feat(plex): multi-rendition audio (client-side switch) + reliable resume-on-F5

- Multi-audio items now ship every audio track as an HLS rendition in one session
  (stream.py var_stream_map -> master.m3u8), so hls.js switches audio CLIENT-SIDE with
  no ffmpeg restart, same timeline, no gap/drift. /session gains ?multi=1 (forces HLS
  even for direct-playable multi-audio files); K is probed from the video variant seg.
- Restore the selected audio track on AUDIO_TRACKS_UPDATED, not MANIFEST_PARSED (the
  renditions aren't parsed yet on MANIFEST_PARSED, so the set was dropped -> after F5 the
  UI showed the restored track but playback stayed on the default).
- Resume position now survives F5: a pagehide keepalive beacon (plexProgressBeacon)
  saves the current position on reload/close/navigate, since React effect cleanup does
  not run on a full reload; seekTo writes the target into absRef immediately so a save
  right after a seek is accurate.
This commit is contained in:
npeter83 2026-07-08 22:25:03 +02:00
parent 0a0703b769
commit 3c622fd44c
4 changed files with 155 additions and 43 deletions

View file

@ -63,40 +63,77 @@ def _seek_opts(start_s: float) -> list[str]:
return ["-noaccurate_seek", "-ss", f"{start_s:.3f}"] if start_s > 0 else []
def _ffmpeg_cmd(src: Path, start_s: float, out_dir: Path, audio_ord: int | None, aoff: float = 0.0) -> list[str]:
ao = audio_ord if audio_ord is not None else 0
_HLS_TAIL = [
"-f", "hls",
"-hls_time", str(_SEG_SECONDS),
"-hls_list_size", "0",
# EVENT (not VOD): ffmpeg writes/appends the playlist AS segments complete, so playback can
# start immediately from this session's offset. VOD only writes the playlist at the end. The
# full seekbar comes from our known duration + the seek-restart model, not from the playlist.
"-hls_playlist_type", "event",
"-hls_segment_type", "mpegts",
"-hls_flags", "independent_segments+temp_file",
]
def _probe_audio_langs(src: Path) -> list[str]:
"""One entry per audio stream (its language tag, or "" if untagged). len() = audio-track count.
Used to build the multi-rendition var_stream_map. Empty on probe failure single-audio path."""
try:
out = subprocess.run(
["ffprobe", "-v", "error", "-select_streams", "a", "-show_entries", "stream_tags=language",
"-of", "csv=p=0", str(src)],
capture_output=True, text=True, timeout=10,
)
return [ln.strip() for ln in out.stdout.splitlines()]
except (subprocess.SubprocessError, OSError):
return []
def _ffmpeg_cmd(
src: Path, start_s: float, out_dir: Path, audio_ord: int | None,
aoff: float = 0.0, audio_langs: list[str] | None = None,
) -> list[str]:
args = ["ffmpeg", "-nostdin", "-loglevel", "error"]
# Keep ORIGINAL timestamps on the output so the first segment carries the true absolute PTS of the
# keyframe ffmpeg actually seeked to. hls.js still zero-bases playback to that PTS; we read it back
# (see _probe_first_pts) to learn the real content offset. Without -copyts the segment PTS is
# rewritten toward 0 and the true offset is unknowable, which caused the ~seconds subtitle lead.
args += _seek_opts(start_s) + ["-copyts", "-i", str(src)]
ai = 0 # input index the audio maps come from
if aoff:
# A/V-sync offset: read the file a SECOND time for audio only, with `-itsoffset` shifting the
# audio PTS by ±aoff relative to the (input-0) video. Only when the user set a non-zero offset
# (default single-input path is unchanged, no double read). Maps audio from input 1 then.
# audio PTS by ±aoff relative to the (input-0) video. Maps audio from input 1 then.
args += ["-itsoffset", f"{aoff:.3f}"] + _seek_opts(start_s) + ["-copyts", "-i", str(src)]
amap = f"1:a:{ao}?"
ai = 1
# Video always stream-copied; audio → AAC. Subtitles are NOT muxed here — they're separate WebVTT
# tracks (GET /subtitle) the browser overlays, so choosing a subtitle doesn't restart this session.
if audio_langs:
# MULTI-RENDITION: map EVERY audio track as an alternate HLS rendition in one group, so hls.js
# switches audio CLIENT-SIDE on the same timeline (no session restart, no drift) — the premium
# audio-switch path. Produces master.m3u8 + stream_%v.m3u8 (v0=video, v1..=audio) + seg_%v_%d.ts.
args += ["-map", "0:v:0"]
for i in range(len(audio_langs)):
args += ["-map", f"{ai}:a:{i}?"]
args += ["-sn", "-c:v", "copy", "-c:a", "aac", "-ac", "2", "-b:a", "192k"]
var = ["v:0,agroup:aud"]
for i, lang in enumerate(audio_langs):
entry = f"a:{i},agroup:aud,name:a{i}"
if lang:
entry += f",language:{lang}"
