From a98445ce246509b0d2600df14907f72b448a6d10 Mon Sep 17 00:00:00 2001
From: Christophe Besson <cbesson@gmail.com>
Date: Thu, 17 Sep 2026 10:12:31 +0200
Subject: feat: let the viewer pick the audio track

The streaming path mapped 0:a:0 unconditionally, so a dubbed film played in
whichever language was muxed first and the others were unreachable. The node
now enumerates the tracks in stream_init and honours audio_track in
stream_req; switching is the seek path, since one ffmpeg carries one track.

MNP 3.2, additive: the player draws its selector from the node's own list and
never from a version number, so an older node is never asked for a track it
would ignore. MNP_MIN_SUPPORTED does not move.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---
 .../meshbay-common/src/meshbay_common/__init__.py  |  22 +-
 .../src/meshbay_hub/static/locales/de.js           |   2 +
 .../src/meshbay_hub/static/locales/en.js           |   2 +
 .../src/meshbay_hub/static/locales/es.js           |   2 +
 .../src/meshbay_hub/static/locales/fr.js           |   2 +
 .../src/meshbay_hub/static/locales/it.js           |   2 +
 .../src/meshbay_hub/static/locales/ja.js           |   2 +
 .../src/meshbay_hub/static/locales/nl.js           |   2 +
 .../src/meshbay_hub/static/locales/pl.js           |   2 +
 .../src/meshbay_hub/static/locales/pt-BR.js        |   2 +
 .../src/meshbay_hub/static/locales/zh-CN.js        |   2 +
 .../src/meshbay_hub/static/transport.js            |  18 +-
 .../src/meshbay_hub/static/video-player.js         | 117 ++++++++-
 .../meshbay-hub/tests/test_video_audio_track.py    | 213 +++++++++++++++
 .../src/meshbay_node/indexer/enrich.py             |   7 +-
 .../meshbay-node/src/meshbay_node/media_probe.py   |  85 +++++-
 .../src/meshbay_node/transport/webrtc_server.py    |  46 +++-
 .../tests/test_stream_audio_track_selection.py     | 285 +++++++++++++++++++++
 .../tests/test_stream_audio_transcode.py           |  27 +-
 .../tests/test_stream_video_transcode.py           |  16 +-
 20 files changed, 808 insertions(+), 48 deletions(-)
 create mode 100644 packages/meshbay-hub/tests/test_video_audio_track.py
 create mode 100644 packages/meshbay-node/tests/test_stream_audio_track_selection.py

(limited to 'packages')

