feat(plex): HW HDR→SDR tonemap for 10-bit/HDR transcode (OpenCL on the iGPU)

Upgrade mode-3's 10-bit/HDR branch from "route to software libx264" (BUG 1's
behaviour) to a full-GPU pipeline: VAAPI decode → OpenCL tonemap (hable,
HDR10/HLG → bt709 SDR nv12) → h264_vaapi. Gen 9.5 (UHD 630) can't tonemap in
the VAAPI VPP (tonemap_vaapi fails), so OpenCL is the only HW route — proven on
the prod iGPU at ~4.2x realtime, ~1 CPU core.

- stream.py: `_select_encode` returns (hw, tonemap); a 10-bit/HDR HW-decodable
  source picks the tonemap graph when the OpenCL runtime is present, else drops
  to software. `_opencl_present` gates on a registered ICD. `_venc_opts` /
  `_hwaccel_in_opts` gain the tonemap variant (named vaapi+opencl devices,
  `-filter_hw_device ocl`, hwmap→tonemap_opencl→hwmap reverse). `_want_hw` kept
  as the hw-half wrapper. The runtime HW→SW fallback still backstops a session
  that emits no segment.
- Dockerfile: add the Intel NEO OpenCL runtime (intel-opencl-icd). It was removed
  from Debian trixie, so it's pulled from bookworm (22.43 still supports Gen 9.5;
  the .deb runs on the trixie base — verified live). Harmless without /dev/dri.
- route + FE: thread `tonemap` through the session response; the stats overlay
  mode row reads "transcode · GPU · HDR→SDR" for the tonemap path.
- tests: tonemap command branch, the (hw, tonemap) selection matrix (auto/on ×
  8-bit/10-bit × opencl), start_session tonemap wiring, and the fmtMode tag.

Dockerfile

@@ -60,6 +60,19 @@ RUN apt-get update \
     && rm /tmp/deno.zip \
     && rm -rf /var/lib/apt/lists/*
 
+# Intel NEO OpenCL runtime (intel-opencl-icd) — needed by the Plex transcode's HDR/10-bit HW path, which
+# tone-maps HDR→SDR on the GPU via `tonemap_opencl` (Gen 9.5 / UHD 630 can't tonemap in the VAAPI VPP, so
+# OpenCL is the only HW route; without this the 10-bit branch drops to CPU libx264). NEO was REMOVED from
+# Debian trixie (this base = python:3.12-slim = trixie), so it's pulled from **bookworm** — the 22.43 there
+# still supports Gen 9.5 and its .deb runs on trixie (verified live on the prod UHD 630). The bookworm apt
+# source is added only for this one install and removed after, so it can't drift the rest of the image.
+# Harmless on hosts with no /dev/dri or no OpenCL use (the runtime just sits unused).
+RUN printf 'Types: deb\nURIs: http://deb.debian.org/debian\nSuites: bookworm\nComponents: main\nSigned-By: /usr/share/keyrings/debian-archive-keyring.pgp\n' > /etc/apt/sources.list.d/bookworm.sources \
+    && apt-get update \
+    && apt-get install -y --no-install-recommends -t bookworm intel-opencl-icd \
+    && rm /etc/apt/sources.list.d/bookworm.sources \
+    && rm -rf /var/lib/apt/lists/*
+
 # Install from the PINNED lock (requirements.lock, generated from requirements.txt) so the image is
 # byte-reproducible and the exact versions are the ones pip-audit scanned in CI — not whatever floats
 # up from the floor ranges on build day. Regenerate the lock when requirements.txt changes (see its header).

backend/app/plex/stream.py

@@ -96,12 +96,14 @@ class HlsSession:
         entry: str,
         transcode: bool = False,
         hw: bool = False,
+        tonemap: bool = False,
     ):
         self.key = key
         self.dir = directory
         self.proc = proc
         self.transcode = transcode  # video re-encoded (non-h264 source), not stream-copied
         self.hw = hw  # transcode ran on the VAAPI hardware encoder (else libx264 software)
+        self.tonemap = tonemap  # HW HDR/10-bit path: OpenCL tonemap between VAAPI decode and encode
         self.start_s = start_s  # the REQUESTED seek offset
         # The REAL media start: with `-ss X -c:v copy` ffmpeg can only cut at the nearest earlier
         # keyframe K<=X, and hls.js zero-bases the timeline to K (so video.currentTime=0 is K, not X).
@@ -153,21 +155,26 @@ _HLS_TAIL = [
 # Everything else — the `-copyts`/`_HLS_TAIL`/segment/seek-restart model and the AAC audio path — is
 # unchanged, so hls.js, subtitles, multi-audio and seeking keep working exactly as for remux.
 _VAAPI_DEVICE = "/dev/dri/renderD128"
+# Where an OpenCL ICD registers its driver. The Intel NEO runtime (intel-opencl-icd) drops an .icd file
+# here; its presence is the precondition for the HW HDR→SDR tonemap graph (see _opencl_present).
+_OPENCL_ICD_DIR = Path("/etc/OpenCL/vendors")
 # Codecs the prod iGPU (UHD 630 / Quick Sync Gen 9.5) can hardware-DECODE — the precondition for the
 # full-GPU VAAPI pipeline. A source outside this set (AV1, MPEG-4 ASP/XviD…) falls back to software even
 # when the render node is present, because the GPU can't decode it to feed its own encoder.
 _HW_DECODABLE = {"h264", "hevc", "h265", "vp9", "vc1", "mpeg2video"}
-# Color-transfer values that mark an HDR source (PQ / HLG). Even a HW-decodable HEVC that is HDR must
-# take the software path — see _probe_video_hwsafe.
+# Color-transfer values that mark an HDR source (PQ / HLG). A HW-decodable HDR/10-bit source can't take
+# the plain scale_vaapi=format=nv12 path (Gen 9.5 VPP refuses it) — it needs the OpenCL tonemap graph
+# (_venc_opts tonemap branch) when the OpenCL runtime is present, else software libx264. See _probe_video_hwsafe.
 _HDR_TRANSFERS = {"smpte2084", "arib-std-b67"}
 
