ok trnscode swp

Dockerfile.govod

@@ -0,0 +1,29 @@
+FROM golang:1.23-bookworm AS builder
+
+ARG MEMORIES_COMMIT=951b79a38f5f6438c956b98d63f760fdb4cd6c50
+WORKDIR /src
+
+RUN git clone https://github.com/pulsejet/memories.git . \
+    && git checkout "$MEMORIES_COMMIT"
+
+WORKDIR /src/go-vod
+COPY patches/go-vod-sheepflix.patch /tmp/go-vod-sheepflix.patch
+RUN git apply /tmp/go-vod-sheepflix.patch \
+    && gofmt -w transcoder/handler.go transcoder/stream.go \
+    && CGO_ENABLED=0 go build -ldflags="-s -w" -o /go-vod .
+
+# go-vod itself uses Jellyfin's FFmpeg distribution. Pin the stable Jellyfin
+# runtime so driver/codec behavior does not change unexpectedly on rebuild.
+FROM jellyfin/jellyfin:10.11.11
+USER root
+
+RUN ln -sf /usr/lib/jellyfin-ffmpeg/ffmpeg /usr/local/bin/ffmpeg \
+    && ln -sf /usr/lib/jellyfin-ffmpeg/ffprobe /usr/local/bin/ffprobe
+
+COPY --from=builder /go-vod /app/go-vod
+COPY govod-config.json /app/config.json
+
+EXPOSE 8080
+ENV LIBVA_DRIVER_NAME=i965
+
+ENTRYPOINT ["/app/go-vod", "/app/config.json"]

Dockerfile.transcode

@@ -1,35 +0,0 @@
-FROM golang:1.23-alpine AS builder
-
-ARG GO_TRANSCODE_COMMIT=f8137e68c7b8da88c51aec9563eec9585c10c723
-WORKDIR /src
-
-RUN apk add --no-cache git
-RUN git clone https://github.com/m1k1o/go-transcode.git . \
-    && git checkout "$GO_TRANSCODE_COMMIT"
-
-COPY patches/go-transcode-vaapi.patch /tmp/go-transcode-vaapi.patch
-RUN git apply /tmp/go-transcode-vaapi.patch \
-    && gofmt -w hlsvod/manager.go hlsvod/probe.go hlsvod/transcode.go \
-    && CGO_ENABLED=0 go build -o /go-transcode .
-
-# Keep Alpine 3.17: upstream documents an FFmpeg 6 segmenting regression with
-# newer Alpine releases. Install both Intel and Mesa VAAPI drivers so this also
-# works if the Unraid host exposes an AMD render device.
-FROM alpine:3.17
-WORKDIR /app
-
-RUN apk add --no-cache \
-        bash \
-        ffmpeg \
-        intel-media-driver \
-        libva-intel-driver \
-        mesa-va-gallium
-
-COPY --from=builder /go-transcode /app/go-transcode
-
-EXPOSE 8080
-ENV TRANSCODE_BIND=:8080 \
-    TRANSCODE_VAAPI_DEVICE=/dev/dri/renderD128
-
-ENTRYPOINT ["/app/go-transcode"]
-CMD ["serve"]

config.example.yaml

@@ -1,28 +0,0 @@
-# go-transcode VOD configuration. Hardware encoding is enabled by the
-# TRANSCODE_VAAPI_DEVICE environment variable in docker-compose.yml.
-
-vod:
-  media-dir: /media
-  transcode-dir: /tmp/go-transcode
-  cache: true
-  cache-dir: /cache
-  # Avoid expensive full-file keyframe scans. FFmpeg forces HLS boundaries.
-  video-keyframes: false
-  video-profiles:
-    480p:
-      width: 854
-      height: 480
-      bitrate: 1200
-
-    720p:
-      width: 1280
-      height: 720
-      bitrate: 3000
-
-    1080p:
-      width: 1920
-      height: 1080
-      bitrate: 6000
-
-  audio-profile:
-    bitrate: 192

docker-compose.yml

@@ -15,19 +15,17 @@ services:
   transcode:
     build:
       context: .
-      dockerfile: Dockerfile.transcode
+      dockerfile: Dockerfile.govod
     devices:
       - /dev/dri:/dev/dri
     environment:
-      - TRANSCODE_VAAPI_DEVICE=/dev/dri/renderD128
       - LIBVA_DRIVER_NAME=i965
     volumes:
       - /mnt/user/plex:/media:ro
-      - ./transcode-config.yaml:/app/config.yaml:ro
-      - transcode-cache-v5:/cache
+      - govod-cache:/cache
     ports:
       - "8090:8080"
     restart: unless-stopped
 