if i == 0:
entry += ",default:yes"
var.append(entry)
args += _HLS_TAIL + [
"-master_pl_name", "master.m3u8",
"-var_stream_map", " ".join(var),
"-hls_segment_filename", str(out_dir / "seg_%v_%d.ts"), str(out_dir / "stream_%v.m3u8"),
]
else:
amap = f"0:a:{ao}?"
# Video always stream-copied; the selected audio track (default = first) is transcoded to AAC.
# Subtitles are NOT muxed here — they're served as standalone WebVTT tracks (GET /subtitle) that
# the browser overlays, so choosing a subtitle no longer restarts this session (and external
# sidecar subs, which aren't in the file, work — the old `-map 0:s` broke on those).
args += ["-map", "0:v:0", "-map", amap, "-sn"]
args += ["-c:v", "copy", "-c:a", "aac", "-ac", "2", "-b:a", "192k"]
args += [
"-f", "hls",
"-hls_time", str(_SEG_SECONDS),
"-hls_list_size", "0",
# EVENT (not VOD): ffmpeg writes/appends the playlist AS segments complete, so playback can
# start immediately from this session's offset. VOD only writes the playlist at the end. The
# full seekbar comes from our known duration + the seek-restart model, not from the playlist.
"-hls_playlist_type", "event",
"-hls_segment_type", "mpegts",
"-hls_flags", "independent_segments+temp_file",
]
args += ["-hls_segment_filename", str(out_dir / "seg_%d.ts"), str(out_dir / "index.m3u8")]
ao = audio_ord if audio_ord is not None else 0
args += ["-map", "0:v:0", "-map", f"{ai}:a:{ao}?", "-sn"]
args += ["-c:v", "copy", "-c:a", "aac", "-ac", "2", "-b:a", "192k"]
args += _HLS_TAIL + ["-hls_segment_filename", str(out_dir / "seg_%d.ts"), str(out_dir / "index.m3u8")]
return args
@ -146,43 +183,48 @@ def start_session(
start_s: float,
audio_ord: int | None = None,
aoff: float = 0.0,
multi: bool = False,
) -> HlsSession | None:
"""(Re)start the HLS remux for an item at the given offset, with an optional selected audio track
(ordinal among audio streams). Subtitles are separate WebVTT tracks (not muxed here). Returns
None if the local file can't be read."""
"""(Re)start the HLS remux for an item at the given offset. `multi` (item has >1 audio track) maps
every audio track as an HLS rendition so the client switches audio without a restart; otherwise a
single audio track (`audio_ord`) is muxed. Subtitles are separate WebVTT tracks (not muxed here).
Returns None if the local file can't be read."""
src = paths.local_media_path(db, item.file_path)
if src is None:
return None
key = item.rating_key
start_s = max(0.0, float(start_s))
audio_langs = _probe_audio_langs(src) if multi else []
is_multi = len(audio_langs) > 1 # only worth a master playlist when there really are ≥2 tracks
with _lock:
old = _sessions.pop(key, None)
if old is not None:
_kill(old)
directory = _HLS_ROOT / f"{key}_{int(start_s)}_{audio_ord}_{aoff:+.2f}"
tag = "multi" if is_multi else str(audio_ord)
directory = _HLS_ROOT / f"{key}_{int(start_s)}_{tag}_{aoff:+.2f}"
shutil.rmtree(directory, ignore_errors=True)
directory.mkdir(parents=True, exist_ok=True)
proc = subprocess.Popen(
_ffmpeg_cmd(src, start_s, directory, audio_ord, aoff),
_ffmpeg_cmd(src, start_s, directory, audio_ord, aoff, audio_langs if is_multi else None),
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL,
)
s = HlsSession(key, directory, proc, start_s, "index.m3u8")
s = HlsSession(key, directory, proc, start_s, "master.m3u8" if is_multi else "index.m3u8")
_sessions[key] = s
_enforce_cap()
# Learn the REAL start offset K from the first segment (outside the lock — this blocks on ffmpeg
# producing seg_0, and we must not hold up other sessions). The client reads `media_start_s`, so
# Learn the REAL start offset K from the first VIDEO segment (outside the lock — this blocks on
# ffmpeg producing it, and we must not hold up other sessions). The client reads `media_start_s`, so
# its absolute clock + subtitle shift key off the true keyframe, not the requested offset. On
# timeout/probe failure we keep the provisional requested offset (old behaviour, small drift).
if start_s > 0:
seg0 = directory / "seg_0.ts"
seg0 = directory / ("seg_0_0.ts" if is_multi else "seg_0.ts") # multi: video variant = v0
if wait_for(seg0, timeout=25.0):
k = _probe_first_pts(seg0)
if k is not None and k >= 0:
s.media_start_s = k
log.info(
"plex hls session start key=%s req=%.1f real=%.3f audio=%s",
key, start_s, s.media_start_s, audio_ord,
"plex hls session start key=%s req=%.1f real=%.3f multi=%s audio=%s",
key, start_s, s.media_start_s, is_multi, audio_ord,
)
return s