 
 def _probe_video_hwsafe(src: Path) -> bool:
-    """Whether the source is safe for the VAAPI HW transcode pipeline — i.e. 8-bit SDR. A 10-bit (p010)
-    or HDR source breaks `scale_vaapi=format=nv12` on Gen 9.5 VPP (`VAProfile not supported` → ffmpeg
-    emits zero segments and the item never plays), so it must take the software libx264 path, which
-    CPU-decodes and downconverts to 8-bit yuv420p cleanly. Ground truth from ffprobe (bit depth via
-    pix_fmt / bits_per_raw_sample, HDR via color_transfer); the Plex mirror doesn't store these.
+    """Whether the source is safe for the PLAIN VAAPI transcode pipeline — i.e. 8-bit SDR. A 10-bit
+    (p010) or HDR source breaks `scale_vaapi=format=nv12` on Gen 9.5 VPP (`VAProfile not supported` →
+    ffmpeg emits zero segments and the item never plays), so it takes the OpenCL tonemap graph instead
+    (when the OpenCL runtime is present) or software libx264 (which CPU-decodes and downconverts to 8-bit
+    yuv420p cleanly). Ground truth from ffprobe (bit depth via pix_fmt / bits_per_raw_sample, HDR via
+    color_transfer); the Plex mirror doesn't store these.
 
     Returns True (assume HW-safe) on any probe failure: the runtime HW→SW fallback in start_session is
     the backstop for a HW session that still produces nothing, so a probe miss only costs the rare
@@ -213,18 +220,48 @@ def _probe_video_hwsafe(src: Path) -> bool:
     return True
 
 
-def _hwaccel_in_opts(hw: bool) -> list[str]:
+def _opencl_present() -> bool:
+    """Whether an OpenCL runtime (Intel NEO / intel-opencl-icd) is installed — the precondition for the
+    HW HDR→SDR tonemap graph. Gen 9.5 VPP can't tonemap natively (tonemap_vaapi fails), so the working
+    HW path for a 10-bit/HDR source is VAAPI-decode → OpenCL tonemap → VAAPI-encode, which needs an ICD
+    registered here. Absent on the dev box and until the image bakes NEO in → those sources take
+    software libx264. Its own function so it's monkeypatchable in tests, mirroring _render_node_present."""
+    try:
+        return any(_OPENCL_ICD_DIR.glob("*.icd"))
+    except OSError:
+        return False
+
+
+def _hwaccel_in_opts(hw: bool, tonemap: bool = False) -> list[str]:
     """VAAPI decode options, placed before `-i` so the source is decoded (and kept) on the GPU.
     `-hwaccel_output_format vaapi` keeps frames as GPU surfaces all the way to `h264_vaapi` — no
-    round-trip to system memory. Empty for the software path."""
+    round-trip to system memory. Empty for the software path.
+
+    The tonemap variant additionally derives an OpenCL device FROM the VAAPI one (`opencl=ocl@va`) so
+    tonemap_opencl can map the decoded surfaces zero-copy (see _venc_opts); it needs named devices, so it
+    uses `-init_hw_device` + `-hwaccel_device` rather than the plain path's implicit `-vaapi_device`."""
     if not hw:
         return []
+    if tonemap:
+        return [
+            "-init_hw_device",
+            f"vaapi=va:{_VAAPI_DEVICE}",
+            "-init_hw_device",
+            "opencl=ocl@va",
+            "-hwaccel",
+            "vaapi",
+            "-hwaccel_output_format",
+            "vaapi",
+            "-hwaccel_device",
+            "va",
+        ]
     return ["-hwaccel", "vaapi", "-hwaccel_output_format", "vaapi", "-vaapi_device", _VAAPI_DEVICE]
 
