Support for multiple audio encoding modes when using segmented encoding

checks.gradle

@@ -15,6 +15,9 @@ jacocoTestCoverageVerification {
                     '*QueueService.migrateQueues()',
                     '*.ShutdownHandler.*',
                     '*FfmpegExecutor.runFfmpeg$lambda$?(java.lang.Process)',
+                    '*FfmpegExecutorKt.getProgressRegex()',
+                    '*FilterSettings.*',
+                    'se.svt.oss.encore.service.EncoreService.handleProgress.1.1.emit(int, kotlin.coroutines.Continuation)',
             ]
             limit {
                 counter = 'LINE'

encore-common/src/main/kotlin/se/svt/oss/encore/config/EncodingProperties.kt

@@ -5,8 +5,13 @@
 package se.svt.oss.encore.config
 
 import org.springframework.boot.context.properties.NestedConfigurationProperty
+import se.svt.oss.encore.model.AudioEncodingMode
 import se.svt.oss.encore.model.profile.ChannelLayout
 
+data class SegmentedEncodingProperties(
+    val audioEncodingMode: AudioEncodingMode = AudioEncodingMode.ENCODE_WITH_VIDEO,
+)
+
 data class EncodingProperties(
     @NestedConfigurationProperty
     val audioMixPresets: Map<String, AudioMixPreset> = mapOf("default" to AudioMixPreset()),
@@ -15,4 +20,6 @@ data class EncodingProperties(
     val flipWidthHeightIfPortrait: Boolean = true,
     val exitOnError: Boolean = true,
     val globalParams: LinkedHashMap<String, Any?> = linkedMapOf(),
+    @NestedConfigurationProperty
+    val segmentedEncoding: SegmentedEncodingProperties = SegmentedEncodingProperties(),
 )

encore-common/src/main/kotlin/se/svt/oss/encore/model/AudioEncodingMode.kt

@@ -0,0 +1,28 @@
+// SPDX-FileCopyrightText: 2025 Eyevinn Technology AB
+//
+// SPDX-License-Identifier: EUPL-1.2
+
+package se.svt.oss.encore.model
+
+/**
+ * Defines how audio should be encoded when using segmented encoding.
+ */
+enum class AudioEncodingMode {
+    /**
+     * Encode audio and video together in the same segments.
+     * Creates N tasks of type AUDIOVIDEOSEGMENT.
+     */
+    ENCODE_WITH_VIDEO,
+
+    /**
+     * Encode audio separately from video as a single full-length file (not segmented).
+     * Creates 1 AUDIOFULL task + N VIDEOSEGMENT tasks.
+     */
+    ENCODE_SEPARATELY_FULL,
+
+    /**
+     * Encode audio separately from video, with both audio and video segmented.
+     * Creates N AUDIOSEGMENT tasks + N VIDEOSEGMENT tasks (2N total tasks).
+     */
+    ENCODE_SEPARATELY_SEGMENTED,
+}

encore-common/src/main/kotlin/se/svt/oss/encore/model/EncoreJob.kt

@@ -21,6 +21,55 @@ import se.svt.oss.mediaanalyzer.file.MediaFile
 import java.time.OffsetDateTime
 import java.util.UUID
 
+data class SegmentedEncodingInfo(
+    @field:Schema(
+        description = "Length of each segment in seconds. Should be a multiple of target GOP.",
+        example = "19.2",
+        readOnly = true,
+        nullable = false,
+    )
+    val segmentLength: Double,
+    @field:Schema(
+        description = "Number of video segments",
+        nullable = false,
+        readOnly = true,
+    )
+    val numSegments: Int,
+    @field:Schema(
+        description = "Number of encoding tasks used for this job. This will be equal to numSegments plus numAudioSegments",
+        nullable = false,
+        readOnly = true,
+    )
+    val numTasks: Int,
+    @field:Schema(
+        description = "The audio encoding mode used for this job.",
+        example = "ENCODE_WITH_VIDEO",
+        nullable = false,
+        readOnly = true,
+    )
+    val audioEncodingMode: AudioEncodingMode,
+    @field:Schema(
+        description = "Audio segment padding in seconds (added at start/end of segments to avoid artifacts). Only relevant in ENCODE_SEPARATELY_SEGMENTED mode.",
+        example = "0.04267",
+        nullable = false,
+        readOnly = true,
+    )
+    val audioSegmentPadding: Double = 0.0,
+    @field:Schema(
+        description = "Length of each audio segment in seconds. Only relevant in ENCODE_SEPARATELY_SEGMENTED mode.",
+        example = "256.0",
+        nullable = false,
+        readOnly = true,
+    )
+    val audioSegmentLength: Double = 0.0,
+    @field:Schema(
+        description = "Number of audio segments",
+        nullable = false,
+        readOnly = true,
+    )
+    val numAudioSegments: Int,
+)
+
 @Validated
 @RedisHash("encore-jobs", timeToLive = (60 * 60 * 24 * 7).toLong()) // 1 week ttl
 @Tag(name = "encorejob")
@@ -107,6 +156,29 @@ data class EncoreJob(
     @field:Positive
     val segmentLength: Double? = null,
 
