merge main (0.78.2) into playback-class before release

# Conflicts:
#	backend/app/plex/stream.py

backend/app/config.py

@@ -265,7 +265,7 @@ class Settings(BaseSettings):
     # --- Transcode (playable="transcode") re-encode tuning (E6 S1) ---
     # Hardware-encoder policy: "auto" = VAAPI (h264_vaapi) when the iGPU render node is present AND the
     # source codec is hardware-decodable, else libx264 software; "on" = force VAAPI wherever the render
-    # node exists (skip the decodability gate); "off" = always software. See plex/stream.py `_want_hw`.
+    # node exists (skip the decodability gate); "off" = always software. See plex/stream.py `_select_encode`.
     plex_transcode_hw: str = "auto"
     # Software-fallback libx264 tuning (IGNORED on the VAAPI path). `preset` trades CPU for size; `crf`
     # trades quality for size (lower = better + bigger). `max_height` caps the SOFTWARE transcode's output

backend/app/downloads/edit.py

@@ -447,9 +447,9 @@ def run_ffprobe_json(
 ) -> dict | None:
     """One ffprobe call → its parsed JSON dict, or None on any failure (timeout, non-zero exit,
     missing binary, unparseable output). The SINGLE ffprobe-JSON-invocation point in the downloads
-    package: the browser-playability codec probe here and the upload-ingest probe (`upload._probe`)
-    both go through it, so the subprocess flags/timeout/error handling live in one place instead of
-    two drifting copies."""
+    package: the browser-playability codec probe here and the shared media probe (`probe.probe_media`,
+    used by both the download worker's finalize and the upload ingest) both go through it, so the
+    subprocess flags/timeout/error handling live in one place instead of drifting copies."""
     args = ["ffprobe", "-v", "error"]
     if show_streams:
         args.append("-show_streams")

backend/app/downloads/pipeline.py

@@ -15,7 +15,7 @@ from app.config import settings
 from app.db import SessionLocal
 from app.downloads import edit as editmod
 from app.downloads import errors as dlerrors
-from app.downloads import adapters, credentials, formats, mediakind, service, storage
+from app.downloads import adapters, credentials, formats, mediakind, probe, service, storage
 from app.downloads.jobdb import (
     _edit_staging_dir,
     _job_status,
@@ -475,15 +475,14 @@ def _download(
                 asset.rel_path = rel
                 asset.size_bytes = size
                 asset.duration_s = meta.duration_s
-                asset.width = int(info["width"]) if info.get("width") else None
-                asset.height = int(info["height"]) if info.get("height") else None
+                # Codecs/dims come from a PROBE of the finished file — the ground truth — NOT yt-dlp's
+                # `info` dict, whose codec fields are absent for some sources (Facebook among them):
+                # deriving media_kind from those nulls misclassified a real H.264 video as a
+                # download-only 'file' with no in-app play/edit. Same probe the upload ingest uses.
+                asset.vcodec, asset.acodec, asset.width, asset.height = probe.resolve_media_facts(
+                    final, info
+                )
                 asset.container = ext
-                asset.vcodec = (
-                    info.get("vcodec") if info.get("vcodec") not in (None, "none") else None
-                )
-                asset.acodec = (
-                    info.get("acodec") if info.get("acodec") not in (None, "none") else None
-                )
                 asset.media_kind = mediakind.derive(asset.vcodec, asset.acodec)
                 asset.title = meta.title
                 asset.uploader = meta.uploader