 
 def _venc_opts(
     transcode: bool,
     hw: bool,
+    tonemap: bool = False,
     preset: str = "veryfast",
     crf: int = 20,
     max_height: int = 0,
@@ -237,6 +274,23 @@ def _venc_opts(
     if not transcode:
         return ["-c:v", "copy"]
     force_kf = ["-force_key_frames", f"expr:gte(t,n_forced*{_SEG_SECONDS})"]
+    if hw and tonemap:
+        # HDR/10-bit HW path: Gen 9.5 VPP can't tonemap (scale_vaapi=format=nv12 fails), so the decoded
+        # VAAPI surfaces are mapped to OpenCL (hwmap=derive_device=opencl), tone-mapped HDR→SDR to 8-bit
+        # nv12 there (tonemap_opencl, hable curve, bt709 out), then mapped back to VAAPI (reverse=1) for
+        # h264_vaapi. `-filter_hw_device ocl` tells the filter graph which device to run the OpenCL
+        # filters on. Full GPU pipeline — ~1 CPU core, several×realtime on the UHD 630 (measured).
+        return [
+            "-filter_hw_device",
+            "ocl",
+            "-vf",
+            "hwmap=derive_device=opencl,"
+            "tonemap_opencl=tonemap=hable:transfer=bt709:matrix=bt709:primaries=bt709:format=nv12,"
+            "hwmap=derive_device=vaapi:reverse=1",
+            "-c:v",
+            "h264_vaapi",
+            *force_kf,
+        ]
     if hw:
         # scale_vaapi with format=nv12 downconverts a 10-bit (p010) HEVC source to 8-bit nv12 so
         # h264_vaapi (8-bit High) can encode it; for an 8-bit source it's a cheap format pass-through.
@@ -274,23 +328,44 @@ def _render_node_present() -> bool:
     return Path(_VAAPI_DEVICE).exists()
 
 
-def _want_hw(db: DbSession, item: PlexItem, transcode: bool, src: Path) -> bool:
-    """Whether to use the VAAPI hardware encoder for this transcode session, honouring the admin
-    `plex_transcode_hw` policy: "off" → always software; "on" → VAAPI wherever the render node exists
-    (trusting the admin, skipping the source gates — the runtime HW→SW fallback still backstops a
-    session that produces nothing); "auto" (default) → VAAPI only when the render node is present AND
-    the iGPU can hardware-decode the source AND the source is 8-bit SDR (a 10-bit/HDR source breaks
-    scale_vaapi=format=nv12 on Gen 9.5 → software libx264, which downconverts cleanly)."""
+def _select_encode(db: DbSession, item: PlexItem, transcode: bool, src: Path) -> tuple[bool, bool]:
+    """Pick the transcode video pipeline for this session — returns `(hw, tonemap)`:
+      * `(False, False)` → software libx264 (the fallback: no render node, non-decodable source, or a
+        10-bit/HDR source with no OpenCL runtime to tonemap on the GPU).
+      * `(True, False)`  → full-GPU VAAPI (`scale_vaapi=format=nv12` + `h264_vaapi`) — an 8-bit SDR source.
+      * `(True, True)`   → VAAPI decode → OpenCL tonemap → VAAPI encode — a 10-bit/HDR source, when the
+        OpenCL runtime is present (Gen 9.5 VPP can't tonemap natively).
+
+    Honours the admin `plex_transcode_hw` policy: "off" → always software; "on" → HW wherever the render
+    node exists (trusting the admin, skipping the decodability gate — the runtime HW→SW fallback still
+    backstops a session that produces nothing — but still selecting the tonemap graph for a 10-bit/HDR
+    source so it doesn't hit the scale_vaapi wall); "auto" (default) → HW only when the render node is
+    present AND the iGPU can hardware-decode the source, with tonemap chosen for a 10-bit/HDR source that
+    has an OpenCL runtime (else that source drops to software)."""
     if not transcode:
-        return False
+        return (False, False)
     mode = (sysconfig.get_str(db, "plex_transcode_hw") or "auto").lower()
     if mode == "off" or not _render_node_present():
-        return False
+        return (False, False)
     if mode == "on":
-        return True
+        # A 10-bit/HDR source still needs the tonemap graph (scale_vaapi can't handle it); use it when
+        # OpenCL is present, else plain HW and let the runtime backstop catch a miss. The probe is only
+        # consulted when OpenCL exists (short-circuit), so "on" stays probe-free where it can't matter.
+        tonemap = _opencl_present() and not _probe_video_hwsafe(src)
+        return (True, tonemap)
+    # auto
     if (item.codec_video or "").lower() not in _HW_DECODABLE:
-        return False
-    return _probe_video_hwsafe(src)
+        return (False, False)
+    if _probe_video_hwsafe(src):
+        return (True, False)  # 8-bit SDR → full-GPU scale path
+    # 10-bit / HDR, HW-decodable: HW OpenCL tonemap when the runtime is present, else software.
+    return (True, True) if _opencl_present() else (False, False)
+
+
+def _want_hw(db: DbSession, item: PlexItem, transcode: bool, src: Path) -> bool:
+    """The HW gate on its own (the `hw` half of `_select_encode`). Kept as a thin wrapper for the
+    callers/tests that only care whether a session runs on the GPU at all."""
+    return _select_encode(db, item, transcode, src)[0]
 