+    @field:Schema(
+        description = "Defines how audio should be encoded when using segmented encoding. ENCODE_WITH_VIDEO: audio and video together in segments; ENCODE_SEPARATELY_FULL: audio separately as full file; ENCODE_SEPARATELY_SEGMENTED: audio separately in segments.",
+        example = "ENCODE_WITH_VIDEO",
+        defaultValue = "ENCODE_WITH_VIDEO",
+        nullable = true,
+    )
+    val audioEncodingMode: AudioEncodingMode? = null,
+
+    @field:Schema(
+        description = "Length of audio segments in seconds when using ENCODE_SEPARATELY_SEGMENTED mode. If not specified, a value close to 256s will be calculated that is a multiple of the audio frame size.",
+        example = "256.0",
+        nullable = true,
+    )
+    @field:Positive
+    val audioSegmentLength: Double? = null,
+
+    @field:Schema(
+        description = "Properties for segmented encoding, or null if not used",
+        nullable = true,
+        readOnly = true,
+    )
+    var segmentedEncodingInfo: SegmentedEncodingInfo? = null,
+
     @field:Schema(
         description = "The exception message, if the EncoreJob failed",
         example = "input/output error",

encore-common/src/main/kotlin/se/svt/oss/encore/model/profile/AudioEncode.kt

@@ -17,7 +17,7 @@ import se.svt.oss.encore.model.output.AudioStreamEncode
 import se.svt.oss.encore.model.output.Output
 
 data class AudioEncode(
-    val codec: String = "libfdk_aac",
+    override val codec: String = "libfdk_aac",
     val bitrate: String? = null,
     val samplerate: Int = 48000,
     val channelLayout: ChannelLayout = ChannelLayout.CH_LAYOUT_STEREO,

encore-common/src/main/kotlin/se/svt/oss/encore/model/profile/AudioEncoder.kt

@@ -12,7 +12,8 @@ private val log = KotlinLogging.logger { }
 abstract class AudioEncoder : OutputProducer {
 
     abstract val optional: Boolean
-    abstract val enabled: Boolean
+    abstract val codec: String
+    abstract override val enabled: Boolean
 
     fun logOrThrow(message: String): Output? {
         if (optional || !enabled) {

encore-common/src/main/kotlin/se/svt/oss/encore/model/profile/OutputProducer.kt

@@ -21,5 +21,6 @@ import se.svt.oss.encore.model.output.Output
     JsonSubTypes.Type(value = ThumbnailMapEncode::class, name = "ThumbnailMapEncode"),
 )
 interface OutputProducer {
+    val enabled: Boolean
     fun getOutput(job: EncoreJob, encodingProperties: EncodingProperties): Output?
 }

encore-common/src/main/kotlin/se/svt/oss/encore/model/profile/SimpleAudioEncode.kt

@@ -13,7 +13,7 @@ import se.svt.oss.encore.model.output.AudioStreamEncode
 import se.svt.oss.encore.model.output.Output
 
 data class SimpleAudioEncode(
-    val codec: String = "libfdk_aac",
+    override val codec: String = "libfdk_aac",
     val bitrate: String? = null,
     val samplerate: Int? = null,
     val suffix: String = "_$codec",

encore-common/src/main/kotlin/se/svt/oss/encore/model/profile/ThumbnailEncode.kt

@@ -25,7 +25,7 @@ data class ThumbnailEncode(
     val suffixZeroPad: Int = 2,
     val inputLabel: String = DEFAULT_VIDEO_LABEL,
     val optional: Boolean = false,
-    val enabled: Boolean = true,
+    override val enabled: Boolean = true,
     val intervalSeconds: Double? = null,
     val decodeOutput: Int? = null,
 ) : OutputProducer {

encore-common/src/main/kotlin/se/svt/oss/encore/model/profile/ThumbnailMapEncode.kt