 volumes:
-  transcode-cache-v5:
+  govod-cache:

govod-config.json

@@ -0,0 +1,16 @@
+{
+  "bind": ":8080",
+  "tempdir": "/cache",
+  "chunkSize": 4,
+  "lookBehind": 3,
+  "goalBufferMin": 3,
+  "goalBufferMax": 8,
+  "streamIdleTime": 60,
+  "managerIdleTime": 300,
+  "qf": 22,
+  "vaapi": true,
+  "vaapiLowPower": false,
+  "useTranspose": true,
+  "forceSwTranspose": false,
+  "useGopSize": true
+}

internal/api/api.go

@@ -1,8 +1,11 @@
 package api
 
 import (
+	"bytes"
 	"encoding/json"
+	"fmt"
 	"net/http"
+	"os/exec"
 	"strconv"
 	"strings"
 
@@ -27,6 +30,7 @@ func NewRouter(lib *library.Library) http.Handler {
 	r.Get("/api/search", searchHandler(lib))
 	r.Get("/api/rescan", rescanHandler(lib))
 	r.Get("/api/stream/{id}", streamHandler(lib))
+	r.Get("/api/subtitles/{id}/{index}", subtitleHandler(lib))
 
 	return r
 }
@@ -101,12 +105,15 @@ func detailHandler(lib *library.Library) http.HandlerFunc {
 		probeResult, err := probe.Run(fullPath)
 		if err != nil {
 			writeJSON(w, http.StatusOK, map[string]any{
-				"item":       item,
-				"probe":      nil,
+				"item":        item,
+				"probe":       nil,
 				"probe_error": err.Error(),
 			})
 			return
 		}
+		for i := range probeResult.Subtitles {
+			probeResult.Subtitles[i].URL = fmt.Sprintf("/api/subtitles/%s/%d", id, probeResult.Subtitles[i].Index)
+		}
 
 		writeJSON(w, http.StatusOK, map[string]any{
 			"item":  item,
@@ -142,6 +149,75 @@ func streamHandler(lib *library.Library) http.HandlerFunc {
 	}
 }
 
+func subtitleHandler(lib *library.Library) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		id := chi.URLParam(r, "id")
+		index, err := strconv.Atoi(chi.URLParam(r, "index"))
+		if err != nil || index < 0 {
+			http.Error(w, "invalid subtitle index", http.StatusBadRequest)
+			return
+		}
+
+		_, fullPath, ok := lib.Get(id)
+		if !ok {
+			http.Error(w, "not found", http.StatusNotFound)
+			return
+		}
+
+		metadata, err := probe.Run(fullPath)
+		if err != nil {
+			http.Error(w, "unable to inspect media", http.StatusInternalServerError)
+			return
+		}
+
+		var subtitle *probe.SubTrack
+		for i := range metadata.Subtitles {
+			if metadata.Subtitles[i].Index == index {
+				subtitle = &metadata.Subtitles[i]
+				break
+			}
+		}
+		if subtitle == nil {
+			http.Error(w, "subtitle track not found", http.StatusNotFound)
+			return
+		}
+
+		if !isTextSubtitleCodec(subtitle.Codec) {
+			http.Error(w, fmt.Sprintf("subtitle codec %q requires OCR or burn-in", subtitle.Codec), http.StatusUnsupportedMediaType)
+			return
+		}
+
+		cmd := exec.CommandContext(r.Context(), "ffmpeg",
+			"-v", "error",
+			"-i", fullPath,
+			"-map", fmt.Sprintf("0:%d", index),
+			"-c:s", "webvtt",
+			"-f", "webvtt",
+			"pipe:1",
+		)
+		var stdout, stderr bytes.Buffer
+		cmd.Stdout = &stdout
+		cmd.Stderr = &stderr
+		if err := cmd.Run(); err != nil {
+			http.Error(w, fmt.Sprintf("subtitle conversion failed: %s", strings.TrimSpace(stderr.String())), http.StatusInternalServerError)
+			return
+		}
+
+		w.Header().Set("Content-Type", "text/vtt; charset=utf-8")
+		w.Header().Set("Cache-Control", "public, max-age=86400")
+		_, _ = w.Write(stdout.Bytes())
+	}
+}
+
+func isTextSubtitleCodec(codec string) bool {
+	switch strings.ToLower(codec) {
+	case "ass", "ssa", "subrip", "srt", "text", "webvtt", "mov_text":
+		return true
+	default:
+		return false
+	}
+}
+
 func rescanHandler(lib *library.Library) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
 		go lib.Scan()