backend/app/downloads/probe.py

@@ -0,0 +1,107 @@
+"""Ground-truth media probe for a produced library file — the SINGLE place both ingest paths ask
+"what is actually on disk?".
+
+`media_kind` (and the stored `vcodec`/`acodec`/`width`/`height`) must reflect the real file, not an
+extractor's metadata: yt-dlp's `info` dict omits codec fields for some sources (e.g. Facebook), and
+deriving the kind from those nulls misclassified a genuine H.264 video as a download-only `file`
+(no in-app play). So the download worker (`pipeline`) and the local upload ingest (`upload`) both
+probe the finished file here; `mediakind.derive` then turns the probed codecs into the kind.
+"""
+
+from pathlib import Path
+
+from app.downloads import edit as editmod
+
+# ffprobe container formats that mean "this is a still image / image sequence", NOT playable media.
+# A standalone jpg/png/gif/webp/bmp reports a real video stream (codec mjpeg/png/…) with no
+# `attached_pic` flag, so without this guard it would classify as media_kind='video' and get a Play
+# button + a broken in-app <video> player (and inline serving on the public share). Keyed on the
+# container format so an animated gif/webp also lands as a generic 'file' (no browser plays those in
+# a <video> either). SVG and anything else ffprobe can't decode already fails the probe → 'file'.
+_IMAGE_FORMATS = {
+    "image2",
+    "image2pipe",
+    "png_pipe",
+    "mjpeg",
+    "mjpeg_pipe",
+    "jpeg_pipe",
+    "gif",
+    "webp_pipe",
+    "bmp_pipe",
+    "tiff_pipe",
+    "ico",
+    "svg_pipe",
+}
+
+
+def _facts_from_probe(
+    data: dict,
+) -> tuple[str | None, str | None, int | None, int | None, int | None]:
+    """(vcodec, acodec, width, height, duration_s) parsed from a NON-None ffprobe JSON dict. A still
+    image / image-sequence container yields no vcodec (→ generic 'file'); a video stream flagged
+    `attached_pic` (embedded cover art in an mp3/m4a) is skipped — it's a thumbnail, not a real video
+    track, so an audio file with album art still classifies as audio."""
+    fmt = data.get("format") or {}
+    fmt_names = {n.strip() for n in (fmt.get("format_name") or "").split(",")}
+    if fmt_names & _IMAGE_FORMATS:
+        return None, None, None, None, None
+
+    vcodec = acodec = None
+    width = height = None
+    for stream in data.get("streams", []):
+        kind = stream.get("codec_type")
+        if kind == "video" and vcodec is None:
+            if (stream.get("disposition") or {}).get("attached_pic"):
+                continue  # cover art, not a video track
+            vcodec = stream.get("codec_name")
+            width = stream.get("width")
+            height = stream.get("height")
+        elif kind == "audio" and acodec is None:
+            acodec = stream.get("codec_name")
+
+    duration = None
+    try:
+        raw = fmt.get("duration")
+        if raw is not None:
+            duration = int(float(raw))
+    except (TypeError, ValueError):
+        duration = None
+    return vcodec, acodec, width, height, duration
+
+
+def probe_media(path: Path) -> tuple[str | None, str | None, int | None, int | None, int | None]:
+    """(vcodec, acodec, width, height, duration_s) from one ffprobe call; all None if ffprobe fails
+    or the file isn't playable media (an unprobeable file — or a still image — is treated as a generic
+    non-media file, so it becomes a download-only 'file' rather than a broken 'video')."""
+    data = editmod.run_ffprobe_json(path, show_streams=True, show_format=True)
+    if data is None:
+        return None, None, None, None, None
+    return _facts_from_probe(data)
+
+
+def resolve_media_facts(
+    path: Path, info: dict
+) -> tuple[str | None, str | None, int | None, int | None]:
+    """(vcodec, acodec, width, height) for the stored download asset: PROBE the finished file on disk
+    (the ground truth). yt-dlp's `info` omits codec/dimension fields for some sources (Facebook among
+    them), so deriving media_kind from it there labelled a real H.264 video a download-only 'file'.
+
+    Precedence is by whether the probe RAN, not per-field: if ffprobe read the file, its codec verdict
+    is authoritative — INCLUDING a deliberate None (cover-art-only audio, a still image). A per-field
+    `probe or info` fallback would let `info`'s stale vcodec resurrect a video track the file doesn't
+    have (yt-dlp reports a progressive `vcodec` for an audio-mode download whose video it then strips),
+    mislabelling an audio download as a broken in-app 'video'. Only when ffprobe could NOT read the
+    file at all do we fall back to `info` (the pre-probe behaviour, a safety net); the cosmetic
+    dimensions still gap-fill from `info` in either case."""
+    data = editmod.run_ffprobe_json(path, show_streams=True, show_format=True)
+    info_v = info.get("vcodec") if info.get("vcodec") not in (None, "none") else None
+    info_a = info.get("acodec") if info.get("acodec") not in (None, "none") else None
+    info_w = int(info["width"]) if info.get("width") else None
+    info_h = int(info["height"]) if info.get("height") else None
+    if data is None:
+        # ffprobe could not read the file — fall back to the extractor's info entirely.
+        return info_v, info_a, info_w, info_h
+    pv, pa, pw, ph, _dur = _facts_from_probe(data)
+    # Probe ran → its codecs win outright (a None means "no such track", not "unknown"); info only
+    # gap-fills the cosmetic width/height the probe couldn't read.
+    return pv, pa, pw if pw is not None else info_w, ph if ph is not None else info_h