 
 # CPU-nice increment for the SOFTWARE transcode child. The Plex ffmpeg is a Popen in the long-lived API
@@ -345,15 +420,17 @@ def _ffmpeg_cmd(
     audio_langs: list[str] | None = None,
     transcode: bool = False,
     hw: bool = False,
+    tonemap: bool = False,
     preset: str = "veryfast",
     crf: int = 20,
     max_height: int = 0,
 ) -> list[str]:
     # For a transcode session the video is re-encoded (VAAPI or libx264) instead of stream-copied; the
     # hwaccel decode options go before -i, the encoder args replace `-c:v copy` below. Everything else is
-    # identical to remux.
-    venc = _venc_opts(transcode, hw, preset, crf, max_height)
-    args = ["ffmpeg", "-nostdin", "-loglevel", "error", *_hwaccel_in_opts(hw)]
+    # identical to remux. `tonemap` is the HDR/10-bit HW variant (OpenCL tonemap between VAAPI decode and
+    # encode) — it needs matching decode-side (named devices) and encode-side (the OpenCL filter graph) opts.
+    venc = _venc_opts(transcode, hw, tonemap, preset, crf, max_height)
+    args = ["ffmpeg", "-nostdin", "-loglevel", "error", *_hwaccel_in_opts(hw, tonemap)]
     # 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
@@ -491,7 +568,7 @@ def start_session(
         # multi-audio: the video rendition is variant v0 → seg_{i}_0.ts; single-audio → seg_{i}.ts
         return directory / (f"seg_{i}_0.ts" if is_multi else f"seg_{i}.ts")
 
-    def _spawn(hw: bool) -> HlsSession:
+    def _spawn(hw: bool, tonemap: bool = False) -> HlsSession:
         # Software-fallback encode tuning — read only on the SW transcode path (the VAAPI path and
         # remux ignore preset/crf/max_height, and remux must not touch the DB for them).
         preset, crf, max_height = "veryfast", 20, 0
@@ -519,6 +596,7 @@ def start_session(
                     audio_langs if is_multi else None,
                     transcode=transcode,
                     hw=hw,
+                    tonemap=tonemap,
                     preset=preset,
                     crf=crf,
                     max_height=max_height,
@@ -535,6 +613,7 @@ def start_session(
                 "master.m3u8" if is_multi else "index.m3u8",
                 transcode=transcode,
                 hw=hw,
+                tonemap=tonemap,
             )
             _sessions[key] = s
             # Admin-configurable concurrency cap (Configuration → Plex → Max concurrent transcodes);
@@ -542,20 +621,21 @@ def start_session(
             _enforce_cap(max(1, sysconfig.get_int(db, "plex_max_transcodes")))
         return s
 
-    s = _spawn(_want_hw(db, item, transcode, src))
+    s = _spawn(*_select_encode(db, item, transcode, src))
     # Wait for the first video segment outside the lock (this blocks on ffmpeg, and we must not hold up
     # other sessions). Pass the proc so a HW session that dies at VAAPI init fails fast (not after 25s).
     ready = wait_for(_seg(0), timeout=25.0, proc=s.proc)
     if not ready and s.hw:
-        # Runtime HW→SW fallback: the VAAPI session produced no segments (a VAProfile/init failure the
-        # _want_hw gates didn't catch — e.g. an admin-forced "on" on a 10-bit source, or an unforeseen
-        # driver refusal). Respawn on libx264 so a failed GPU session never strands playback; the
-        # software path CPU-decodes and downconverts every source. s.hw is now False, so the route
-        # honestly reports "CPU" and the player's Preparing-stream state stays correct.
+        # Runtime HW→SW fallback: the VAAPI session produced no segments (a VAProfile/init/tonemap
+        # failure the _select_encode gates didn't catch — e.g. an admin-forced "on" on a source the GPU
+        # can't decode, or an unforeseen driver refusal). Respawn on libx264 (no HW, no tonemap) so a
+        # failed GPU session never strands playback; the software path CPU-decodes and downconverts every
+        # source. s.hw is now False, so the route honestly reports "CPU" and the player's Preparing-stream
+        # state stays correct.
         log.warning(
             "plex transcode: HW session produced no segments — falling back to software key=%s", key
         )
-        s = _spawn(False)
+        s = _spawn(False, False)
         ready = wait_for(_seg(0), timeout=25.0, proc=s.proc)
     if ready:
         # For a seek (start_s>0) learn the REAL start offset K from the first segment's absolute PTS,
@@ -572,7 +652,7 @@ def start_session(
             if not wait_for(_seg(i), timeout=_WARMUP_SEG_TIMEOUT_S, proc=s.proc):
                 break
     log.info(
-        "plex hls session start key=%s req=%.1f real=%.3f multi=%s audio=%s transcode=%s hw=%s",
+        "plex hls session start key=%s req=%.1f real=%.3f multi=%s audio=%s transcode=%s hw=%s tonemap=%s",
         key,
         start_s,
         s.media_start_s,
@@ -580,6 +660,7 @@ def start_session(
         audio_ord,
         transcode,
         s.hw,
+        s.tonemap,
     )
     return s
 

backend/app/routes/plex/playback.py

@@ -554,14 +554,16 @@ def stream_session(
         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
     # requested offset — the client's absolute clock and the subtitle cue-shift both key off this.
-    # `transcode`/`hw` let the player show an honest cold-start ("Preparing stream…") and surface the
-    # active mode + encoder in the stats overlay; both segment streams are otherwise identical.
+    # `transcode`/`hw`/`tonemap` let the player show an honest cold-start ("Preparing stream…") and
+    # surface the active mode + encoder in the stats overlay (`tonemap` = the HW HDR→SDR path); both
+    # segment streams are otherwise identical.
     return {
         "mode": "hls",
         "start_s": s.media_start_s,
         "url": f"/api/plex/stream/{it.rating_key}/hls/{s.entry}",
         "transcode": s.transcode,
         "hw": s.hw,
+        "tonemap": s.tonemap,
     }
 