internal/api/api_test.go

@@ -0,0 +1,17 @@
+package api
+
+import "testing"
+
+func TestIsTextSubtitleCodec(t *testing.T) {
+	for _, codec := range []string{"ass", "ssa", "subrip", "srt", "text", "webvtt", "mov_text", "SUBRIP"} {
+		if !isTextSubtitleCodec(codec) {
+			t.Errorf("expected %q to be treated as text", codec)
+		}
+	}
+
+	for _, codec := range []string{"hdmv_pgs_subtitle", "dvd_subtitle", "dvb_subtitle", ""} {
+		if isTextSubtitleCodec(codec) {
+			t.Errorf("expected %q to require bitmap handling", codec)
+		}
+	}
+}

internal/library/library.go

@@ -61,7 +61,7 @@ func (l *Library) Scan() error {
 			RelPath:   f.RelPath,
 			Size:      f.Size,
 			ModTime:   f.ModTime,
-			StreamURL: "/vod/" + f.RelPath + "/index.m3u8",
+			StreamURL: "/vod/" + id + "/media/" + f.RelPath + "/index.m3u8",
 			fullPath:  f.Path,
 		}
 		items = append(items, item)

internal/probe/probe.go

@@ -8,13 +8,13 @@ import (
 )
 
 type ProbeResult struct {
-	Duration   float64       `json:"duration"`
-	Width      int           `json:"width"`
-	Height     int           `json:"height"`
-	VideoCodec string        `json:"video_codec"`
-	Bitrate    int64         `json:"bitrate"`
-	Audio      []AudioTrack  `json:"audio_tracks"`
-	Subtitles  []SubTrack    `json:"subtitle_tracks"`
+	Duration   float64      `json:"duration"`
+	Width      int          `json:"width"`
+	Height     int          `json:"height"`
+	VideoCodec string       `json:"video_codec"`
+	Bitrate    int64        `json:"bitrate"`
+	Audio      []AudioTrack `json:"audio_tracks"`
+	Subtitles  []SubTrack   `json:"subtitle_tracks"`
 }
 
 type AudioTrack struct {
@@ -28,6 +28,7 @@ type SubTrack struct {
 	Index    int    `json:"index"`
 	Codec    string `json:"codec"`
 	Language string `json:"language"`
+	URL      string `json:"url,omitempty"`
 }
 