backend/app/downloads/upload.py

@@ -24,70 +24,9 @@ from sqlalchemy.orm import Session
 
 from app.config import settings
 from app.downloads import edit as editmod
-from app.downloads import mediakind, quota, service, storage
+from app.downloads import mediakind, probe, quota, service, storage
 from app.models import DownloadJob, MediaAsset
 
-# ffprobe container formats that mean "this is a still image / image sequence", NOT playable media.
-# A standalone jpg/png/gif/webp/bmp reports a real video stream (codec mjpeg/png/…) with no
-# `attached_pic` flag, so without this guard it would classify as media_kind='video' and get a Play
-# button + a broken in-app <video> player (and inline serving on the public share). Keyed on the
-# container format so an animated gif/webp also lands as a generic 'file' (no browser plays those in
-# a <video> either). SVG and anything else ffprobe can't decode already fails the probe → 'file'.
-_IMAGE_FORMATS = {
-    "image2",
-    "image2pipe",
-    "png_pipe",
-    "mjpeg",
-    "mjpeg_pipe",
-    "jpeg_pipe",
-    "gif",
-    "webp_pipe",
-    "bmp_pipe",
-    "tiff_pipe",
-    "ico",
-    "svg_pipe",
-}
-
-
-def _probe(path: Path) -> tuple[str | None, str | None, int | None, int | None, int | None]:
-    """(vcodec, acodec, width, height, duration_s) from one ffprobe call; all None if ffprobe fails
-    or the file isn't playable media (an unprobeable file — or a still image — is treated as a generic
-    non-media upload, so it becomes a download-only 'file' rather than broken 'video').
-
-    A video stream flagged `attached_pic` (embedded cover art in an mp3/m4a) is skipped — it's a
-    thumbnail, not a real video track, so an audio file with album art still classifies as audio."""
-    data = editmod.run_ffprobe_json(path, show_streams=True, show_format=True)
-    if data is None:
-        return None, None, None, None, None
-
-    fmt = data.get("format") or {}
-    fmt_names = {n.strip() for n in (fmt.get("format_name") or "").split(",")}
-    if fmt_names & _IMAGE_FORMATS:
-        # A still image / image sequence is a generic file, not video — no vcodec means media_kind='file'.
-        return None, None, None, None, None
-
-    vcodec = acodec = None
-    width = height = None
-    for stream in data.get("streams", []):
-        kind = stream.get("codec_type")
-        if kind == "video" and vcodec is None:
-            if (stream.get("disposition") or {}).get("attached_pic"):
-                continue  # cover art, not a video track
-            vcodec = stream.get("codec_name")
-            width = stream.get("width")
-            height = stream.get("height")
-        elif kind == "audio" and acodec is None:
-            acodec = stream.get("codec_name")
-
-    duration = None
-    try:
-        raw = fmt.get("duration")
-        if raw is not None:
-            duration = int(float(raw))
-    except (TypeError, ValueError):
-        duration = None
-    return vcodec, acodec, width, height, duration
-
 
 def _title_from_filename(filename: str) -> str:
     """A human display title from the original upload name: the stem with separators tidied to
@@ -112,7 +51,7 @@ def ingest_upload(
     rejected upload leaves nothing behind and the caller just cleans up its temp file."""
     quota.check_upload(db, user_id, size)  # the file's size is known → charge it up front
 