View file

@ -1499,6 +1499,7 @@ def stream_session(
start: float = 0.0,
audio: int | None = None,
aoff: float = 0.0,
multi: bool = False,
user: User = Depends(current_user),
db: Session = Depends(get_db),
) -> dict:
@ -1510,11 +1511,13 @@ def stream_session(
it = _item_or_404(db, rating_key)
if it.playable == "transcode":
raise HTTPException(status_code=501, detail="This format needs transcoding (a later phase).")
if it.playable == "direct" and audio is None and not aoff:
# `multi` (item has ≥2 audio tracks) forces the HLS path so ALL audio tracks ship as renditions and
# the client can switch audio without a restart — even for otherwise direct-playable files.
if it.playable == "direct" and audio is None and not aoff and not multi:
return {"mode": "direct", "url": f"/api/plex/stream/{it.rating_key}/file", "start_s": 0.0}
if plex_paths.local_media_path(db, it.file_path) is None:
raise HTTPException(status_code=404, detail="Media file not found on disk")
s = plex_stream.start_session(db, it, start, audio, aoff)
s = plex_stream.start_session(db, it, start, audio, aoff, multi)
if s is None:
raise HTTPException(status_code=404, detail="Media file not found on disk")
# Report the REAL media start (the keyframe ffmpeg landed on, which hls.js zero-bases to), not the
@ -1562,11 +1565,15 @@ def stream_hls_file(
raise HTTPException(status_code=404, detail="Not found")
s = plex_stream.current_session(str(rating_key))
if s is None:
# No session yet: only the entry playlist for a remux item may auto-start one at 0.
# No session yet (e.g. reaped): the ENTRY playlist may auto-start one at 0 — master.m3u8 for a
# multi-audio item, index.m3u8 for a single-audio remux item.
it = _item_or_404(db, rating_key)
if filename != "index.m3u8" or it.playable != "remux":
if filename == "master.m3u8":
s = plex_stream.start_session(db, it, 0.0, multi=True)
elif filename == "index.m3u8" and it.playable == "remux":
s = plex_stream.start_session(db, it, 0.0)
else:
raise HTTPException(status_code=404, detail="No active playback session")
s = plex_stream.start_session(db, it, 0.0)
if s is None:
raise HTTPException(status_code=404, detail="Media file not found on disk")
path = s.dir / filename