 

backend/tests/test_plex_playback_session.py

@@ -58,7 +58,7 @@ def _post(db, user, monkeypatch, rating_key, *, captured):
     def fake_start(db_, it, start, audio, aoff, multi, transcode=False):
         captured["transcode"] = transcode
         return SimpleNamespace(
-            media_start_s=start, entry="index.m3u8", transcode=transcode, hw=False
+            media_start_s=start, entry="index.m3u8", transcode=transcode, hw=False, tonemap=False
         )
 
     monkeypatch.setattr(plex_stream, "start_session", fake_start)

backend/tests/test_plex_stream.py

@@ -267,10 +267,36 @@ class TestTranscodeCommand:
         )
         assert cmd[cmd.index("-vf") + 1] == "scale_vaapi=format=nv12"  # not a software scale
 
+    def test_hdr_tonemap_uses_the_opencl_graph(self):
+        # The HDR/10-bit HW path: VAAPI decode → OpenCL tonemap → VAAPI encode. Gen 9.5 can't tonemap in
+        # the VAAPI VPP, so this is the only HW route for a 10-bit/HDR source.
+        cmd = stream._ffmpeg_cmd(
+            self.SRC, 0.0, self.OUT, None, transcode=True, hw=True, tonemap=True
+        )
+        assert _venc(cmd) == "h264_vaapi"
+        assert "libx264" not in cmd
+        # Named devices + an OpenCL device DERIVED from the VAAPI one, all before -i.
+        assert cmd[cmd.index("-init_hw_device") + 1] == f"vaapi=va:{stream._VAAPI_DEVICE}"
+        assert "opencl=ocl@va" in cmd
+        assert cmd[cmd.index("-hwaccel_device") + 1] == "va"
+        assert cmd.index("-init_hw_device") < cmd.index("-i")
+        # The filter graph runs on the derived OpenCL device and tone-maps HDR→SDR to 8-bit nv12.
+        assert cmd[cmd.index("-filter_hw_device") + 1] == "ocl"
+        vf = cmd[cmd.index("-vf") + 1]
+        assert "hwmap=derive_device=opencl" in vf
+        assert "tonemap_opencl=" in vf and "format=nv12" in vf
+        assert "hwmap=derive_device=vaapi:reverse=1" in vf
+        assert "scale_vaapi" not in vf  # the plain 8-bit path's filter must NOT be used here
+        assert (
+            cmd[cmd.index("-force_key_frames") + 1] == f"expr:gte(t,n_forced*{stream._SEG_SECONDS})"
+        )
+
 
 class TestWantHw:
-    """VAAPI is used only where the render node was passed in AND (in auto) the source is HW-decodable
-    AND 8-bit SDR, with an admin `plex_transcode_hw` policy (auto/on/off) on top."""
+    """VAAPI is used only where the render node was passed in AND (in auto) the source is HW-decodable,
+    with an admin `plex_transcode_hw` policy (auto/on/off) on top. A HW session on a 10-bit/HDR source
+    tone-maps on the GPU via OpenCL (tonemap=True) when the OpenCL runtime is present, else drops to
+    software — see the `_select_encode` cases below."""
 
     SRC = Path("/media/movie.mkv")
 
@@ -283,6 +309,11 @@ class TestWantHw:
         # Default the bit-depth probe so the auto-decodable tests don't shell out to ffprobe.
         monkeypatch.setattr(stream, "_probe_video_hwsafe", lambda src: safe)
 
+    @staticmethod
+    def _opencl(monkeypatch, present):
+        # Pin the OpenCL-runtime probe so the 10-bit routing doesn't depend on the CI box's /etc/OpenCL.
+        monkeypatch.setattr(stream, "_opencl_present", lambda: present)
+
     def test_no_hw_when_not_transcoding(self, monkeypatch):
         self._policy(monkeypatch, "auto")
         monkeypatch.setattr(stream, "_render_node_present", lambda: True)
@@ -322,15 +353,38 @@ class TestWantHw:
             is False
         )
 