-    vcodec, acodec, width, height, duration = _probe(src_path)
+    vcodec, acodec, width, height, duration = probe.probe_media(src_path)
     kind = mediakind.derive(vcodec, acodec)
 
     # An upload never dedups (each file is its own asset): a random ref makes the cache key unique,

backend/app/plex/stream.py

@@ -97,12 +97,14 @@ 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).
@@ -144,7 +146,7 @@ _HLS_TAIL = [
 
 # --- Transcode (playable="transcode") video re-encode -----------------------------------------
 # The remux path stream-copies the video (`-c:v copy`); a non-h264 source can't be copied and must be
-# re-encoded to H.264. Two encoders, chosen per-session by `_want_hw`:
+# re-encoded to H.264. The encoder + tonemap are chosen per-session by `_select_encode`:
 #   * VAAPI hardware (`h264_vaapi`) — the prod default once the iGPU is passed in. A full GPU pipeline
 #     (decode + scale + encode all on-device) offloads the CPU almost entirely, so several concurrent
 #     1080p transcodes are fine and the API stays responsive.
@@ -154,25 +156,125 @@ _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"
-# Codecs the prod iGPU can hardware-DECODE (the precondition for the full-GPU VAAPI pipeline; a source
-# outside it — AV1, MPEG-4 ASP/XviD… — falls back to software even with the render node present). THE
-# single source is plex.sync, so the runtime encoder choice here and the catalog `transcode_hw` facet
-# can't disagree by codec; aliased to the old name to keep the call site below unchanged.
+# 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 it (AV1, MPEG-4 ASP/XviD…) falls back to software even with
+# the render node present, because the GPU can't decode it to feed its own encoder. THE single source is
+# plex.sync, so the runtime encoder choice here and the catalog `transcode_hw` facet can't disagree by
+# codec; aliased to the old name to keep the call sites below unchanged.
 _HW_DECODABLE = HW_DECODABLE_CODECS
+# Color-transfer values that mark a true-HDR source (PQ / HLG). A HW-decodable HDR 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. A 10-bit source
+# WITHOUT one of these transfers is SDR, not HDR, and must not be tone-mapped. See _probe_video_kind.
+_HDR_TRANSFERS = {"smpte2084", "arib-std-b67"}
 