View file

@ -133,6 +133,10 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
const wrapRef = useRef<HTMLDivElement>(null);
const sessionStartRef = useRef(0); // absolute offset the current HLS/direct session begins at
const durationRef = useRef(0);
// ≥2 audio tracks → multi-rendition HLS: all tracks ship in one session and hls.js switches audio
// client-side (no restart). Ref so the stable loadSession/changeAudio callbacks read it fresh.
const multiAudioRef = useRef(false);
multiAudioRef.current = (detail?.audio_streams.length ?? 0) > 1;
const [playing, setPlaying] = useState(false);
const [abs, setAbs] = useState(0); // absolute current time
@ -213,7 +217,10 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
setLoadError(null);
let sess;
try {
sess = await api.plexSession(id, startAt, audioRef.current, aoffRef.current);
// Multi-audio → request the multi-rendition session (no single `audio` ordinal); the track is
// selected client-side after the manifest loads.
const m = multiAudioRef.current;
sess = await api.plexSession(id, startAt, m ? null : audioRef.current, aoffRef.current, m);
} catch (e: any) {
// Surface WHY playback can't start instead of an eternal spinner. 404 = the physical file
// isn't reachable (missing media bind-mount or a wrong plex_path_map); 501 = the codec
@ -247,6 +254,21 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
setReady(true);
if (aliveRef.current) video.play().catch(() => {});
});
// Restore the selected audio track ONCE hls.js has parsed the renditions — doing it on
// MANIFEST_PARSED was too early (the track list wasn't populated, so the set was dropped and
// playback stayed on the default track while the UI showed the restored selection).
hls.on(Hls.Events.AUDIO_TRACKS_UPDATED, () => {
if (multiAudioRef.current) {
const want = audioRef.current ?? 0;
if (hls.audioTrack !== want) {
try {
hls.audioTrack = want;
} catch {
/* out of range — ignore */
}
}
}
});
// A fatal HLS error (e.g. the remux session died / the file became unreadable) would
// otherwise leave the spinner up forever — surface it instead.
hls.on(Hls.Events.ERROR, (_e, data) => {
@ -353,8 +375,17 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
api.plexProgress(id, Math.floor(absRef.current), durationRef.current).catch(() => {});
};
const iv = window.setInterval(flush, 10000);
// Save on page unload too (F5/close/navigate): React effect cleanup does NOT run on a full reload,
// so without this the resume point lags up to 10s — e.g. a seek right before F5 was lost. Uses a
// keepalive beacon so the POST survives the unload.
const onHide = () => {
if (absRef.current > 0 && durationRef.current > 0)
api.plexProgressBeacon(id, Math.floor(absRef.current), durationRef.current);
};
window.addEventListener("pagehide", onHide);
return () => {
window.clearInterval(iv);
window.removeEventListener("pagehide", onHide);
flush();
qc.invalidateQueries({ queryKey: ["plex-browse"] });
qc.invalidateQueries({ queryKey: ["plex-show"] });
@ -368,6 +399,7 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
if (!video) return;
const dur = durationRef.current || target;
const T = Math.max(0, Math.min(target, dur - 1));
absRef.current = T; // reflect the target immediately so a save/beacon right after a seek is accurate
const rel = T - sessionStartRef.current;
if (rel >= 0 && rel <= (video.seekable.length ? video.seekable.end(0) : 0) + 0.5) {
video.currentTime = rel;
@ -461,7 +493,16 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
setAudioOrd(ord);
// Remember the LANGUAGE (not the ord) so F5 / the next item restores this choice. Menu stays open.
patchPrefs({ audioLang: ord == null ? "" : (detail?.audio_streams.find((s) => s.ord === ord)?.language ?? "") });
loadSession(absRef.current);
if (multiAudioRef.current && hlsRef.current) {
// Multi-rendition: switch the audio track CLIENT-SIDE — instant, same timeline, no restart.
try {
hlsRef.current.audioTrack = ord ?? 0;
} catch {
/* ignore */
}
} else {
loadSession(absRef.current); // single-audio: re-map the track server-side (session restart)
}
},
[loadSession, detail, patchPrefs],
);

View file

@ -1218,10 +1218,11 @@ export const api = {
req(`/api/plex/show/${encodeURIComponent(id)}`),
plexItem: (id: string): Promise<PlexItemDetail> =>
req(`/api/plex/item/${encodeURIComponent(id)}`),
plexSession: (id: string, start = 0, audio?: number | null, aoff = 0): Promise<PlexPlaySession> => {
plexSession: (id: string, start = 0, audio?: number | null, aoff = 0, multi = false): Promise<PlexPlaySession> => {
const p = new URLSearchParams({ start: String(Math.max(0, Math.floor(start))) });
if (audio != null) p.set("audio", String(audio));
if (aoff) p.set("aoff", String(aoff));
if (multi) p.set("multi", "1"); // ≥2 audio tracks → multi-rendition HLS (client-side audio switch)
return req(`/api/plex/stream/${encodeURIComponent(id)}/session?${p.toString()}`, { method: "POST" });
},
// WebVTT URL for a text subtitle track (used directly as a <track src>). Same-origin → cookie-authed.
@ -1234,6 +1235,27 @@ export const api = {
method: "POST",
body: JSON.stringify({ position_seconds, duration_seconds }),
}),
// Fire-and-forget progress save that SURVIVES page unload (F5/close/navigate): a normal fetch is
// cancelled on unload and React effect cleanup doesn't run on a full reload, so the last position
// (esp. right after a seek) would be lost. `keepalive` lets the POST complete during unload; we
// replicate req()'s credentials + account header (no retry — there's no page left to retry on).
plexProgressBeacon: (id: string, position_seconds: number, duration_seconds: number): void => {
const active = getActiveAccount();
try {
void fetch(`/api/plex/item/${encodeURIComponent(id)}/progress`, {
method: "POST",
keepalive: true,
credentials: "include",
headers: {
"Content-Type": "application/json",
...(active != null ? { [ACTIVE_ACCOUNT_HEADER]: String(active) } : {}),
},
body: JSON.stringify({ position_seconds, duration_seconds }),
}).catch(() => {});
} catch {
/* ignore */
}
},
plexSetState: (id: string, status: "new" | "watched" | "hidden"): Promise<unknown> =>
req(`/api/plex/item/${encodeURIComponent(id)}/state`, {
method: "POST",