 type ffprobeOutput struct {

patches/go-transcode-vaapi.patch

@@ -1,139 +0,0 @@
-diff --git a/hlsvod/manager.go b/hlsvod/manager.go
-index 662c41d..51460f8 100644
---- a/hlsvod/manager.go
-+++ b/hlsvod/manager.go
-@@ -378,12 +378,17 @@ func (m *ManagerCtx) waitForSegment(index int) (chan struct{}, bool) {
- 
- func (m *ManagerCtx) transcodeSegments(offset, limit int) error {
- 	logger := m.logger.With().Int("offset", offset).Int("limit", limit).Logger()
-+	inputCodec := ""
-+	if m.metadata.Video != nil {
-+		inputCodec = m.metadata.Video.CodecName
-+	}
- 
- 	segmentTimes := m.breakpoints[offset : offset+limit+1]
- 	logger.Info().Interface("segments-times", segmentTimes).Msg("transcoding segments")
- 
- 	segments, err := TranscodeSegments(m.ctx, m.config.FFmpegBinary, TranscodeConfig{
- 		InputFilePath: m.config.MediaPath,
-+		InputCodec:    inputCodec,
- 		OutputDirPath: m.config.TranscodeDir,
- 		SegmentPrefix: m.config.SegmentPrefix, // This does not need to match.
- 
-diff --git a/hlsvod/probe.go b/hlsvod/probe.go
-index c39461c..58c3edd 100644
---- a/hlsvod/probe.go
-+++ b/hlsvod/probe.go
-@@ -83,6 +83,7 @@ func ProbeMedia(ctx context.Context, ffprobeBinary string, inputFilePath string)
- 			}
- 
- 			data.Video = &ProbeVideoData{
-+				CodecName: stream.CodecName,
- 				Width:    stream.Width,
- 				Height:   stream.Height,
- 				Duration: duration,
-@@ -118,6 +119,7 @@ func ProbeMedia(ctx context.Context, ffprobeBinary string, inputFilePath string)
- }
- 
- type ProbeVideoData struct {
-+	CodecName  string
- 	Width      int
- 	Height     int
- 	Duration   time.Duration
-diff --git a/hlsvod/transcode.go b/hlsvod/transcode.go
-index 9bbe037..c0297fe 100644
---- a/hlsvod/transcode.go
-+++ b/hlsvod/transcode.go
-@@ -5,6 +5,7 @@ import (
- 	"context"
- 	"fmt"
- 	"log"
-+	"os"
- 	"os/exec"
- 	"path"
- 	"strings"
-@@ -13,6 +14,7 @@ import (
- 
- type TranscodeConfig struct {
- 	InputFilePath string // Transcoded video input.
-+	InputCodec    string // ffprobe codec name, e.g. h264 or hevc.
- 	OutputDirPath string // Segments output path.
- 	SegmentPrefix string // e.g. prefix-000001.ts
- 	SegmentOffset int    // Start segment number.
-@@ -69,6 +71,21 @@ func TranscodeSegments(ctx context.Context, ffmpegBinary string, config Transcod
- 		}...)
- 	}
- 
-+	// Haswell can decode H.264 in hardware, but not HEVC Main 10. Use full
-+	// hardware transcoding for H.264 and software decode plus VAAPI upload for
-+	// other codecs. Scaling and H.264 encoding always remain on the GPU.
-+	vaapiDevice := os.Getenv("TRANSCODE_VAAPI_DEVICE")
-+	hardwareDecode := vaapiDevice != "" && config.InputCodec == "h264"
-+	if vaapiDevice != "" {
-+		args = append(args, []string{
-+			"-init_hw_device", "vaapi=va:" + vaapiDevice,
-+			"-filter_hw_device", "va",
-+		}...)
-+		if hardwareDecode {
-+			args = append(args, "-hwaccel", "vaapi", "-hwaccel_output_format", "vaapi")
-+		}
-+	}
-+
- 	// Input specs
- 	args = append(args, []string{
- 		"-i", config.InputFilePath, // Input file
-@@ -89,14 +106,30 @@ func TranscodeSegments(ctx context.Context, ffmpegBinary string, config Transcod
- 			scale = fmt.Sprintf("scale=%d:-2", profile.Width)
- 		}
- 
--		args = append(args, []string{
--			"-vf", scale,
--			"-c:v", "libx264",
--			"-preset", "faster",
--			"-profile:v", "high",
--			"-level:v", "4.0",
--			"-b:v", fmt.Sprintf("%dk", profile.Bitrate),
--		}...)
-+		if vaapiDevice != "" {
-+			if profile.Width >= profile.Height {
-+				scale = fmt.Sprintf("scale_vaapi=w=-2:h=%d:format=nv12", profile.Height)
-+			} else {
-+				scale = fmt.Sprintf("scale_vaapi=w=%d:h=-2:format=nv12", profile.Width)
-+			}
-+			if !hardwareDecode {
-+				scale = "format=nv12,hwupload," + scale
-+			}
-+			args = append(args, []string{
-+				"-vf", scale, "-c:v", "h264_vaapi", "-profile:v", "high",
-+				"-level:v", "4.0", "-rc_mode", "VBR",
-+				"-b:v", fmt.Sprintf("%dk", profile.Bitrate),
-+				"-maxrate", fmt.Sprintf("%dk", profile.Bitrate*3/2),
-+				"-bufsize", fmt.Sprintf("%dk", profile.Bitrate*2),
-+				"-bf", "2", "-compression_level", "0",
-+			}...)
-+		} else {
-+			args = append(args, []string{
-+				"-vf", scale, "-c:v", "libx264", "-preset", "faster",
-+				"-profile:v", "high", "-level:v", "4.0",
-+				"-b:v", fmt.Sprintf("%dk", profile.Bitrate),
-+			}...)
-+		}
- 	}
- 
- 	// Audio specs
-@@ -106,12 +127,15 @@ func TranscodeSegments(ctx context.Context, ffmpegBinary string, config Transcod
- 		args = append(args, []string{
- 			"-c:a", "aac",
- 			"-b:a", fmt.Sprintf("%dk", profile.Bitrate),
-+			"-af", "aresample=async=1:min_hard_comp=0.100",
- 		}...)
- 	}
- 
- 	// Segmenting specs
- 	args = append(args, []string{
- 		"-f", "segment",
-+		"-muxdelay", "0",
-+		"-muxpreload", "0",
- 		"-segment_time_delta", "0.2",
- 		"-segment_format", "mpegts",
- 		"-segment_times", commaSeparatedSegTimes,