 
-def _hwaccel_in_opts(hw: bool) -> list[str]:
+# The three source classes that drive the transcode-pipeline choice (see _probe_video_kind).
+_VK_HWSAFE = "hwsafe"  # 8-bit SDR — the plain scale_vaapi=format=nv12 GPU path
+_VK_HDR = "hdr"  # PQ/HLG HDR — needs the OpenCL tonemap graph (or software if no runtime)
+_VK_SDR10 = "sdr10"  # 10-bit *SDR* — a plain 10→8-bit downconvert, NEVER the HDR tone curve
+
+
+def _probe_video_kind(src: Path) -> str:
+    """Classify the source for the transcode pipeline — one of `_VK_HWSAFE` / `_VK_HDR` / `_VK_SDR10`:
+
+      * `_VK_HWSAFE` — 8-bit SDR: safe for the plain VAAPI path (`scale_vaapi=format=nv12` + h264_vaapi).
+      * `_VK_HDR`    — PQ/HLG HDR (any depth): the plain scale path breaks on Gen 9.5 VPP, so it needs the
+        OpenCL tonemap graph (`tonemap_opencl=hable`) when the runtime is present, else software libx264.
+      * `_VK_SDR10`  — 10-bit but *SDR* (bt709 etc.): NOT HDR. Applying the HDR hable tone curve here would
+        visibly alter an SDR picture (crushed highlights/midtones) where only a 10→8-bit downconvert was
+        needed, so it must NOT take the tonemap graph — it goes to software libx264, which CPU-decodes and
+        downconverts to 8-bit yuv420p exactly. (Kept off the plain scale_vaapi 10-bit path too: that path
+        is unverified on Gen 9.5 VPP, and software is the known-correct downconvert.)
+
+    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 `_VK_HWSAFE` on any probe failure: the runtime HW→SW fallback
+    in start_session backstops a HW session that still produces nothing, so a probe miss only costs the
+    rare wasted first attempt, never a stranded item."""
+    try:
+        out = subprocess.run(
+            [
+                "ffprobe",
+                "-v",
+                "error",
+                "-select_streams",
+                "v:0",
+                "-show_entries",
+                "stream=pix_fmt,bits_per_raw_sample,color_transfer",
+                "-of",
+                "default=noprint_wrappers=1",
+                str(src),
+            ],
+            capture_output=True,
+            text=True,
+            timeout=10,
+        )
+    except (subprocess.SubprocessError, OSError):
+        return _VK_HWSAFE
+    pix_fmt = bits = transfer = ""
+    for line in out.stdout.splitlines():
+        k, _, v = line.strip().partition("=")
+        v = v.strip().lower()
+        if k == "pix_fmt":
+            pix_fmt = v
+        elif k == "bits_per_raw_sample":
+            bits = v
+        elif k == "color_transfer":
+            transfer = v
+    if transfer in _HDR_TRANSFERS:
+        return _VK_HDR  # PQ/HLG — needs a real HDR→SDR tonemap, not a plain downconvert
+    # >8-bit but not HDR = 10-bit SDR. bits_per_raw_sample OR the pix_fmt bit-depth suffix
+    # (yuv420p10le, p010le, yuv444p12le…) signals >8-bit even when the other is absent.
+    if (bits.isdigit() and int(bits) > 8) or re.search(r"(?:10|12|16)(?:le|be)$", pix_fmt):
+        return _VK_SDR10
+    return _VK_HWSAFE
+
+
+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,
@@ -185,6 +287,23 @@ 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.
@@ -222,20 +341,53 @@ def _render_node_present() -> bool:
     return Path(_VAAPI_DEVICE).exists()
 
