Add end of stream flag to AudioDecoder and increase min api to 23

app/build.gradle

@@ -25,7 +25,7 @@ android {
     namespace 'com.github.jing332.tts_server_android'
     defaultConfig {
         applicationId 'com.github.jing332.tts_server_android'
-        minSdk 21
+        minSdk 23
         targetSdk 34
         versionCode gitCommits
         versionName version

app/src/main/java/com/github/jing332/tts_server_android/help/audio/AudioDecoder.kt

@@ -4,15 +4,13 @@ import android.media.MediaCodec
 import android.media.MediaCodec.BufferInfo
 import android.media.MediaExtractor
 import android.media.MediaFormat
-import android.os.Build
 import android.os.SystemClock
 import android.text.TextUtils
 import android.util.Log
 import com.github.jing332.tts_server_android.help.audio.AudioDecoderException.Companion.ERROR_CODE_NO_AUDIO_TRACK
 import com.github.jing332.tts_server_android.utils.GcManager
 import kotlinx.coroutines.delay
 import kotlinx.coroutines.isActive
-import okio.ByteString.Companion.toByteString
 import java.io.IOException
 import java.io.InputStream
 import java.nio.ByteBuffer
@@ -62,12 +60,7 @@ class AudioDecoder {
         private fun getFormats(srcData: ByteArray): List<MediaFormat> {
             kotlin.runCatching {
                 val mediaExtractor = MediaExtractor()
-                if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.M)
-                    mediaExtractor.setDataSource(ByteArrayMediaDataSource(srcData))
-                else
-                    mediaExtractor.setDataSource(
-                        "data:" + "" + ";base64," + srcData.toByteString().base64()
-                    )
+                mediaExtractor.setDataSource(ByteArrayMediaDataSource(srcData))
 
                 val formats = mutableListOf<MediaFormat>()
                 for (i in 0 until mediaExtractor.trackCount) {
@@ -116,6 +109,7 @@ class AudioDecoder {
             }
         }
         mediaCodec!!.reset()
+        mediaFormat.setInteger(MediaFormat.KEY_PRIORITY, 0)
         mediaCodec!!.configure(mediaFormat, null, null, 0)
         return mediaCodec as MediaCodec
     }
@@ -135,12 +129,7 @@ class AudioDecoder {
                 onRead.invoke(data)
                 return
             }
-            if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.M)
-                mediaExtractor.setDataSource(ByteArrayMediaDataSource(srcData))
-            else
-                mediaExtractor.setDataSource(
-                    "data:" + currentMime + ";base64," + srcData.toByteString().base64()
-                )
+            mediaExtractor.setDataSource(ByteArrayMediaDataSource(srcData))
 
             decodeInternal(mediaExtractor, sampleRate) {
                 onRead.invoke(it)
@@ -165,13 +154,7 @@ class AudioDecoder {
     ) {
         val mediaExtractor = MediaExtractor()
         try {
-            if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.M)
-                mediaExtractor.setDataSource(InputStreamMediaDataSource(ins))
-            else
-                throw AudioDecoderException(
-                    AudioDecoderException.ERROR_CODE_NOT_SUPPORT_A5,
-                    "音频流解码器不支持 Android 5"
-                )
+            mediaExtractor.setDataSource(InputStreamMediaDataSource(ins))
 
             decodeInternal(mediaExtractor, sampleRate, timeoutUs) {
                 onRead.invoke(it)
@@ -208,31 +191,38 @@ class AudioDecoder {
             bufferInfo.presentationTimeUs = mediaExtractor.sampleTime
 
             inputBuffer = mediaCodec.getInputBuffer(inputIndex)
-            if (inputBuffer != null) {
-                inputBuffer.clear()
-            } else
+            if (inputBuffer == null) {
                 continue
+            }
 
             //从流中读取的采样数量
             val sampleSize = mediaExtractor.readSampleData(inputBuffer, 0)
             if (sampleSize > 0) {
                 bufferInfo.size = sampleSize
                 //入队解码
-                mediaCodec.queueInputBuffer(inputIndex, 0, sampleSize, 0, 0)
+                mediaCodec.queueInputBuffer(inputIndex, 0, sampleSize, mediaExtractor.sampleTime, 0)
                 //移动到下一个采样点
                 if (!mediaExtractor.nextSample(startNanos)) {
                     Log.d(TAG, "nextSample(): 已到达流末尾EOF")
                 }
-            } else
-                break
+            } else {
+                mediaCodec.queueInputBuffer(
+                    inputIndex,
+                    0,
+                    0,
+                    0,
+                    MediaCodec.BUFFER_FLAG_END_OF_STREAM
+                )
+            }
+
+            //取解码后的数据/
+            val outputIndex = mediaCodec.dequeueOutputBuffer(bufferInfo, timeoutUs)
 