diff --git a/packages/meshbay-common/src/meshbay_common/__init__.py b/packages/meshbay-common/src/meshbay_common/__init__.py
index 80e6d92..c4c4bf4 100644
--- a/packages/meshbay-common/src/meshbay_common/__init__.py
+++ b/packages/meshbay-common/src/meshbay_common/__init__.py
@@ -181,5 +181,25 @@ __version__ = "0.14.0"
 # next one it reaches and keeps its own copy meanwhile (§6.4). `MNP_MIN_SUPPORTED`
 # does not move. That is the difference from 3.0, where the requirement — not the
 # messages — is what made it MAJOR.
-MNP_VERSION = "3.1"
+# **3.2 (2026-09-17): the viewer picks the audio track.**
+#
+# `stream_init` carries `audio_tracks` (ordinal, language, title, codec,
+# channels) and the `audio_track` actually used; `stream_req` gains an optional
+# `audio_track`. Until now the streaming path mapped `0:a:0` unconditionally,
+# so a dubbed film played in whichever language was muxed first — across a real
+# library that is one language, and the others were unreachable.
+#
+# **Additive, and MINOR because nothing is required, but the reason differs
+# from 3.1's.** A playlist a node cannot store is a feature a client keeps to
+# itself; an `audio_track` a node ignores is the *wrong language*, served
+# silently, which is a wrong answer and not a missing one. What makes this
+# MINOR anyway is that the client cannot get into that position: the selector
+# is drawn from `audio_tracks` in the node's own `stream_init`, so a 3.1 node
+# sends no list, the client shows no selector, and no `audio_track` is ever
+# sent to a peer that would ignore it. **The capability is discovered from the
+# answer, never from the version number** — and that is not the opt-in
+# compatibility switch 3.0 refused, because there is no second branch on the
+# node: a 3.2 node always enumerates, always honours what it is asked for, and
+# always says which track it used. `MNP_MIN_SUPPORTED` does not move.
+MNP_VERSION = "3.2"
 MHP_VERSION = "0.1"
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/de.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/de.js
index cb92b18..806fa38 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/de.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/de.js
@@ -249,6 +249,8 @@ export default {
   'video.buffering': 'Wird gepuffert …',
   'video.resumed_at': "Fortgesetzt bei {time}",
   'video.from_start': "Von vorn beginnen",
+  'video.audio_track': 'Tonspur',
+  'video.audio_track_n': 'Spur {n}',
   'video.close': 'Schließen (Esc)',
   'preview.pdf_fallback': 'Dieser Browser zeigt das PDF nicht direkt an. Laden Sie es '
     + 'stattdessen herunter — entschlüsselt wurde es ohnehin hier.',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/en.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/en.js
index 3b84c3c..92114c9 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/en.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/en.js
@@ -248,6 +248,8 @@ export default {
   'video.buffering': 'Buffering...',
   'video.resumed_at': "Resumed at {time}",
   'video.from_start': "Start from the beginning",
+  'video.audio_track': 'Audio track',
+  'video.audio_track_n': 'Track {n}',
   'video.close': 'Close (Esc)',
   'preview.pdf_fallback': 'This browser will not display the PDF inline. Download it instead — it was decrypted here either way.',
   'preview.too_large': 'This file is {size}, more than this page can hold in memory ({limit}). Download it instead — a download is written straight to disk.',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/es.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/es.js
index 145f464..d9b8f63 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/es.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/es.js
@@ -247,6 +247,8 @@ export default {
   'video.buffering': 'Almacenando en búfer...',
   'video.resumed_at': "Reanudado en {time}",
   'video.from_start': "Empezar desde el principio",
+  'video.audio_track': 'Pista de audio',
+  'video.audio_track_n': 'Pista {n}',
   'video.close': 'Cerrar (Esc)',
   'preview.pdf_fallback': 'Este navegador no mostrará el PDF integrado. Descárguelo '
     + 'en su lugar — en cualquier caso se descifró aquí.',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/fr.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/fr.js
index a4f32de..d507d28 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/fr.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/fr.js
@@ -248,6 +248,8 @@ export default {
   'video.buffering': 'Mise en mémoire tampon...',
   'video.resumed_at': "Reprise à {time}",
   'video.from_start': "Reprendre depuis le début",
+  'video.audio_track': 'Piste audio',
+  'video.audio_track_n': 'Piste {n}',
   'video.close': 'Fermer (Échap)',
   'preview.pdf_fallback': 'Ce navigateur n’affichera pas le PDF directement. '
     + 'Téléchargez-le plutôt — il a été déchiffré ici dans les deux cas.',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/it.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/it.js
index 1da0793..b6c68dd 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/it.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/it.js
@@ -248,6 +248,8 @@ export default {
   'video.buffering': 'Buffering in corso...',
   'video.resumed_at': "Ripreso da {time}",
   'video.from_start': "Riparti dall'inizio",
+  'video.audio_track': 'Traccia audio',
+  'video.audio_track_n': 'Traccia {n}',
   'video.close': 'Chiudi (Esc)',
   'preview.pdf_fallback': 'Questo browser non mostrerà il PDF nella pagina. Lo scarichi '
     + 'invece — in ogni caso è stato decifrato qui.',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/ja.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/ja.js
index 971ecb2..99e2e1f 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/ja.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/ja.js
@@ -245,6 +245,8 @@ export default {
   'video.buffering': 'バッファリング中…',
   'video.resumed_at': "{time} から再開しました",
   'video.from_start': "最初から再生する",
+  'video.audio_track': '音声トラック',
+  'video.audio_track_n': 'トラック {n}',
   'video.close': '閉じる（Esc）',
   'preview.pdf_fallback': 'このブラウザーはページ内に PDF を表示しません。'
     + 'ダウンロードしてご覧ください。いずれにせよ復号はここで行われています。',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/nl.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/nl.js
index dae63d2..269fa70 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/nl.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/nl.js
@@ -249,6 +249,8 @@ export default {
   'video.buffering': 'Bezig met bufferen...',
   'video.resumed_at': "Hervat op {time}",
   'video.from_start': "Vanaf het begin afspelen",
+  'video.audio_track': 'Audiospoor',
+  'video.audio_track_n': 'Spoor {n}',
   'video.close': 'Sluiten (Esc)',
   'preview.pdf_fallback': 'Deze browser toont de PDF niet in de pagina zelf. Download '
     + 'hem in plaats daarvan — ontsleuteld werd hij hoe dan ook hier.',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/pl.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/pl.js
index 639d518..363296c 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/pl.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/pl.js
@@ -254,6 +254,8 @@ export default {
   'video.buffering': 'Buforowanie...',
   'video.resumed_at': "Wznowiono od {time}",
   'video.from_start': "Odtwórz od początku",
+  'video.audio_track': 'Ścieżka dźwiękowa',
+  'video.audio_track_n': 'Ścieżka {n}',
   'video.close': 'Zamknij (Esc)',
   'preview.pdf_fallback': 'Ta przeglądarka nie wyświetli pliku PDF na stronie. Proszę '
     + 'go pobrać — i tak został odszyfrowany tutaj.',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/pt-BR.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/pt-BR.js
index 371b84f..2a90956 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/pt-BR.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/pt-BR.js
@@ -249,6 +249,8 @@ export default {
   'video.buffering': 'Armazenando em buffer...',
   'video.resumed_at': "Retomado em {time}",
   'video.from_start': "Começar do início",
+  'video.audio_track': 'Faixa de áudio',
+  'video.audio_track_n': 'Faixa {n}',
   'video.close': 'Fechar (Esc)',
   'preview.pdf_fallback': 'Este navegador não exibirá o PDF na própria página. Baixe '
     + 'o arquivo — de todo modo ele foi descriptografado aqui.',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/locales/zh-CN.js b/packages/meshbay-hub/src/meshbay_hub/static/locales/zh-CN.js
index 96ba4fe..e3aa440 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/locales/zh-CN.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/locales/zh-CN.js
@@ -243,6 +243,8 @@ export default {
   'video.buffering': '正在缓冲…',
   'video.resumed_at': "已从 {time} 继续播放",
   'video.from_start': "从头开始播放",
+  'video.audio_track': '音轨',
+  'video.audio_track_n': '音轨 {n}',
   'video.close': '关闭（Esc）',
   'preview.pdf_fallback': '此浏览器不会在页面内显示该 PDF。请改为下载——无论如何它都已在本地解密。',
   'preview.too_large': '该文件为 {size}，超出本页面可在内存中保存的上限（{limit}）。请改为下载——下载会直接写入磁盘。',
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/transport.js b/packages/meshbay-hub/src/meshbay_hub/static/transport.js
index a6b3f52..262b00e 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/transport.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/transport.js
@@ -2264,12 +2264,24 @@ class MeshBayTransport {
    * MediaSource consumes a segment at a time — which is fine for a clip and
    * fatal for anything worth streaming.
    */
-  requestStream(fileId, credits = STREAM_CREDITS, start = 0) {
+  requestStream(fileId, credits = STREAM_CREDITS, start = 0, audioTrack = null) {
     // `start` is a seek: the node retires whatever this session was streaming
     // and spawns ffmpeg again from there. Omitted or zero is the film's
     // beginning, which is what an 0.1 node understands.
-    console.log('[stream] sending stream_req start:', start, 'credits:', credits);
-    this._send({ type: 'stream_req', v: '0.1', file_id: fileId, credits, start });
+    //
+    // `audioTrack` is omitted entirely unless the caller has one, and the
+    // caller only has one because a `stream_init` listed the tracks. A node
+    // too old to enumerate them is therefore never sent a field it would
+    // ignore — which matters more here than it looks: ignoring it would not
+    // degrade the stream, it would serve a different language without saying
+    // so.
+    const req = { type: 'stream_req', v: '0.1', file_id: fileId, credits, start };
+    if (Number.isInteger(audioTrack) && audioTrack >= 0) {
+      req.audio_track = audioTrack;
+    }
+    console.log('[stream] sending stream_req start:', start, 'credits:', credits,
+                'audio_track:', req.audio_track ?? '-');
+    this._send(req);
   }
 
   /** Room for `n` more segments. */
diff --git a/packages/meshbay-hub/src/meshbay_hub/static/video-player.js b/packages/meshbay-hub/src/meshbay_hub/static/video-player.js
index 4f8a4e4..4970262 100644
--- a/packages/meshbay-hub/src/meshbay_hub/static/video-player.js
+++ b/packages/meshbay-hub/src/meshbay_hub/static/video-player.js
@@ -1,7 +1,7 @@
 import {
   html, useState, useEffect, useCallback, useRef,
 } from './vendor/htm-preact.js';
-import { t } from './i18n.js';
+import { t, getLocale } from './i18n.js';
 import { Icon } from './icon.js';
 import { formatSize } from './file-utils.js';
 import { loadAuth } from './hub-client.js';
@@ -39,6 +39,63 @@ const RESUME_MIN_S = 30;
 const RESUME_MAX_FRACTION = 0.97;
 const QUEUE_HIGH_WATER = 12;
 