 
-def _want_hw(db: DbSession, item: PlexItem, transcode: bool) -> 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-decodability gate); "auto" (default) → VAAPI only when the
-    render node is present AND the iGPU can hardware-decode the source (else the full-GPU pipeline can't
-    feed itself → software libx264)."""
+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 true-HDR (PQ/HLG) source, when
+        the OpenCL runtime is present (Gen 9.5 VPP can't tonemap natively).
+
+    A 10-bit *SDR* source is deliberately NOT a `(True, True)` case: it needs a plain 10→8-bit downconvert,
+    not the HDR tone curve (which would visibly darken/flatten an SDR picture), so it drops to software.
+
+    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 routing a true-HDR source through the tonemap
+    graph 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 only for a true-HDR (PQ/HLG)
+    source that has an OpenCL runtime, while a 10-bit SDR 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
-    return (item.codec_video or "").lower() in _HW_DECODABLE
+        # Trust the admin: HW wherever the render node exists, skipping the decodability gate (the runtime
+        # backstop catches a miss). Only a true-HDR (PQ/HLG) source needs the tonemap graph — a 10-bit SDR
+        # source takes the plain scale path here, NOT the HDR tone curve. Probe only when OpenCL exists
+        # (short-circuit), so "on" stays probe-free where tonemap can't apply anyway.
+        tonemap = _opencl_present() and _probe_video_kind(src) == _VK_HDR
+        return (True, tonemap)
+    # auto
+    if (item.codec_video or "").lower() not in _HW_DECODABLE:
+        return (False, False)
+    kind = _probe_video_kind(src)
+    if kind == _VK_HWSAFE:
+        return (True, False)  # 8-bit SDR → full-GPU scale path
+    if kind == _VK_HDR:
+        # PQ/HLG: HW OpenCL tonemap when the runtime is present, else software libx264.
+        return (True, True) if _opencl_present() else (False, False)
+    # _VK_SDR10 — 10-bit SDR: a plain 10→8-bit downconvert, done exactly on the CPU. Never the HDR tone
+    # curve (it would alter an SDR picture), and off the unverified scale_vaapi 10-bit path — software.
+    return (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`). A readability shim for the tests that
+    assert only whether a session runs on the GPU (production decides via `_select_encode` directly)."""
+    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
@@ -290,15 +442,17 @@ 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
@@ -427,67 +581,104 @@ def start_session(
         return None
     key = item.rating_key
     start_s = max(0.0, float(start_s))
-    hw = _want_hw(db, item, transcode)
-    # 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
-    if transcode and not hw:
-        preset = sysconfig.get_str(db, "plex_transcode_preset") or "veryfast"
-        crf = sysconfig.get_int(db, "plex_transcode_crf")
-        max_height = sysconfig.get_int(db, "plex_transcode_max_height")
     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
-    # Nice the SOFTWARE transcode child (not the HW path, and not remux) so a CPU-bound re-encode can't
-    # starve the co-located API/playback. preexec_fn runs in the forked child (POSIX); on a platform
-    # without fork (the API always runs in a Linux container in prod) it's simply omitted.
-    preexec = _sw_child_preexec if (transcode and not hw and os.name == "posix") else None
-    with _lock:
-        old = _sessions.pop(key, None)
-        if old is not None:
-            _kill(old)
-        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,
-                audio_langs if is_multi else None,
-                transcode=transcode,
-                hw=hw,
-                preset=preset,
-                crf=crf,
-                max_height=max_height,
-            ),
-            stdout=subprocess.DEVNULL,
-            stderr=subprocess.DEVNULL,
-            preexec_fn=preexec,
-        )
-        s = HlsSession(
-            key,
-            directory,
-            proc,
-            start_s,
-            "master.m3u8" if is_multi else "index.m3u8",
-            transcode=transcode,
-            hw=hw,
-        )
-        _sessions[key] = s
-        # Admin-configurable concurrency cap (Configuration → Plex → Max concurrent transcodes);
-        # at least 1 so playback can't be capped to zero.
-        _enforce_cap(max(1, sysconfig.get_int(db, "plex_max_transcodes")))
+    tag = "multi" if is_multi else str(audio_ord)
+    directory = _HLS_ROOT / f"{key}_{int(start_s)}_{tag}_{aoff:+.2f}"
 
-    # Wait for the first video segment, then warm up a few more — all outside the lock (this blocks on
-    # ffmpeg producing them, and we must not hold up other sessions).
     def _seg(i: int) -> Path:
         # 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")
 