-    def test_auto_falls_back_to_software_for_a_10bit_or_hdr_source(self, monkeypatch):
-        # BUG 1: a decodable HEVC that is 10-bit/HDR still breaks scale_vaapi=format=nv12 → software.
+    def test_auto_10bit_source_falls_back_to_software_without_opencl(self, monkeypatch):
+        # A decodable HEVC that is 10-bit/HDR can't use scale_vaapi=format=nv12 (Gen 9.5), and with no
+        # OpenCL runtime to tonemap on the GPU it drops to software libx264 (BUG 1's behaviour).
         self._policy(monkeypatch, "auto")
         self._hwsafe(monkeypatch, False)
+        self._opencl(monkeypatch, False)
         monkeypatch.setattr(stream, "_render_node_present", lambda: True)
-        assert (
-            stream._want_hw(None, _FakeItem(codec_video="hevc"), transcode=True, src=self.SRC)
-            is False
-        )
+        assert stream._select_encode(
+            None, _FakeItem(codec_video="hevc"), transcode=True, src=self.SRC
+        ) == (False, False)
+
+    def test_auto_10bit_source_uses_hw_tonemap_with_opencl(self, monkeypatch):
+        # With the OpenCL runtime present, the same 10-bit/HDR HEVC tone-maps on the GPU (VAAPI decode →
+        # OpenCL tonemap → VAAPI encode) instead of dropping to CPU — the HW HDR path.
+        self._policy(monkeypatch, "auto")
+        self._hwsafe(monkeypatch, False)
+        self._opencl(monkeypatch, True)
+        monkeypatch.setattr(stream, "_render_node_present", lambda: True)
+        assert stream._select_encode(
+            None, _FakeItem(codec_video="hevc"), transcode=True, src=self.SRC
+        ) == (True, True)
+
+    def test_auto_8bit_source_uses_plain_hw_not_tonemap(self, monkeypatch):
+        # An 8-bit SDR source takes the plain full-GPU scale path — never the tonemap graph — even where
+        # OpenCL is available.
+        self._policy(monkeypatch, "auto")
+        self._hwsafe(monkeypatch, True)
+        self._opencl(monkeypatch, True)
+        monkeypatch.setattr(stream, "_render_node_present", lambda: True)
+        assert stream._select_encode(
+            None, _FakeItem(codec_video="hevc"), transcode=True, src=self.SRC
+        ) == (True, False)
 
     def test_off_forces_software_even_for_a_decodable_source(self, monkeypatch):
         self._policy(monkeypatch, "off")
@@ -356,6 +410,17 @@ class TestWantHw:
             is False
         )
 
+    def test_on_still_picks_the_tonemap_graph_for_a_10bit_source(self, monkeypatch):
+        # "on" trusts the GPU but must still route a 10-bit/HDR source through the tonemap graph (plain
+        # scale_vaapi would hit the VPP wall) when OpenCL is present.
+        self._policy(monkeypatch, "on")
+        self._hwsafe(monkeypatch, False)
+        self._opencl(monkeypatch, True)
+        monkeypatch.setattr(stream, "_render_node_present", lambda: True)
+        assert stream._select_encode(
+            None, _FakeItem(codec_video="hevc"), transcode=True, src=self.SRC
+        ) == (True, True)
+
 
 class TestColdStartWarmup:
     """The cold-start buffer warm-up: before reporting a session ready, start_session waits for the
@@ -450,7 +515,9 @@ class TestTranscodeStartSession:
     def test_software_transcode_threads_the_knobs_into_the_command(
         self, capturing_env, monkeypatch
     ):
-        monkeypatch.setattr(stream, "_want_hw", lambda db, item, transcode, src: False)
+        monkeypatch.setattr(
+            stream, "_select_encode", lambda db, item, transcode, src: (False, False)
+        )
         s = stream.start_session(
             db=None, item=_FakeItem(codec_video="mpeg4"), start_s=0.0, transcode=True
         )
@@ -465,11 +532,13 @@ class TestTranscodeStartSession:
             assert kwargs.get("preexec_fn") is stream._sw_child_preexec
 
     def test_hardware_transcode_uses_the_gpu_and_never_nices(self, capturing_env, monkeypatch):
-        monkeypatch.setattr(stream, "_want_hw", lambda db, item, transcode, src: True)
+        monkeypatch.setattr(
+            stream, "_select_encode", lambda db, item, transcode, src: (True, False)
+        )
         s = stream.start_session(
             db=None, item=_FakeItem(codec_video="hevc"), start_s=0.0, transcode=True
         )
-        assert s is not None and s.transcode is True and s.hw is True
+        assert s is not None and s.transcode is True and s.hw is True and s.tonemap is False
         assert len(_CapturingProc.calls) == 1  # first segment arrived — no HW→SW fallback respawn
         (cmd, *_), kwargs = _CapturingProc.calls[0]
         assert _venc(cmd) == "h264_vaapi"
@@ -479,13 +548,31 @@ class TestTranscodeStartSession:
         )  # no SW cap
         assert kwargs.get("preexec_fn") is None  # GPU encode is cheap — never niced
 