+// ffprobe reports a container's language tag as ISO 639-2, and in either of
+// its two variants for the dozen languages that have both — a bibliographic
+// one (fre, ger, dut) and a terminological one (fra, deu, nld), with real
+// files in this library using each. `Intl.DisplayNames` wants 639-1, so both
+// variants are folded onto the same two-letter code here. Only what a media
+// container actually carries is listed; anything unmapped falls through to the
+// raw tag, which is more useful than "Unknown".
+const _ISO639 = {
+  ara: 'ar', ben: 'bn', bul: 'bg', cat: 'ca', ces: 'cs', cze: 'cs',
+  chi: 'zh', dan: 'da', deu: 'de', dut: 'nl', ell: 'el', eng: 'en',
+  est: 'et', fas: 'fa', fin: 'fi', fra: 'fr', fre: 'fr', ger: 'de',
+  gle: 'ga', gre: 'el', heb: 'he', hin: 'hi', hrv: 'hr', hun: 'hu',
+  ice: 'is', ind: 'id', isl: 'is', ita: 'it', jpn: 'ja', kor: 'ko',
+  lav: 'lv', lit: 'lt', may: 'ms', msa: 'ms', nld: 'nl', nor: 'no',
+  per: 'fa', pol: 'pl', por: 'pt', ron: 'ro', rum: 'ro', rus: 'ru',
+  slk: 'sk', slo: 'sk', slv: 'sl', spa: 'es', srp: 'sr', swe: 'sv',
+  tam: 'ta', tha: 'th', tur: 'tr', ukr: 'uk', urd: 'ur', vie: 'vi',
+  zho: 'zh',
+};
+
+/**
+ * What to call one audio track, in the reader's language.
+ *
+ * The container's own `title` tag is preferred when there is one: a muxer that
+ * bothered to write "VFQ" or "Director's commentary" has said something the
+ * language code cannot, and two tracks tagged with the same language are
+ * otherwise indistinguishable in the menu — which is common, since a stereo
+ * downmix usually sits beside the surround track it came from.
+ */
+function audioTrackLabel(track) {
+  const code = (track.lang || '').toLowerCase();
+  let name = null;
+  const iso = _ISO639[code] || (code.length === 2 ? code : null);
+  if (iso) {
+    try {
+      name = new Intl.DisplayNames([getLocale()], { type: 'language' }).of(iso);
+      // `Intl.DisplayNames` follows each locale's prose convention, which is
+      // lower case in French, Spanish and Italian among others. A menu entry
+      // is not prose, and "français" beside "AC3 5.1" reads like a bug. Only
+      // this branch needs it: a raw tag is a code and is shown as written,
+      // and the numbered fallback comes from the catalogues already cased.
+      if (name) name = name.charAt(0).toUpperCase() + name.slice(1);
+    } catch { /* no Intl.DisplayNames, or a code it does not know */ }
+  }
+  if (!name && code && code !== 'und') name = code;
+  if (!name) name = t('video.audio_track_n', { n: track.i + 1 });
+  // Two tracks in the same language are one menu entry repeated without
+  // this, and a library where a stereo downmix sits beside the surround
+  // track it came from is the ordinary case. The container's title wins; the
+  // channel layout is the fallback, written in the "5.1" notation that needs
+  // no catalogue entry in any of the ten languages.
+  let detail = track.title;
+  if (!detail && track.ch > 2) detail = `${track.ch - 1}.1`;
+  else if (!detail && track.ch) detail = `${track.ch}.0`;
+  return detail ? `${name} — ${detail}` : name;
+}
+
 function _mseSupported(codec) {
   if (!window.MediaSource) return false;
   const mime = `video/mp4; codecs="${codec}"`;
@@ -151,6 +208,16 @@ function VideoPlayer({ entry, transportRef, gekRef, onClose, onDownload }) {
   // live; the render needs to reach it for "start from the beginning".
   const requestSeekRef = useRef(null);
   const [resumedFrom, setResumedFrom] = useState(0);
+  // The audio tracks this node reported for this file, and which one is
+  // playing. An empty list means either a file with one track or a node too
+  // old to enumerate them — both draw no selector, which is why nothing here
+  // needs to know which of the two it is.
+  const [audioTracks, setAudioTracks] = useState([]);
+  const [audioTrack, setAudioTrack] = useState(0);
+  const [audioMenuOpen, setAudioMenuOpen] = useState(false);
+  // Read inside the effect's closures, which are built once and would
+  // otherwise capture the first track forever.
+  const audioTrackRef = useRef(null);
   const [castActive, setCastActive] = useState(false);
   const [castUrl, setCastUrl] = useState(null);
   const [castPickerOpen, setCastPickerOpen] = useState(false);
@@ -412,7 +479,7 @@ function VideoPlayer({ entry, transportRef, gekRef, onClose, onDownload }) {
         outstandingRef.current = STREAM_WINDOW;
         setPhase('loading');
         console.log('[seek] request', +target.toFixed(1), 'outstanding:', STREAM_WINDOW);
-        t.requestStream(entry.id, STREAM_WINDOW, target);
+        t.requestStream(entry.id, STREAM_WINDOW, target, audioTrackRef.current);
       }, SEEK_DEBOUNCE_MS);
     };
     requestSeekRef.current = requestSeek;
@@ -557,6 +624,16 @@ function VideoPlayer({ entry, transportRef, gekRef, onClose, onDownload }) {
         const mime = `video/mp4; codecs="${msg.codec}"`;
         castCodecRef.current = msg.codec;
 
+        // What the node offered, and what it actually used — which is not
+        // always what was asked for: a file replaced on disk since the list
+        // was drawn falls back to the first track, and the selector must show
+        // the truth rather than the request.
+        setAudioTracks(Array.isArray(msg.audio_tracks) ? msg.audio_tracks : []);
+        if (Number.isInteger(msg.audio_track)) {
+          audioTrackRef.current = msg.audio_track;
+          setAudioTrack(msg.audio_track);
+        }
+
         if (!window.MediaSource || !MediaSource.isTypeSupported(mime)) {
           setError(t('video.err_mse', { codec: msg.codec }));
           setPhase('error');
@@ -769,7 +846,7 @@ function VideoPlayer({ entry, transportRef, gekRef, onClose, onDownload }) {
       outstandingRef.current = STREAM_WINDOW;
       const resumeAt = readResumePosition(entry.id);
       if (resumeAt) setResumedFrom(resumeAt);
-      transport.requestStream(entry.id, STREAM_WINDOW, resumeAt);
+      transport.requestStream(entry.id, STREAM_WINDOW, resumeAt, audioTrackRef.current);
     };
 