-            //取解码后的数据
-            var outputIndex = mediaCodec.dequeueOutputBuffer(bufferInfo, timeoutUs)
             //不一定能一次取完，所以要循环取
             var outputBuffer: ByteBuffer?
             val pcmData = ByteArray(bufferInfo.size)
 
-            while (coroutineContext.isActive && outputIndex >= 0) {
+            if (outputIndex >= 0) {
                 outputBuffer = mediaCodec.getOutputBuffer(outputIndex)
                 if (outputBuffer != null) {
                     outputBuffer.get(pcmData)
@@ -241,7 +231,10 @@ class AudioDecoder {
 
                 onRead.invoke(pcmData)
                 mediaCodec.releaseOutputBuffer(/* index = */ outputIndex, /* render = */ false)
-                outputIndex = mediaCodec.dequeueOutputBuffer(bufferInfo, timeoutUs)
+            }
+
+            if ((bufferInfo.flags and MediaCodec.BUFFER_FLAG_END_OF_STREAM) != 0) {
+                break
             }
         }
     }
@@ -262,7 +255,7 @@ class AudioDecoder {
         for (i in 0 until trackCount) {
             trackFormat = getTrackFormat(i)
             mime = trackFormat.getString(MediaFormat.KEY_MIME)
-            if (!TextUtils.isEmpty(mime) && mime!!.startsWith("audio")) {
+            if (!TextUtils.isEmpty(mime) && mime!!.startsWith("audio/")) {
                 audioTrackIndex = i
                 break
             }

app/src/main/java/com/github/jing332/tts_server_android/help/audio/ExoAudioPlayer.kt

@@ -20,7 +20,6 @@ import kotlinx.coroutines.coroutineScope
 import kotlinx.coroutines.launch
 import java.io.InputStream
 
-@SuppressLint("UnsafeOptInUsageError")
 class ExoAudioPlayer(val context: Context) {
     companion object {
         const val TAG = "AudioPlayer"
@@ -34,7 +33,7 @@ class ExoAudioPlayer(val context: Context) {
 
     // APP内音频播放器 必须在主线程调用
     private val exoPlayer by lazy {
-        ExoPlayer.Builder(context).setSkipSilenceEnabled(true).build().apply {
+        ExoPlayer.Builder(context).build().apply {
             playWhenReady = true
             addListener(object : Player.Listener {
                 @SuppressLint("SwitchIntDef")

app/src/main/java/com/github/jing332/tts_server_android/help/audio/InputStreamMediaDataSource.kt

@@ -1,20 +1,22 @@
 package com.github.jing332.tts_server_android.help.audio
 
 import android.media.MediaDataSource
-import android.os.Build
 import android.util.Log
-import androidx.annotation.RequiresApi
 import okio.buffer
 import okio.source
 import java.io.InputStream
+import kotlin.math.max
 
-@RequiresApi(Build.VERSION_CODES.M)
 class InputStreamMediaDataSource(private val inputStream: InputStream) : MediaDataSource() {
     companion object {
         const val TAG = "InputStreamDataSource"
     }
 
     private val bufferedInputStream = inputStream.source().buffer()
+    private var cache = ByteArray(1024)
+    private var limit = 0L
+    private var finishedRead = false
+
 
     override fun close() {
         Log.d(TAG, "close")
@@ -24,9 +26,27 @@ class InputStreamMediaDataSource(private val inputStream: InputStream) : MediaDa
     override fun readAt(position: Long, buffer: ByteArray, offset: Int, size: Int): Int {
         Log.d(TAG, "readAt: pos=$position, offset=$offset, size=$size")
         kotlin.runCatching {
-            return bufferedInputStream.read(buffer, offset, size).apply {
-                Log.d(TAG, "readAt: readLen=$this")
+            val newSize = position.toInt() + size
+            if (!finishedRead && newSize > cache.size) {
+                cache = cache.copyOf(32 - Integer.numberOfLeadingZeros((newSize * 2) - 1))
+            }
+            while (!finishedRead && newSize > limit) {
+                val readSize = bufferedInputStream.read(cache, limit.toInt(), size).apply {
+                    Log.d(TAG, "readAt: readLen=$this")
+                }
+                if (readSize == -1) {
+                    finishedRead = true
+                } else {
+                    limit += readSize
+                }
+            }
+            val bytesToRead = max(limit - position, -1)
+            if (bytesToRead > 0) {
+                cache.copyInto(buffer, offset, position.toInt(), limit.toInt())
             }
+
+            return bytesToRead.toInt()
+
         }.onFailure {
             Log.d(TAG, it.stackTraceToString())
         }

app/src/main/java/com/github/jing332/tts_server_android/help/audio/exo/DecoderAudioSink.kt