patches/go-vod-sheepflix.patch

@@ -0,0 +1,69 @@
+diff --git a/transcoder/handler.go b/transcoder/handler.go
+index 8c5bb08..0ad7458 100644
+--- a/transcoder/handler.go
++++ b/transcoder/handler.go
+@@ -37,6 +37,14 @@ func NewHandler(c *Config) *Handler {
+ }
+ 
+ func (h *Handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
++	w.Header().Set("Access-Control-Allow-Origin", "*")
++	w.Header().Set("Access-Control-Allow-Methods", "GET, HEAD, OPTIONS")
++	w.Header().Set("Access-Control-Allow-Headers", "Range, Content-Type")
++	if r.Method == http.MethodOptions {
++		w.WriteHeader(http.StatusNoContent)
++		return
++	}
++
+ 	// Check version if monitoring is enabled
+ 	if h.c.VersionMonitor && !h.versionOk(w, r) {
+ 		return
+@@ -55,15 +63,15 @@ func (h *Handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
+ 	}
+ 
+ 	// Serve actual file from manager
+-	if len(parts) < 3 {
++	if len(parts) < 5 || parts[0] != "vod" {
+ 		log.Println("Invalid URL", url)
+ 		w.WriteHeader(http.StatusBadRequest)
+ 		return
+ 	}
+ 
+ 	// Get streamid and chunk
+-	streamid := parts[0]
+-	path := "/" + strings.Join(parts[1:len(parts)-1], "/")
++	streamid := parts[1]
++	path := "/" + strings.Join(parts[2:len(parts)-1], "/")
+ 	chunk := parts[len(parts)-1]
+ 
+ 	// Check if POST request to create temp file
+diff --git a/transcoder/stream.go b/transcoder/stream.go
+index cf353fc..f8694fc 100644
+--- a/transcoder/stream.go
++++ b/transcoder/stream.go
+@@ -370,8 +370,12 @@ func (s *Stream) transcodeArgs(startAt float64, isHls bool) []string {
+ 	// Check whether hwaccel should be used
+ 	if s.c.VAAPI {
+ 		CV = ENCODER_VAAPI
+-		extra := "-hwaccel vaapi -hwaccel_device /dev/dri/renderD128 -hwaccel_output_format vaapi -init_hw_device vaapi=memories:/dev/dri/renderD128 -filter_hw_device memories"
++		extra := "-init_hw_device vaapi=memories:/dev/dri/renderD128 -filter_hw_device memories"
+ 		args = append(args, strings.Split(extra, " ")...)
++		if s.m.probe.CodecName == CODEC_H264 {
++			extra = "-hwaccel vaapi -hwaccel_device /dev/dri/renderD128 -hwaccel_output_format vaapi"
++			args = append(args, strings.Split(extra, " ")...)
++		}
+ 	} else if s.c.NVENC {
+ 		CV = ENCODER_NVENC
+ 		extra := "-hwaccel cuda"
+@@ -397,7 +401,11 @@ func (s *Stream) transcodeArgs(startAt float64, isHls bool) []string {
+ 	scalerArgs = append(scalerArgs, "force_original_aspect_ratio=decrease")
+ 
+ 	if CV == ENCODER_VAAPI {
+-		format = "format=nv12|vaapi,hwupload"
++		if s.m.probe.CodecName == CODEC_H264 {
++			format = "format=nv12|vaapi,hwupload"
++		} else {
++			format = "format=nv12,hwupload"
++		}
+ 		scaler = "scale_vaapi"
+ 		scalerArgs = append(scalerArgs, "format=nv12")
+ 	} else if CV == ENCODER_NVENC {

transcode-config.yaml

@@ -1,26 +0,0 @@
-cors: true
-
-vod:
-  media-dir: /media
-  cache: true
-  cache-dir: /cache
-  # Full-file keyframe probing is expensive and can be OOM-killed on Unraid.
-  # FFmpeg will force segment-boundary keyframes during transcoding instead.
-  video-keyframes: false
-
-  video-profiles:
-    1080p:
-      width: 1920
-      height: 1080
-      bitrate: 6000
-    720p:
-      width: 1280
-      height: 720
-      bitrate: 3000
-    480p:
-      width: 854
-      height: 480
-      bitrate: 1200
-
-  audio-profile:
-    bitrate: 192