     // Closing the tab, or backgrounding it on a phone, never runs a React
@@ -954,6 +1031,40 @@ function VideoPlayer({ entry, transportRef, gekRef, onClose, onDownload }) {
     }}>
       <div class="video-top-bar">
         <span class="video-title">${entry.name} (${formatSize(entry.size)})</span>
+        ${audioTracks.length > 1 && html`
+          <div class="cast-wrapper" style="position:relative">
+            <button class="video-close ${audioMenuOpen ? 'cast-active' : ''}"
+              onClick=${() => setAudioMenuOpen(!audioMenuOpen)}
+              title="${t('video.audio_track')}">
+              <${Icon} name="volume" /></button>
+            ${audioMenuOpen && html`
+              <div class="cast-picker">
+                ${audioTracks.map((track) => html`
+                  <button class="cast-picker-item" onClick=${() => {
+                    setAudioMenuOpen(false);
+                    if (track.i === audioTrackRef.current) return;
+                    // A different track is a different ffmpeg, so this is the
+                    // seek path verbatim — and it has to be, because the new
+                    // stream opens with an init segment the SourceBuffer can
+                    // only accept after the abort/remove that `reinitAt` does.
+                    // Resuming where the film already was is the whole point:
+                    // the viewer changed language, not position.
+                    audioTrackRef.current = track.i;
+                    setAudioTrack(track.i);
+                    const v = videoRef.current;
+                    const seek = requestSeekRef.current;
+                    if (v && seek) seek(v.currentTime);
+                  }}>
+                    ${track.i === audioTrack
+                      ? html`<${Icon} name="check" />`
+                      : html`<span style="display:inline-block;width:14px"></span>`}
+                    ${' '}${audioTrackLabel(track)}
+                  </button>
+                `)}
+              </div>
+            `}
+          </div>
+        `}
         ${platform.capabilities.lanCast && html`
           <div class="cast-wrapper" style="position:relative">
             <button class="video-close ${castActive ? 'cast-active' : ''}"
diff --git a/packages/meshbay-hub/tests/test_video_audio_track.py b/packages/meshbay-hub/tests/test_video_audio_track.py
new file mode 100644
index 0000000..534da6d
--- /dev/null
+++ b/packages/meshbay-hub/tests/test_video_audio_track.py
@@ -0,0 +1,213 @@
+"""
+Choosing the audio track from the player.
+
+A dubbed film carries several audio tracks and the node used to map the first
+one unconditionally, so it played in whichever language was muxed first. Three
+things have to hold on this side:
+
+**The selector exists only where the node said there was something to choose.**
+It is drawn from the `audio_tracks` list in `stream_init` and from nothing else
+— there is no version check in the player — so a node too old to enumerate them
+draws no selector and is never sent an `audio_track` it would ignore. Ignoring
+it would not degrade the stream; it would serve a different language in
+silence, which is the failure this shape exists to make impossible.
+
+**Changing track is the seek path.** One ffmpeg produces one audio track, so
+there is nothing to switch inside a running stream: the node has to be asked
+again, and the new stream opens with an init segment the SourceBuffer can only
+accept after `reinitAt`'s abort/remove.
+
+**The label is derived, not translated.** ffprobe reports ISO 639-2 in either
+of its two variants, and the ten catalogues have no room for a language list —
+`Intl.DisplayNames` does that, from a fold that is real logic and is run here
+rather than read.
+"""
+
+import json
+import shutil
+import subprocess
+from pathlib import Path
+
+import pytest
+
+STATIC = Path(__file__).resolve().parents[1] / "src" / "meshbay_hub" / "static"
+APP = STATIC / "video-player.js"
+TRANSPORT = STATIC / "transport.js"
+
+pytestmark = pytest.mark.skipif(
+    shutil.which("node") is None or not APP.exists(),
+    reason="node or the SPA sources are not available")
+
+
+@pytest.fixture(scope="module")
+def app():
+    return APP.read_text()
+
+
+def _player(app: str) -> str:
+    i = app.index("function VideoPlayer(")
+    nxt = app.find("\nfunction ", i + 1)
+    return app[i:nxt if nxt > 0 else len(app)]
+
+
+def _lift(app: str, name: str) -> str:
+    """One top-level function, as text, for node to execute."""
+    start = app.index(f"function {name}(")
+    depth, i, seen = 0, start, False
+    while i < len(app):
+        if app[i] == "{":
+            depth += 1
+            seen = True
+        elif app[i] == "}":
+            depth -= 1
+            if seen and depth == 0:
+                return app[start:i + 1]
+        i += 1
+    raise AssertionError(f"{name} never closes")
+
+
+# ── The label, run rather than read ───────────────────────────────────────────
+
+def _label_cases(tmp_path, app, cases, locale="en"):
+    script = tmp_path / "label.mjs"
+    src = "\n".join([
+        app[app.index("const _ISO639 = {"):app.index("};", app.index("const _ISO639 = {")) + 2],
+        _lift(app, "audioTrackLabel"),
+    ])
+    script.write_text(
+        # The environment is modelled; the function under test is the shipped
+        # text above. `t` is only reached for a track with no usable language.
+        f"const getLocale = () => '{locale}';\n"
+        "const t = (k, p) => `${k}:${p.n}`;\n"
+        + src
+        + "\nconst out = JSON.parse(process.argv[2]).map(audioTrackLabel);\n"
+          "console.log(JSON.stringify(out));\n")
+    proc = subprocess.run(
+        ["node", str(script), json.dumps(cases)],
+        capture_output=True, text=True)
+    assert proc.returncode == 0, proc.stderr
+    return json.loads(proc.stdout)
+
+
+def test_both_iso_639_2_variants_name_the_same_language(tmp_path, app):
+    """Real files in one library use each: `fre` and `fra` are both French,
+    `ger` and `deu` both German. A fold that covers only one variant leaves
+    half a collection labelled with a three-letter code."""
+    out = _label_cases(tmp_path, app, [
+        {"i": 0, "lang": "fre", "ch": 2},
+        {"i": 1, "lang": "fra", "ch": 2},
+        {"i": 2, "lang": "ger", "ch": 2},
+        {"i": 3, "lang": "deu", "ch": 2},
+    ])
+    assert out[0] == out[1], f"fre and fra must agree: {out}"
+    assert out[2] == out[3], f"ger and deu must agree: {out}"
+    assert "French" in out[0] and "German" in out[2], out
+
+
+def test_the_container_title_wins_over_the_language_name(tmp_path, app):
+    """Two tracks in one language are the same menu entry twice without it."""
+    out = _label_cases(tmp_path, app, [
+        {"i": 0, "lang": "fre", "title": "Surround", "ch": 6},
+        {"i": 1, "lang": "fre", "title": "Stereo", "ch": 2},
+    ])
+    assert out[0] != out[1]
+    assert "Surround" in out[0] and "Stereo" in out[1]
+
+
+def test_an_untitled_track_falls_back_to_its_channel_layout(tmp_path, app):
+    """Still two distinguishable entries, and "5.1" needs no catalogue entry."""
+    out = _label_cases(tmp_path, app, [
+        {"i": 0, "lang": "fre", "ch": 6},
+        {"i": 1, "lang": "fre", "ch": 2},
+    ])
+    assert out[0] != out[1], f"two French tracks must not read alike: {out}"
+    assert "5.1" in out[0] and "2.0" in out[1], out
+
+
+def test_an_untagged_track_is_numbered_not_called_unknown(tmp_path, app):
+    """`und`, or no tag at all — common enough in real files, and a number the
+    viewer can act on beats a word that tells them nothing."""
+    out = _label_cases(tmp_path, app, [
+        {"i": 0, "lang": "und", "ch": 2},
+        {"i": 1, "ch": 2},
+    ])
+    assert out[0].startswith("video.audio_track_n:1"), out
+    assert out[1].startswith("video.audio_track_n:2"), out
+
+
+def test_a_language_the_fold_does_not_know_shows_its_tag(tmp_path, app):
+    """More useful than "Unknown": the tag is what the file actually says."""
+    out = _label_cases(tmp_path, app, [{"i": 0, "lang": "qaa", "ch": 2}])
+    assert "qaa" in out[0], out
+
+
+def test_a_language_name_is_cased_for_a_menu_not_for_prose(tmp_path, app):
+    """`Intl.DisplayNames` returns "français", which is right in a sentence and
+    wrong in a menu beside "AC3 5.1". The locales where this bites are the ones
+    this project ships most of."""
+    out = _label_cases(tmp_path, app, [
+        {"i": 0, "lang": "fre", "ch": 6},
+        {"i": 1, "lang": "eng", "ch": 6},
+    ], locale="fr")
+    assert out == ["Français — 5.1", "Anglais — 5.1"], out
+
+
+# ── The capability gate ───────────────────────────────────────────────────────
+
+def test_the_selector_is_drawn_from_the_nodes_list_and_no_version(app):
+    """The whole negotiation. A version comparison anywhere near this is the
+    bug: the floor is below the current version, so an older node is reachable,
+    and the only honest signal that it can switch track is that it enumerated."""
+    player = _player(app)
+    assert "audioTracks.length > 1" in player, \
+        "the selector must be gated on the list the node sent"
+    menu = player[player.index("audioTracks.length > 1"):]
+    menu = menu[:menu.index("platform.capabilities.lanCast")]
+    for forbidden in ("MNP_VERSION", "peerVersion", "3.2", "v_min"):
+        assert forbidden not in menu, (
+            f"the selector branches on {forbidden}, not on what the node said")
+
+
+def test_no_track_is_ever_sent_to_a_node_that_did_not_offer_one():
+    """`requestStream` must omit the field rather than default it.
+
+    A node that ignores `audio_track` does not degrade — it serves a different
+    language and says nothing. Sending 0 "harmlessly" to every node is what
+    makes that reachable, so the field has to be absent unless the caller was
+    given a real one.
+    """
+    src = TRANSPORT.read_text()
+    fn = src[src.index("requestStream(fileId"):]
+    fn = fn[:fn.index("\n  /** Room for")]
+    assert "Number.isInteger(audioTrack)" in fn, \
+        "an omitted track must not become 0 on the wire"
+    assert "req.audio_track = audioTrack" in fn
+    assert "audio_track: audioTrack" not in fn, \
+        "the field must be added conditionally, never built into the literal"
+
+
+def test_changing_track_restarts_the_stream_where_the_film_already_was(app):
+    """Setting the ref alone changes nothing: one ffmpeg carries one audio
+    track, so the node has to be asked again — and at the current position,
+    because the viewer changed language and not place."""
+    player = _player(app)
+    handler = player[player.index("audioTracks.length > 1"):]
+    handler = handler[:handler.index("platform.capabilities.lanCast")]
+    assert "audioTrackRef.current = track.i" in handler
+    assert "requestSeekRef.current" in handler, \
+        "a track change must go through the seek path"
+    assert "seek(v.currentTime)" in handler, \
+        "a track change must resume where the film already was"
+
+
+def test_the_player_believes_the_node_about_which_track_is_playing(app):
+    """`stream_init` reports the track actually used, which is not always the
+    one asked for — a list drawn before the file was replaced on disk can name
+    a track that is gone, and the node falls back to the first. Showing the
+    request instead of the answer would tick the wrong entry for the rest of
+    the film."""
+    player = _player(app)
+    assert "Number.isInteger(msg.audio_track)" in player
+    init = player[player.index("Number.isInteger(msg.audio_track)"):]
+    init = init[:init.index("}")]
+    assert "audioTrackRef.current = msg.audio_track" in init
diff --git a/packages/meshbay-node/src/meshbay_node/indexer/enrich.py b/packages/meshbay-node/src/meshbay_node/indexer/enrich.py
index 19ab9ce..6eeb1ef 100644
--- a/packages/meshbay-node/src/meshbay_node/indexer/enrich.py
+++ b/packages/meshbay-node/src/meshbay_node/indexer/enrich.py
@@ -238,11 +238,12 @@ class Enricher:
                 duration = cached_meta["duration"]
             else:
                 try:
-                    _codec, duration, _has_audio, width, height, _raw = await asyncio.wait_for(
+                    probe = await asyncio.wait_for(
                         probe_video(str(file_path)), timeout=PROBE_TIMEOUT_SECS)
+                    duration = probe.duration
                     fields["duration"] = int(duration) if duration else None
-                    fields["width"] = width
-                    fields["height"] = height
+                    fields["width"] = probe.width
+                    fields["height"] = probe.height
                 except Exception as e:
                     log.warning("Probe failed for %s: %s", file_path, e)
                 await self._media_cache.put_video_meta(
diff --git a/packages/meshbay-node/src/meshbay_node/media_probe.py b/packages/meshbay-node/src/meshbay_node/media_probe.py
index ad668fa..7858ebe 100644
--- a/packages/meshbay-node/src/meshbay_node/media_probe.py
+++ b/packages/meshbay-node/src/meshbay_node/media_probe.py
@@ -8,6 +8,7 @@ already imports from) can call it too without a circular import.
 
