDir, renderSidebarVideo, getVideoEntriesForSource, getAudioEntriesForSource, overlapping }) {
   function resolveReplayWindow({ sources = [], nowMs, guardMs, durationMs }) {
     const tentativeEnd = nowMs - guardMs;
     const sourceEnds = [];
@@ -122,18 +122,22 @@ function createReplayBuilder({ execFileAsync, fsp, ensureDir, renderSidebarVideo
         normalizedVideos.push({ path: videoTrimmed, source });
         usedSources.push(source);
 
-        if (source.type === 'rover') {
-          const audioEntries = overlapping(getAudioEntriesForRover(sourceId), tStart, tEnd);
-          if (audioEntries.length) {
-            const audioConcat = path.join(tmpDir, `audio-${i}.m4a`);
-            await concatFiles(audioEntries.map((entry) => entry.filePath), audioConcat);
-            const audioTrimmed = path.join(tmpDir, `audio-${i}.trim.m4a`);
-            const firstAudioStartMs = audioEntries[0].startMs;
-            const ass = Math.max(0, (tStart - firstAudioStartMs) / 1000);
-            const ato = Math.max(ass + 0.1, (tEnd - firstAudioStartMs) / 1000);
-            await execFileAsync(FFMPEG_BIN, ['-y','-hide_banner','-loglevel','error','-ss',ass.toFixed(3),'-to',ato.toFixed(3),'-i',audioConcat,'-vn','-ac','1','-ar','48000','-c:a','aac','-b:a','96k',audioTrimmed]);
-            normalizedAudios.push(audioTrimmed);
-          }
+        const audioEntries = overlapping(getAudioEntriesForSource(source), tStart, tEnd);
+        if (audioEntries.length) {
+          /*
+            Audio workers are separate from selected video sources, even for PTZ
+            where the live camera path carries inline Opus. Trim the matching
+            source-owned audio window here and let the final graph mix every
+            selected source's audio together.
+          */
+          const audioConcat = path.join(tmpDir, `audio-${i}.m4a`);
+          await concatFiles(audioEntries.map((entry) => entry.filePath), audioConcat);
+          const audioTrimmed = path.join(tmpDir, `audio-${i}.trim.m4a`);
+          const firstAudioStartMs = audioEntries[0].startMs;
+          const ass = Math.max(0, (tStart - firstAudioStartMs) / 1000);
+          const ato = Math.max(ass + 0.1, (tEnd - firstAudioStartMs) / 1000);
+          await execFileAsync(FFMPEG_BIN, ['-y','-hide_banner','-loglevel','error','-ss',ass.toFixed(3),'-to',ato.toFixed(3),'-i',audioConcat,'-vn','-ac','1','-ar','48000','-c:a','aac','-b:a','96k',audioTrimmed]);
+          normalizedAudios.push(audioTrimmed);
         }
       }
 

server/src/services/replayEngineV2/segmentStore.js

@@ -97,6 +97,24 @@ function createSegmentStore({ getActiveSegmentRoot }) {
     return segmentIndex.get(key) || [];
   }
 
+  function getAudioEntriesForSource(source) {
+    /*
+      Rover audio is published as a separate "<rover>-audio" stream, while PTZ
+      replay audio is split into an internal "ptz-camera-audio" worker from the
+      same live camera stream. Keep this mapping close to the segment index so
+      replayBuilder can ask for "audio that belongs to this selected source"
+      without knowing every worker naming convention.
+    */
+    const type = String(source?.type || '');
+    const id = String(source?.id || '');
+    if (type === 'rover') return getAudioEntriesForRover(id);
+    if (type === 'ptz') {
+      const key = sourceKey({ sourceType: 'ptz', kind: 'audio', id: `${id}-audio` });
+      return segmentIndex.get(key) || [];
+    }
+    return [];
+  }
+
   function overlapping(entries, startMs, endMs) {
     return entries.filter((entry) => entry.endMs > startMs && entry.startMs < endMs);
   }
@@ -157,6 +175,7 @@ function createSegmentStore({ getActiveSegmentRoot }) {
     cleanupOldFiles,
     getVideoEntriesForSource,
     getAudioEntriesForRover,
+    getAudioEntriesForSource,
     overlapping,
     bootstrapIndexFromDisk,
     getReplayHealthSnapshot,

server/src/services/replayEngineV2/sources.js

@@ -47,8 +47,10 @@ function listDesiredSources() {
     if (!streamUrl) continue;
     sources.push({ id: String(camera.id), sourceType: 'room', kind: 'video', label: camera.name || camera.id, inputUrl: streamUrl });
   }
-  const ptzSource = ptzCameraService.getReplayWorkerSource();
-  if (ptzSource) sources.push(ptzSource);
+  // The PTZ service owns its own worker list because video and microphone audio
+  // both come from the same live MediaMTX path, unlike rovers where audio is a
+  // separate published stream.
+  sources.push(...ptzCameraService.getReplayWorkerSources());
   return sources;
 }