+    def test_hdr_tonemap_transcode_threads_the_opencl_graph_and_flags_the_session(
+        self, capturing_env, monkeypatch
+    ):
+        # A HW tonemap session (10-bit/HDR + OpenCL) must build the OpenCL graph in the argv and carry
+        # tonemap=True on the session, still on the GPU (hw=True) and never niced.
+        monkeypatch.setattr(stream, "_select_encode", lambda db, item, transcode, src: (True, True))
+        s = stream.start_session(
+            db=None, item=_FakeItem(codec_video="hevc"), start_s=0.0, transcode=True
+        )
+        assert s is not None and s.transcode is True and s.hw is True and s.tonemap is True
+        (cmd, *_), kwargs = _CapturingProc.calls[0]
+        assert _venc(cmd) == "h264_vaapi"
+        assert "opencl=ocl@va" in cmd and cmd[cmd.index("-filter_hw_device") + 1] == "ocl"
+        assert "tonemap_opencl=" in cmd[cmd.index("-vf") + 1]
+        assert kwargs.get("preexec_fn") is None
+
     def test_a_hw_session_that_produces_nothing_falls_back_to_software(
         self, capturing_env, monkeypatch
     ):
         # BUG 1 backstop: _want_hw picked HW but the VAAPI session emits no first segment (a VAProfile
         # refusal the gates didn't catch). start_session must respawn on libx264 and report hw=False so
         # the item still plays and the route/player honestly show CPU.
-        monkeypatch.setattr(stream, "_want_hw", lambda db, item, transcode, src: True)
+        monkeypatch.setattr(
+            stream, "_select_encode", lambda db, item, transcode, src: (True, False)
+        )
         seen = {"n": 0}
 
         def flaky_wait(path, timeout=20.0, proc=None):
@@ -510,7 +597,9 @@ class TestTranscodeStartSession:
     def test_remux_reads_no_transcode_knobs_and_stream_copies(self, capturing_env, monkeypatch):
         # A remux start must stream-copy the video and spawn no niced child (the transcode knobs and
         # the preexec nice are gated behind `transcode=True`).
-        monkeypatch.setattr(stream, "_want_hw", lambda db, item, transcode, src: False)
+        monkeypatch.setattr(
+            stream, "_select_encode", lambda db, item, transcode, src: (False, False)
+        )
         s = stream.start_session(db=None, item=_FakeItem(), start_s=0.0)  # transcode defaults False
         assert s is not None and s.transcode is False
         (cmd, *_), kwargs = _CapturingProc.calls[0]

frontend/src/components/PlexPlayer.tsx

@@ -101,6 +101,9 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
   // Which encoder the current transcode session ran: true=VAAPI hardware, false=libx264 software,
   // null=not a transcode. Feeds the stats overlay's mode row ("transcode · GPU" / "· CPU").
   const hwRef = useRef<boolean | null>(null);
+  // Whether the current HW transcode tone-mapped a 10-bit/HDR source on the GPU (OpenCL). Feeds the
+  // stats mode row's "· HDR→SDR" tag; null/false for an 8-bit HW, software, or non-transcode session.
+  const tonemapRef = useRef<boolean | null>(null);
   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.
@@ -234,6 +237,7 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
       // re-encodes the video (hw = which encoder), and its first segment is genuinely slower than remux.
       const isTranscode = sess.mode === "hls" && !!sess.transcode;
       hwRef.current = isTranscode ? (sess.hw ?? false) : null;
+      tonemapRef.current = isTranscode ? (sess.tonemap ?? false) : null;
       setSessionIsTranscode(isTranscode);
       // hls.js plays our non-zero-start copyts stream ~1s BEHIND the media clock (see HLS_SUB_LAG);
       // fold that into the session start so the seekbar clock, seeking AND the subtitle shift are all
@@ -1206,6 +1210,7 @@ export default function PlexPlayer({ itemId, onClose, queue }: Props) {
           hlsRef={hlsRef}
           playable={detail?.playable ?? ""}
           hwRef={hwRef}
+          tonemapRef={tonemapRef}
           sessionStartRef={sessionStartRef}
           absRef={absRef}
           durationRef={durationRef}

frontend/src/components/PlexStatsOverlay.tsx

@@ -13,6 +13,7 @@ export function PlexStatsOverlay({
   hlsRef,
   playable,
   hwRef,
+  tonemapRef,
   sessionStartRef,
   absRef,
   durationRef,
@@ -22,6 +23,7 @@ export function PlexStatsOverlay({
   hlsRef: RefObject<Hls | null>;
   playable: string;
   hwRef: RefObject<boolean | null>; // active transcode encoder: true=GPU, false=CPU, null=not transcode
+  tonemapRef: RefObject<boolean | null>; // HW HDR→SDR tonemap engaged (adds a "· HDR→SDR" mode tag)
   sessionStartRef: RefObject<number>;
   absRef: RefObject<number>;
   durationRef: RefObject<number>;
@@ -78,6 +80,7 @@ export function PlexStatsOverlay({
         buildStatRows({
           playable,
           hwEncoder: hwRef.current,
+          tonemap: tonemapRef.current,
           transport: hls ? "hls" : "direct",
           videoWidth: v?.videoWidth ?? 0,
           videoHeight: v?.videoHeight ?? 0,
@@ -94,7 +97,7 @@ export function PlexStatsOverlay({
     sample();
     const iv = setInterval(sample, 1000);
     return () => clearInterval(iv);
-  }, [videoRef, hlsRef, playable, hwRef, sessionStartRef, absRef, durationRef]);
+  }, [videoRef, hlsRef, playable, hwRef, tonemapRef, sessionStartRef, absRef, durationRef]);
 
   return (
     <div

frontend/src/lib/api/plex.ts

@@ -392,6 +392,7 @@ export interface PlexPlaySession {
   start_s: number;
   transcode?: boolean; // hls only: video re-encoded (non-h264 source) vs stream-copied (remux)
   hw?: boolean; // hls transcode only: ran on the VAAPI hardware encoder (else libx264 software)
+  tonemap?: boolean; // hls HW transcode only: 10-bit/HDR source tone-mapped HDR→SDR on the GPU (OpenCL)
 }
 