 import asyncio
 import json
+from dataclasses import dataclass, field
 
 _H264_PROFILES = {"Baseline": "42", "Main": "4d", "High": "64", "High 10": "6e"}
 
@@ -21,12 +22,46 @@ _H264_PROFILES = {"Baseline": "42", "Main": "4d", "High": "64", "High 10": "6e"}
 BROWSER_INCOMPATIBLE_VIDEO_CODECS = frozenset({"hevc"})
 
 
-async def probe_video(
-    path: str,
-) -> tuple[str | None, float, bool, int | None, int | None, str | None]:
+@dataclass(frozen=True)
+class AudioTrack:
     """
-    Probe video file with ffprobe, return (MSE codec string, duration,
-    has_audio, width, height, raw video codec name).
+    One selectable audio track.
+
+    **`ordinal` is the position among the audio streams, not the container
+    stream index**, because that is what `-map 0:a:<n>` takes. A file whose
+    audio sits at container indices 1, 2 and 3 has ordinals 0, 1 and 2, and
+    mapping `0:a:1` on the container index would silently serve the third
+    track — the failure this field's name exists to prevent.
+    """
+    ordinal: int
+    language: str | None
+    title: str | None
+    codec_name: str | None
+    channels: int | None
+
+
+@dataclass
+class VideoProbe:
+    """
+    What one ffprobe call says about a video file.
+
+    A dataclass rather than the tuple this used to return: the tuple had six
+    positional fields, `has_audio` sat third and `raw_codec_name` sixth, and
+    adding a seventh for the track list would have made every call site a
+    counting exercise.
+    """
+    codec: str | None
+    duration: float
+    has_audio: bool
+    width: int | None
+    height: int | None
+    raw_codec_name: str | None
+    audio_tracks: list[AudioTrack] = field(default_factory=list)
+
+
+async def probe_video(path: str) -> VideoProbe:
+    """
+    Probe a video file with ffprobe.
 
     The audio half of the codec string is always "mp4a.40.2" (AAC-LC) or
     absent — never the source's real audio codec — because the streaming
@@ -51,6 +86,13 @@ async def probe_video(
     H264. It answered it by refusing until 2026-09-09, which read to the
     operator as a broken file rather than as an unwired code path.
 
+    **Every audio track is reported, not just the first.** The streaming path
+    transcodes audio unconditionally, so serving the second track costs exactly
+    what serving the first costs and the choice is the viewer's to make; a
+    library of dubbed films is one where the first track is a language half the
+    group does not want. `has_audio` stays as the single question the muxing
+    decisions ask, and is now `bool(audio_tracks)`.
+
     width/height come from the same ffprobe call (one extra `-show_entries`
     field, no second process spawn) — resolution is deliberately never
     guessed from the filename (docs/mediacenter.md §3.5).
@@ -58,7 +100,9 @@ async def probe_video(
     from meshbay_node.platform import ffprobe_cmd
     proc = await asyncio.create_subprocess_exec(
         ffprobe_cmd(), "-v", "error",
-        "-show_entries", "stream=codec_name,profile,level,codec_type,width,height",
+        "-show_entries",
+        "stream=codec_name,profile,level,codec_type,width,height,channels",
+        "-show_entries", "stream_tags=language,title",
         "-show_entries", "format=duration",
         "-of", "json", path,
         stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE,
@@ -69,9 +113,9 @@ async def probe_video(
 
     v_codec = ""
     raw_codec_name: str | None = None
-    has_audio = False
     width: int | None = None
     height: int | None = None
+    audio_tracks: list[AudioTrack] = []
     for s in info.get("streams", []):
         if s.get("codec_type") == "video" and not v_codec:
             cn = s.get("codec_name", "")
@@ -89,9 +133,26 @@ async def probe_video(
             width = s.get("width")
             height = s.get("height")
         elif s.get("codec_type") == "audio":
-            has_audio = True
+            tags = s.get("tags") or {}
+            audio_tracks.append(AudioTrack(
+                # Counted here, never read from `s["index"]` — see AudioTrack.
+                ordinal=len(audio_tracks),
+                language=(tags.get("language") or "").strip() or None,
+                title=(tags.get("title") or "").strip() or None,
+                codec_name=s.get("codec_name") or None,
+                channels=s.get("channels"),
+            ))
 