@@ -26,7 +26,7 @@ data class ThumbnailMapEncode(
     val rows: Int = 20,
     val quality: Int = 5,
     val optional: Boolean = true,
-    val enabled: Boolean = true,
+    override val enabled: Boolean = true,
     val suffix: String = "_${cols}x${rows}_${tileWidth}x${tileHeight}_thumbnail_map",
     val format: String = "jpg",
     val inputLabel: String = DEFAULT_VIDEO_LABEL,

encore-common/src/main/kotlin/se/svt/oss/encore/model/profile/VideoEncode.kt

@@ -34,7 +34,7 @@ interface VideoEncode : OutputProducer {
     val codec: String
     val inputLabel: String
     val optional: Boolean
-    val enabled: Boolean
+    override val enabled: Boolean
     val cropTo: FractionString?
     val padTo: FractionString?
 

encore-common/src/main/kotlin/se/svt/oss/encore/model/queue/QueueItem.kt

@@ -14,5 +14,18 @@ data class QueueItem(
     val id: String,
     val priority: Int = 0,
     val created: LocalDateTime = LocalDateTime.now(),
-    val segment: Int? = null,
+    val task: Task? = null,
+)
+
+enum class TaskType {
+    AUDIOVIDEOSEGMENT,
+    VIDEOSEGMENT,
+    AUDIOFULL,
+    AUDIOSEGMENT,
+}
+
+data class Task(
+    val type: TaskType,
+    val taskNo: Int,
+    val segment: Int,
 )

encore-common/src/main/kotlin/se/svt/oss/encore/process/SegmentUtil.kt

@@ -5,30 +5,30 @@
 package se.svt.oss.encore.process
 
 import se.svt.oss.encore.model.EncoreJob
-import se.svt.oss.mediaanalyzer.file.MediaContainer
-import kotlin.math.ceil
-
-fun EncoreJob.segmentLengthOrThrow() = segmentLength ?: throw RuntimeException("No segmentLength in job!")
-
-fun EncoreJob.numSegments(): Int {
-    val segLen = segmentLengthOrThrow()
-    val readDuration = duration
-    return if (readDuration != null) {
-        ceil(readDuration / segLen).toInt()
-    } else {
-        val segments =
-            inputs.map { ceil(((it.analyzed as MediaContainer).duration - (it.seekTo ?: 0.0)) / segLen).toInt() }.toSet()
-        if (segments.size > 1) {
-            throw RuntimeException("Inputs differ in length")
-        }
-        segments.first()
+
+fun EncoreJob.segmentLengthOrThrow() = segmentedEncodingInfoOrThrow().segmentLength
+
+fun EncoreJob.segmentedEncodingInfoOrThrow() = segmentedEncodingInfo ?: throw RuntimeException("No segmentedEncodingInfo in job!")
+
+fun EncoreJob.segmentDuration(segmentNumber: Int): Double {
+    val numSegments = segmentedEncodingInfoOrThrow().numSegments
+    return when {
+        duration == null -> segmentLengthOrThrow()
+        segmentNumber < numSegments - 1 -> segmentLengthOrThrow()
+        segmentNumber == numSegments - 1 ->
+            // This correctly handles the case where the duration is an exact multiple of the segment length
+            duration!! - segmentLengthOrThrow() * (numSegments - 1)
+        else -> throw IllegalArgumentException("segmentNumber $segmentNumber is out of range for job with $numSegments segments")
     }
 }
 
-fun EncoreJob.segmentDuration(segmentNumber: Int): Double = when {
-    duration == null -> segmentLengthOrThrow()
-    segmentNumber < numSegments() - 1 -> segmentLengthOrThrow()
-    else -> duration!! % segmentLengthOrThrow()
+fun EncoreJob.baseName(segmentNumber: Int) = "${baseName}_%05d".format(segmentNumber)
+
+fun EncoreJob.segmentSuffixFromFilename(file: String): String {
+    val regex = Regex("${baseName}_\\d{5}(.*)")
+    val match = regex.find(file) ?: throw RuntimeException("Could not find segment suffix for file $file")
+    return match.groupValues[1]
 }
 
-fun EncoreJob.baseName(segmentNumber: Int) = "${baseName}_%05d".format(segmentNumber)
+fun EncoreJob.targetFilenameFromSegmentFilename(segmentFile: String) =
+    segmentFile.replace(Regex("^${baseName}_\\d{5}"), baseName)