-    if wait_for(_seg(0), timeout=25.0):
+    def _spawn(
+        hw: bool, tonemap: bool = False, *, replace_only: "HlsSession | None" = None
+    ) -> "HlsSession | None":
+        # `replace_only` makes the supersede identity-aware for the runtime fallback: the whole
+        # check-and-replace happens under ONE _lock, so a newer same-key session (a concurrent
+        # seek-restart) installed between the fallback's decision and here is never clobbered. When the
+        # registered session is no longer `replace_only`, we abort and return None (the caller yields to
+        # whoever is registered now). The INITIAL spawn passes replace_only=None and always supersedes —
+        # that is the seek-restart's intended last-writer-wins.
+        # 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
+        if transcode and not hw:
+            preset = sysconfig.get_str(db, "plex_transcode_preset") or "veryfast"
+            crf = sysconfig.get_int(db, "plex_transcode_crf")
+            max_height = sysconfig.get_int(db, "plex_transcode_max_height")
+        # Nice the SOFTWARE transcode child (not the HW path, and not remux) so a CPU-bound re-encode
+        # can't starve the co-located API/playback. preexec_fn runs in the forked child (POSIX); on a
+        # platform without fork (the API always runs in a Linux container in prod) it's simply omitted.
+        preexec = _sw_child_preexec if (transcode and not hw and os.name == "posix") else None
+        with _lock:
+            if replace_only is not None and _sessions.get(key) is not replace_only:
+                return None  # superseded/culled while we waited out of lock — don't clobber the newer one
+            old = _sessions.pop(key, None)
+            if old is not None:
+                _kill(old)
+            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,
+                    audio_langs if is_multi else None,
+                    transcode=transcode,
+                    hw=hw,
+                    tonemap=tonemap,
+                    preset=preset,
+                    crf=crf,
+                    max_height=max_height,
+                ),
+                stdout=subprocess.DEVNULL,
+                stderr=subprocess.DEVNULL,
+                preexec_fn=preexec,
+            )
+            s = HlsSession(
+                key,
+                directory,
+                proc,
+                start_s,
+                "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);
+            # at least 1 so playback can't be capped to zero.
+            _enforce_cap(max(1, sysconfig.get_int(db, "plex_max_transcodes")))
+        return s
+
+    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:
+        # We waited on _seg(0) OUTSIDE _lock, so `s.proc` dying is ambiguous: it may have failed on its
+        # own (VAProfile/init/tonemap refusal), OR a concurrent same-key _spawn (a rapid seek-restart) or
+        # _enforce_cap may have killed it to install/keep a newer session. Respawn on libx264 (no HW, no
+        # tonemap) so a genuinely-failed GPU session never strands playback — but pass `replace_only=s` so
+        # the respawn is a no-op if we've been superseded: the identity check and the pop+replace share
+        # ONE lock in _spawn, so we can never tear down the newer session. 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
+        )
+        respawned = _spawn(False, False, replace_only=s)
+        if respawned is None:
+            # Superseded/culled mid-fallback — yield whoever is registered now (a newer seek-restart), or
+            # our own now-dead session if the key was cap-culled. Never None here: the POST /session route
+            # turns a None return into a "Media file not found on disk" 404 — dishonest for a file that
+            # exists. A (dead) session instead degrades to a retryable 404 ("No active playback session" /
+            # "Not available yet") the client re-requests.
+            return current_session(key) or s
+        s = respawned
+        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,
         # so the client's absolute clock + subtitle shift key off the true keyframe, not the requested
         # offset. On probe failure we keep the provisional requested offset (old behaviour, small drift).
@@ -499,17 +690,18 @@ def start_session(
         # at the first segment that isn't ready in time — ffmpeg produces them faster than realtime once
         # warm, so a miss means something slow and we proceed rather than block the request further.
         for i in range(1, _WARMUP_SEGMENTS):
-            if not wait_for(_seg(i), timeout=_WARMUP_SEG_TIMEOUT_S):
+            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,
         is_multi,
         audio_ord,
         transcode,
-        hw,
+        s.hw,
+        s.tonemap,
     )
     return s
 
@@ -522,10 +714,14 @@ def current_session(key: str) -> HlsSession | None:
         return s
 
 
-def wait_for(path: Path, timeout: float = 20.0) -> bool:
+def wait_for(path: Path, timeout: float = 20.0, proc: subprocess.Popen | None = None) -> bool:
     """Wait until a session file (playlist / segment) exists and is non-empty. Segments are
     produced ~faster than realtime, so a segment just ahead of playback appears quickly; a segment
-    far beyond the generated region won't (the frontend restarts the session at a seek instead)."""
+    far beyond the generated region won't (the frontend restarts the session at a seek instead).
+
+    With `proc` given, also stop as soon as that ffmpeg exits: a VAAPI session that fails to init
+    (10-bit/HDR VAProfile refusal) dies within ~1s producing nothing, and without this we'd block the
+    whole `timeout` before the HW→SW fallback could kick in — turning the fix into a stall."""
     end = time.time() + timeout
     while time.time() < end:
         try:
@@ -533,6 +729,12 @@ def wait_for(path: Path, timeout: float = 20.0) -> bool:
                 return True
         except OSError:
             pass
+        if proc is not None and proc.poll() is not None:
+            # the producer exited — no more segments are coming; one last check then give up.
+            try:
+                return path.exists() and path.stat().st_size > 0
+            except OSError:
+                return False
         time.sleep(0.15)
     return path.exists()
 

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,
     }