-    if not v_codec:
-        return None, duration, has_audio, width, height, raw_codec_name
-    codec = f"{v_codec},mp4a.40.2" if has_audio else v_codec
-    return codec, duration, has_audio, width, height, raw_codec_name
+    has_audio = bool(audio_tracks)
+    codec = None
+    if v_codec:
+        codec = f"{v_codec},mp4a.40.2" if has_audio else v_codec
+    return VideoProbe(
+        codec=codec,
+        duration=duration,
+        has_audio=has_audio,
+        width=width,
+        height=height,
+        raw_codec_name=raw_codec_name,
+        audio_tracks=audio_tracks,
+    )
diff --git a/packages/meshbay-node/src/meshbay_node/transport/webrtc_server.py b/packages/meshbay-node/src/meshbay_node/transport/webrtc_server.py
index 99ba3c9..b39941d 100644
--- a/packages/meshbay-node/src/meshbay_node/transport/webrtc_server.py
+++ b/packages/meshbay-node/src/meshbay_node/transport/webrtc_server.py
@@ -5948,11 +5948,14 @@ class WebRTCPeerSession:
         file_hash = bytes.fromhex(entry.id)
 
         try:
-            codec_str, duration, has_audio, _width, _height, raw_video_codec = \
-                await _probe_video(str(file_path))
+            probe = await _probe_video(str(file_path))
         except Exception as e:
             self._send({"type": "error", "detail": f"Probe failed: {e}"})
             return
+        codec_str = probe.codec
+        duration = probe.duration
+        has_audio = probe.has_audio
+        raw_video_codec = probe.raw_codec_name
 
         # No video stream at all is the only thing this path cannot serve, and
         # it is the only thing refused here. A source with no MSE codec string
@@ -6051,8 +6054,23 @@ class WebRTCPeerSession:
             codec_str = "avc1.640029,mp4a.40.2" if has_audio else "avc1.640029"
         else:
             codec_args = ["-c:v", "copy"]
+        # Which audio track. A dubbed film carries several and the first one is
+        # not a neutral default — it is whatever the person who muxed the file
+        # happened to put first, which across a real library is overwhelmingly
+        # one language. Out of range falls back to the first rather than
+        # refusing: the client's list comes from a `stream_init` that may
+        # predate the file being replaced on disk, and a viewer who asked for
+        # the second track of a file that now has one wants the film, not an
+        # error. `stream_init` says which track was actually used, the same way
+        # it says which `start` was actually used and for the same reason.
+        try:
+            audio_track = int(msg.get("audio_track", 0) or 0)
+        except (TypeError, ValueError):
+            audio_track = 0
+        if not 0 <= audio_track < len(probe.audio_tracks):
+            audio_track = 0
         if has_audio:
-            map_args += ["-map", "0:a:0"]
+            map_args += ["-map", f"0:a:{audio_track}"]
             # Downmixed to stereo: a WEB-DL's 5.1 track becomes 6-channel AAC
             # with no "-ac", which ffprobe and VLC accept fine but which some
             # browsers' MSE decoder rejects outright once real fragments are
@@ -6081,6 +6099,22 @@ class WebRTCPeerSession:
             # this is what the client adds back (`SourceBuffer.timestampOffset`)
             # to put the fragments where they belong on the timeline.
             "start": start,
+            # The track list is how a client discovers that this node can
+            # switch language at all — there is no version check anywhere in
+            # the player. A node that does not send it gets no selector, and
+            # the client then never sends `audio_track` to a peer that would
+            # ignore it and serve the wrong language without saying so.
+            "audio_tracks": [
+                {
+                    "i": tr.ordinal,
+                    "lang": tr.language,
+                    "title": tr.title,
+                    "codec": tr.codec_name,
+                    "ch": tr.channels,
+                }
+                for tr in probe.audio_tracks
+            ],
+            "audio_track": audio_track if has_audio else None,
         })
 