 // Discover surface (E5 S4): the consolidated rail payload. Every rail is a plain PlexCard list the

frontend/src/lib/plexStats.test.ts

@@ -118,6 +118,11 @@ describe("buildStatRows", () => {
     expect(cpu.find((r) => r.key === "mode")?.value).toBe("transcode · CPU");
   });
 
+  it("tags the mode row with HDR→SDR for a HW tonemap session", () => {
+    const rows = buildStatRows({ ...base, playable: "transcode", hwEncoder: true, tonemap: true });
+    expect(rows.find((r) => r.key === "mode")?.value).toBe("transcode · GPU · HDR→SDR");
+  });
+
   it("shows an empty classification honestly as a dash", () => {
     expect(buildStatRows({ ...base, playable: "" }).find((r) => r.key === "mode")?.value).toBe(
       DASH,
@@ -133,5 +138,9 @@ describe("fmtMode", () => {
     expect(fmtMode("transcode")).toBe("transcode");
     expect(fmtMode("remux", true)).toBe("remux"); // not a transcode → no encoder suffix
     expect(fmtMode("")).toBe(DASH);
+    // The HDR→SDR tag rides on the GPU encoder token for a tonemap session.
+    expect(fmtMode("transcode", true, true)).toBe("transcode · GPU · HDR→SDR");
+    expect(fmtMode("transcode", true, false)).toBe("transcode · GPU");
+    expect(fmtMode("transcode", null, true)).toBe("transcode"); // no encoder → no tag either
   });
 });

frontend/src/lib/plexStats.ts

@@ -12,6 +12,9 @@ export interface PlexStatsInput {
   // software, null/undefined = not a transcode (or unknown) — so the mode row can read "transcode · GPU"
   // vs "transcode · CPU", the visible confirmation of which S1 path engaged.
   hwEncoder?: boolean | null;
+  // For a HW transcode of a 10-bit/HDR source: true = the GPU OpenCL tonemap path (HDR→SDR) engaged, so
+  // the mode row can read "transcode · GPU · HDR→SDR". Absent/false for an 8-bit HW or software session.
+  tonemap?: boolean | null;
   videoWidth: number; // intrinsic decoded dims (0 while the browser can't decode / pre-metadata)
   videoHeight: number;
   droppedFrames: number; // getVideoPlaybackQuality().droppedVideoFrames
@@ -94,20 +97,30 @@ export function fmtCodec(video: string | undefined, audio: string | undefined):
 }
 
 // Active mode, with the transcode encoder appended when known: "transcode · GPU" (VAAPI hardware) or
-// "transcode · CPU" (libx264 software) — the at-a-glance confirmation of which S1 path engaged. Non-
-// transcode modes (direct/remux) and an unknown encoder read as the bare mode. Raw tokens (not i18n'd),
-// consistent with the other technical values in this overlay (e.g. transport "hls").
-export function fmtMode(playable: string, hwEncoder?: boolean | null): string {
+// "transcode · CPU" (libx264 software) — the at-a-glance confirmation of which S1 path engaged — and a
+// "· HDR→SDR" tag when the GPU OpenCL tonemap path handled a 10-bit/HDR source ("transcode · GPU ·
+// HDR→SDR"). Non-transcode modes (direct/remux) and an unknown encoder read as the bare mode. Raw tokens
+// (not i18n'd), consistent with the other technical values in this overlay (e.g. transport "hls").
+export function fmtMode(
+  playable: string,
+  hwEncoder?: boolean | null,
+  tonemap?: boolean | null,
+): string {
   const mode = playable || DASH;
   if (playable !== "transcode" || hwEncoder == null) return mode;
-  return `${mode} · ${hwEncoder ? "GPU" : "CPU"}`;
+  const base = `${mode} · ${hwEncoder ? "GPU" : "CPU"}`;
+  return tonemap ? `${base} · HDR→SDR` : base;
 }
 
 // Build the ordered, i18n-keyed rows the overlay renders. Rows whose data is absent still show a dash
 // (honest "not available" beats a missing row) — except HLS-only rows, dropped entirely for direct play.
 export function buildStatRows(s: PlexStatsInput): StatRow[] {
   const rows: StatRow[] = [
-    { key: "mode", labelKey: "plex.player.stats.mode", value: fmtMode(s.playable, s.hwEncoder) },
+    {
+      key: "mode",
+      labelKey: "plex.player.stats.mode",
+      value: fmtMode(s.playable, s.hwEncoder, s.tonemap),
+    },
     { key: "transport", labelKey: "plex.player.stats.transport", value: s.transport },
     {
       key: "res",