         # A client that says nothing gets the old behaviour, which is why this
@@ -6097,8 +6131,10 @@ class WebRTCPeerSession:
         self._stream_started_at = time.monotonic()
         self._stream_segments = 0
         reason = "eof"
-        log.info("stream: stream_init sent file=%s paced=%s credits=%d start=%.1fs",
-                 file_id[:12], paced, self._stream_credit, start)
+        log.info("stream: stream_init sent file=%s paced=%s credits=%d start=%.1fs "
+                 "audio=%s/%d",
+                 file_id[:12], paced, self._stream_credit, start,
+                 audio_track if has_audio else "-", len(probe.audio_tracks))
         try:
             while True:
                 if paced and not await self._await_stream_credit():
diff --git a/packages/meshbay-node/tests/test_stream_audio_track_selection.py b/packages/meshbay-node/tests/test_stream_audio_track_selection.py
new file mode 100644
index 0000000..53781f9
--- /dev/null
+++ b/packages/meshbay-node/tests/test_stream_audio_track_selection.py
@@ -0,0 +1,285 @@
+"""
+The viewer picks which audio track is streamed.
+
+A dubbed film carries several audio tracks and the streaming path used to map
+`0:a:0` unconditionally, so it played in whichever language happened to be
+muxed first. Across a real library that is overwhelmingly one language, and the
+others could not be reached at all.
+
+The tracks here are told apart by **amplitude**, not by their tags: each is the
+same tone at a different volume, so an assertion about which track was served
+is measured from the decoded audio of the reassembled stream and cannot be
+satisfied by mapping the wrong one. Tags would only prove that the node copied
+a string it was given.
+
+Like the other streaming tests, these spawn real ffmpeg/ffprobe against small
+synthetic files rather than asserting against the source text.
+"""
+
+import re
+import shutil
+import subprocess
+from pathlib import Path
+
+import pytest
+from cryptography.hazmat.primitives.asymmetric.ed25519 import Ed25519PrivateKey
+
+from meshbay_common.crypto import generate_gek
+from meshbay_common.webcrypto import chunk_key_aes, decrypt_chunk_aes
+from meshbay_node.indexer.group_index import GroupIndex
+from meshbay_node.transport.webrtc_server import WebRTCPeerSession, _probe_video
+
+from conftest import needs_subprocess, one_root
+
+_HAVE_FFMPEG = shutil.which("ffmpeg") and shutil.which("ffprobe")
+pytestmark = [
+    pytest.mark.asyncio,
+    pytest.mark.skipif(not _HAVE_FFMPEG, reason="ffmpeg/ffprobe not installed"),
+    needs_subprocess,
+]
+
+# Each track's tone is attenuated by a different amount, far enough apart that
+# an AAC round trip cannot blur one into another. Index in this list is the
+# audio ordinal the node is asked for.
+_TRACK_GAIN = [1.0, 0.1, 0.01]          # 0 dB, -20 dB, -40 dB
+_TRACK_LANG = ["fre", "eng", "spa"]
+_TRACK_TITLE = ["Surround", "Original", None]
+
+
+def _make_multitrack_clip(path: Path) -> None:
+    """~1s of H264 video plus three audio tracks at descending volumes.
+
+    The video is muxed first, so the audio streams sit at container indices 1,
+    2 and 3 while their audio *ordinals* are 0, 1 and 2 — the gap that
+    `-map 0:a:<n>` is indexed by and that a probe reading `s["index"]` would
+    get wrong.
+    """
+    args = ["ffmpeg", "-hide_banner", "-loglevel", "error", "-y",
+            "-f", "lavfi", "-i", "testsrc=size=320x240:rate=25:duration=1"]
+    for gain in _TRACK_GAIN:
+        args += ["-f", "lavfi",
+                 "-i", f"sine=frequency=440:duration=1:sample_rate=48000,"
+                       f"volume={gain}"]
+    args += ["-map", "0:v:0"]
+    for i in range(len(_TRACK_GAIN)):
+        args += ["-map", f"{i + 1}:a:0"]
+    args += ["-c:v", "libx264", "-preset", "ultrafast", "-c:a", "aac"]
+    for i, lang in enumerate(_TRACK_LANG):
+        args += [f"-metadata:s:a:{i}", f"language={lang}"]
+        if _TRACK_TITLE[i]:
+            args += [f"-metadata:s:a:{i}", f"title={_TRACK_TITLE[i]}"]
+    args.append(str(path))
+    subprocess.run(args, check=True, capture_output=True)
+
+
+def _session(tmp_path: Path, video_path: Path, gek: bytes):
+    import blake3
+    file_bytes = video_path.read_bytes()
+    file_id = blake3.blake3(file_bytes).hexdigest()
+
+    sk_node = Ed25519PrivateKey.generate()
+    index = GroupIndex(group_id="g" * 32, sk_node=sk_node, gek=gek)
+    from meshbay_common.protocol import IndexEntry
+    index.add_entry(IndexEntry(
+        id=file_id, name=video_path.name, path=video_path.parent.name,
+        size=len(file_bytes), type="video", added_at=0))
+
+    session = WebRTCPeerSession.__new__(WebRTCPeerSession)
+    session._ctx = {
+        "roots": one_root(video_path.parent),
+        "index": index,
+        "gek": gek,
+        "sk_node": sk_node,
+        "max_concurrent_streams": 4,
+    }
+    session._group_id = None
+    session._user_id = "tester"
+    session._stream_stopped = False
+    session._stream_keepalives = 0
+    session.sent = []
+    session._send = session.sent.append
+    session._audit = lambda *a, **k: None
+    return session, file_id
+
+
+def _reassemble(sent: list[dict], gek: bytes, file_id: str) -> bytes:
+    file_hash = bytes.fromhex(file_id)
+    segments = sorted(
+        (m for m in sent if m.get("type") == "stream_data"),
+        key=lambda m: m["segment_index"])
+    out = b""
+    for m in segments:
+        key = chunk_key_aes(gek, file_hash, m["segment_index"])
+        out += decrypt_chunk_aes(key, m["nonce"], m["ct"])
+    return out
+
+
+def _mean_volume_db(path: Path) -> float:
+    """What ffmpeg's volumedetect measures in the decoded audio."""
+    proc = subprocess.run(
+        ["ffmpeg", "-hide_banner", "-i", str(path), "-af", "volumedetect",
+         "-f", "null", "-"],
+        capture_output=True, text=True)
+    m = re.search(r"mean_volume:\s*(-?\d+(?:\.\d+)?) dB", proc.stderr)
+    assert m, f"volumedetect said nothing usable: {proc.stderr[-400:]}"
+    return float(m.group(1))
+
+
+async def _stream(tmp_path: Path, clip: Path, msg_extra: dict):
+    gek = generate_gek()
+    session, file_id = _session(tmp_path, clip, gek)
+    await session._stream_video_inner(
+        {"file_id": file_id, "start": 0, "credits": 0, **msg_extra})
+    errors = [m for m in session.sent if m.get("type") == "error"]
+    assert not errors, f"streaming must not fail: {errors}"
+    init = next(m for m in session.sent if m.get("type") == "stream_init")
+    return session, file_id, gek, init
+
+
+async def _streamed_volume(tmp_path: Path, clip: Path, msg_extra: dict, tag: str):
+    session, file_id, gek, init = await _stream(tmp_path, clip, msg_extra)
+    out = tmp_path / f"out-{tag}.mp4"
+    out.write_bytes(_reassemble(session.sent, gek, file_id))
+    return _mean_volume_db(out), init
+
+
+async def test_probe_enumerates_every_track_by_ordinal_not_container_index(tmp_path):
+    """The trap this feature is one wrong line away from.
+
+    `-map 0:a:1` counts audio streams; `s["index"]` counts every stream in the
+    container. With a video stream muxed first the two never agree, and a probe
+    reporting container indices would make the client ask for track 1 and be
+    served track 2 — silently, since both are real audio.
+    """
+    clip = tmp_path / "clip.mkv"
+    _make_multitrack_clip(clip)
+
+    probe = await _probe_video(str(clip))
+
+    assert [tr.ordinal for tr in probe.audio_tracks] == [0, 1, 2]
+    assert [tr.language for tr in probe.audio_tracks] == _TRACK_LANG
+    assert probe.has_audio is True
+    # The container really does disagree, or the assertion above proves nothing.
+    raw = subprocess.run(
+        ["ffprobe", "-v", "error", "-select_streams", "a",
+         "-show_entries", "stream=index", "-of", "csv=p=0", str(clip)],
+        check=True, capture_output=True, text=True)
+    assert [int(x) for x in raw.stdout.split()] == [1, 2, 3], (
+        "the fixture must put audio at container indices that differ from the "
+        "ordinals, or it cannot tell the two apart")
+
+
+async def test_probe_reports_the_title_tag_when_the_muxer_wrote_one(tmp_path):
+    """Two tracks in one language are one menu entry repeated without it."""
+    clip = tmp_path / "clip.mkv"
+    _make_multitrack_clip(clip)
+
+    probe = await _probe_video(str(clip))
+
+    assert [tr.title for tr in probe.audio_tracks] == _TRACK_TITLE
+    assert all(tr.channels == 1 for tr in probe.audio_tracks)
+
+
+async def test_the_requested_audio_track_is_the_one_streamed(tmp_path):
+    """Measured from the decoded audio, not from a tag the node echoed back."""
+    clip = tmp_path / "clip.mkv"
+    _make_multitrack_clip(clip)
+
+    first, init_first = await _streamed_volume(tmp_path, clip, {"audio_track": 0}, "0")
+    third, init_third = await _streamed_volume(tmp_path, clip, {"audio_track": 2}, "2")
+
+    assert init_first["audio_track"] == 0
+    assert init_third["audio_track"] == 2
+    # -40 dB of attenuation between them; anything under 15 dB of measured
+    # separation means the same track was served twice.
+    assert first - third > 15, (
+        f"track 0 ({first} dB) and track 2 ({third} dB) decoded to the same "
+        "loudness, so the requested track was not the one mapped")
+
+
+async def test_no_audio_track_asked_for_still_means_the_first(tmp_path):
+    """A client that says nothing gets exactly what it got before."""
+    clip = tmp_path / "clip.mkv"
+    _make_multitrack_clip(clip)
+
+    silent, init_silent = await _streamed_volume(tmp_path, clip, {}, "default")
+    explicit, _ = await _streamed_volume(tmp_path, clip, {"audio_track": 0}, "explicit")
+
+    assert init_silent["audio_track"] == 0
+    assert abs(silent - explicit) < 2
+
+
+async def test_an_out_of_range_track_falls_back_to_the_first_and_says_so(tmp_path):
+    """The client's list can predate the file being replaced on disk.
+
+    A viewer who asked for the second track of a file that now has one wants
+    the film, not an error — and `stream_init` has to report the track actually
+    used, the same way it reports the `start` actually used.
+    """
+    clip = tmp_path / "clip.mkv"
+    _make_multitrack_clip(clip)
+
+    got, init = await _streamed_volume(tmp_path, clip, {"audio_track": 9}, "oob")
+    expected, _ = await _streamed_volume(tmp_path, clip, {"audio_track": 0}, "base")
+
+    assert init["audio_track"] == 0, "stream_init must not echo the impossible request"
+    assert abs(got - expected) < 2
+
+
+async def test_a_malformed_track_number_is_not_an_error(tmp_path):
+    clip = tmp_path / "clip.mkv"
+    _make_multitrack_clip(clip)
+
+    for bad in ("two", None, -1, 1.5):
+        _, _, _, init = await _stream(tmp_path, clip, {"audio_track": bad})
+        assert init["audio_track"] in (0, 1), f"{bad!r} produced {init['audio_track']!r}"
+
+
+async def test_stream_init_lists_the_tracks_for_the_client_to_choose_from(tmp_path):
+    """The list is the whole capability negotiation.
+
+    The player draws its selector from this and from nothing else — there is no
+    version check in it — so a node that sends no list gets no selector and is
+    never sent an `audio_track` it would ignore and answer in the wrong
+    language.
+    """
+    clip = tmp_path / "clip.mkv"
+    _make_multitrack_clip(clip)
+
+    _, _, _, init = await _stream(tmp_path, clip, {})
+
+    assert [tr["i"] for tr in init["audio_tracks"]] == [0, 1, 2]
+    assert [tr["lang"] for tr in init["audio_tracks"]] == _TRACK_LANG
+    assert init["audio_tracks"][0]["title"] == "Surround"
+    assert init["audio_tracks"][2]["title"] is None
+
+
+async def test_a_single_track_file_reports_a_list_of_one(tmp_path):
+    """Not an empty list: the player decides on length, and a one-entry list is
+    how it knows there is nothing to choose rather than nothing to report."""
+    clip = tmp_path / "mono.mkv"
+    subprocess.run(
+        ["ffmpeg", "-hide_banner", "-loglevel", "error", "-y",
+         "-f", "lavfi", "-i", "testsrc=size=320x240:rate=25:duration=1",
+         "-f", "lavfi", "-i", "sine=frequency=440:duration=1:sample_rate=48000",
+         "-c:v", "libx264", "-preset", "ultrafast", "-c:a", "aac", str(clip)],
+        check=True, capture_output=True)
+
+    _, _, _, init = await _stream(tmp_path, clip, {})
+
+    assert len(init["audio_tracks"]) == 1
+    assert init["audio_track"] == 0
+
+
+async def test_a_file_with_no_audio_reports_no_track_at_all(tmp_path):
+    clip = tmp_path / "silent.mkv"
+    subprocess.run(
+        ["ffmpeg", "-hide_banner", "-loglevel", "error", "-y",
+         "-f", "lavfi", "-i", "testsrc=size=320x240:rate=25:duration=1",
+         "-c:v", "libx264", "-preset", "ultrafast", "-an", str(clip)],
+        check=True, capture_output=True)
+
+    _, _, _, init = await _stream(tmp_path, clip, {"audio_track": 1})
+
+    assert init["audio_tracks"] == []
+    assert init["audio_track"] is None
diff --git a/packages/meshbay-node/tests/test_stream_audio_transcode.py b/packages/meshbay-node/tests/test_stream_audio_transcode.py
index 959cc91..92518ba 100644
--- a/packages/meshbay-node/tests/test_stream_audio_transcode.py
+++ b/packages/meshbay-node/tests/test_stream_audio_transcode.py
@@ -150,15 +150,15 @@ async def test_probe_video_reports_aac_regardless_of_source_audio_codec(tmp_path
     clip = tmp_path / "clip.mkv"
     _make_clip(clip, acodec="eac3", channels=6)
 
-    codec, duration, has_audio, width, height, raw_codec = await _probe_video(str(clip))
+    probe = await _probe_video(str(clip))
 
-    assert has_audio is True
-    assert duration > 0
-    assert codec is not None
-    assert "eac3" not in codec and "ec-3" not in codec
-    assert "mp4a.40.2" in codec
-    assert (width, height) == (320, 240)
-    assert raw_codec == "h264"
+    assert probe.has_audio is True
+    assert probe.duration > 0
+    assert probe.codec is not None
+    assert "eac3" not in probe.codec and "ec-3" not in probe.codec
+    assert "mp4a.40.2" in probe.codec
+    assert (probe.width, probe.height) == (320, 240)
+    assert probe.raw_codec_name == "h264"
 
 
 async def test_probe_video_handles_no_audio_track(tmp_path):
@@ -170,10 +170,11 @@ async def test_probe_video_handles_no_audio_track(tmp_path):
         check=True, capture_output=True,
     )
 
-    codec, duration, has_audio, width, height, raw_codec = await _probe_video(str(clip))
+    probe = await _probe_video(str(clip))
 
-    assert has_audio is False
-    assert codec is not None and "," not in codec, \
+    assert probe.has_audio is False
+    assert probe.audio_tracks == []
+    assert probe.codec is not None and "," not in probe.codec, \
         "no audio track must not produce a dangling ',' or a fake audio codec"
-    assert (width, height) == (320, 240)
-    assert raw_codec == "h264"
+    assert (probe.width, probe.height) == (320, 240)
+    assert probe.raw_codec_name == "h264"
diff --git a/packages/meshbay-node/tests/test_stream_video_transcode.py b/packages/meshbay-node/tests/test_stream_video_transcode.py
index 69f4c2d..b165af4 100644
--- a/packages/meshbay-node/tests/test_stream_video_transcode.py
+++ b/packages/meshbay-node/tests/test_stream_video_transcode.py
@@ -191,10 +191,10 @@ async def test_probe_video_reports_raw_codec_name_for_hevc(tmp_path):
     clip = tmp_path / "clip.mkv"
     _make_hevc_clip(clip)
 
-    codec, duration, has_audio, width, height, raw_codec = await _probe_video(str(clip))
+    probe = await _probe_video(str(clip))
 
-    assert raw_codec == "hevc"
-    assert codec is not None and codec.startswith("hev1.")
+    assert probe.raw_codec_name == "hevc"
+    assert probe.codec is not None and probe.codec.startswith("hev1.")
 
 
 @needs_mp3
@@ -206,13 +206,13 @@ async def test_probe_video_reports_no_codec_string_for_mpeg4(tmp_path):
     clip = tmp_path / "clip.avi"
     _make_mpeg4_clip(clip)
 
-    codec, duration, has_audio, width, height, raw_codec = await _probe_video(str(clip))
+    probe = await _probe_video(str(clip))
 
-    assert raw_codec == "mpeg4"
-    assert codec is None, (
+    assert probe.raw_codec_name == "mpeg4"
+    assert probe.codec is None, (
         "a codec string for MPEG-4 Part 2 would be one no browser can act on")
-    assert has_audio is True
-    assert (width, height) == (320, 240)
+    assert probe.has_audio is True
+    assert (probe.width, probe.height) == (320, 240)
 
 
 @needs_mp3
-- 
cgit v1.2.3