@@ -4,11 +4,9 @@ import androidx.media3.common.AudioAttributes
 import androidx.media3.common.AuxEffectInfo
 import androidx.media3.common.Format
 import androidx.media3.common.PlaybackParameters
-import androidx.media3.common.audio.AudioProcessor
 import androidx.media3.common.util.UnstableApi
 import androidx.media3.exoplayer.audio.AudioSink
 import androidx.media3.exoplayer.audio.AudioSink.SINK_FORMAT_SUPPORTED_WITH_TRANSCODING
-import androidx.media3.exoplayer.audio.SilenceSkippingAudioProcessor
 import java.nio.ByteBuffer
 
 @UnstableApi
@@ -17,14 +15,6 @@ import java.nio.ByteBuffer
  */
 class DecoderAudioSink(private val onPcmBuffer: (ByteBuffer) -> Unit) : AudioSink {
     private var timeUs: Long = 0L
-    private var skippingAudioProcessor =
-        SilenceSkippingAudioProcessor(
-            SilenceSkippingAudioProcessor.DEFAULT_MINIMUM_SILENCE_DURATION_US,
-            SilenceSkippingAudioProcessor.DEFAULT_SILENCE_RETENTION_RATIO,
-            SilenceSkippingAudioProcessor.DEFAULT_MAX_SILENCE_TO_KEEP_DURATION_US,
-            SilenceSkippingAudioProcessor.DEFAULT_MIN_VOLUME_TO_KEEP_PERCENTAGE,
-            SilenceSkippingAudioProcessor.DEFAULT_SILENCE_THRESHOLD_LEVEL
-        )
 
     companion object {
         const val TAG = "DecoderAudioSink"
@@ -46,9 +36,6 @@ class DecoderAudioSink(private val onPcmBuffer: (ByteBuffer) -> Unit) : AudioSin
         specifiedBufferSize: Int,
         outputChannels: IntArray?
     ) {
-        skippingAudioProcessor.setEnabled(true)
-        skippingAudioProcessor.configure(AudioProcessor.AudioFormat(inputFormat))
-        skippingAudioProcessor.flush()
     }
 
     override fun play() {
@@ -63,39 +50,18 @@ class DecoderAudioSink(private val onPcmBuffer: (ByteBuffer) -> Unit) : AudioSin
         presentationTimeUs: Long,
         encodedAccessUnitCount: Int
     ): Boolean {
-        while (!skippingAudioProcessor.isEnded) {
-            val outBuf = skippingAudioProcessor.output
-            if (outBuf.hasRemaining()) {
-                onPcmBuffer.invoke(outBuf)
-            }
-            if (!buffer.hasRemaining()) {
-                break
-            }
-            skippingAudioProcessor.queueInput(buffer)
-        }
+        onPcmBuffer.invoke(buffer)
         timeUs += presentationTimeUs
         return true
     }
 
     override fun playToEndOfStream() {
-        skippingAudioProcessor.queueEndOfStream()
-        while (!skippingAudioProcessor.isEnded) {
-            val outBuf = skippingAudioProcessor.output
-            if (outBuf.hasRemaining()) {
-                onPcmBuffer.invoke(outBuf)
-            }
-            skippingAudioProcessor.queueEndOfStream()
-        }
-    }
 
-    override fun isEnded(): Boolean {
-        return skippingAudioProcessor.isEnded
     }
 
-    override fun hasPendingData(): Boolean
-    {
-        return !skippingAudioProcessor.isEnded
-    }
+    override fun isEnded(): Boolean = true
+
+    override fun hasPendingData(): Boolean = true
 
     override fun setPlaybackParameters(playbackParameters: PlaybackParameters) {
 
@@ -108,7 +74,7 @@ class DecoderAudioSink(private val onPcmBuffer: (ByteBuffer) -> Unit) : AudioSin
 
     }
 
-    override fun getSkipSilenceEnabled(): Boolean = true
+    override fun getSkipSilenceEnabled(): Boolean = false
 
     override fun setAudioAttributes(audioAttributes: AudioAttributes) {
 
@@ -137,15 +103,16 @@ class DecoderAudioSink(private val onPcmBuffer: (ByteBuffer) -> Unit) : AudioSin
     }
 
     override fun pause() {
+
     }
 
     override fun flush() {
-        timeUs = 0
-        skippingAudioProcessor.flush()
+
     }
 
+
     override fun reset() {
-        flush()
-        skippingAudioProcessor.reset()
+
+
     }
-}
+}

app/src/main/java/com/github/jing332/tts_server_android/help/audio/exo/ExoAudioDecoder.kt

@@ -2,17 +2,11 @@ package com.github.jing332.tts_server_android.help.audio.exo
 
 import android.annotation.SuppressLint
 import android.content.Context
-import androidx.media3.common.C
 import androidx.media3.common.PlaybackException
 import androidx.media3.common.Player
 import androidx.media3.exoplayer.DefaultRenderersFactory
 import androidx.media3.exoplayer.ExoPlayer
 import androidx.media3.exoplayer.audio.AudioSink
-import androidx.media3.exoplayer.audio.DefaultAudioSink
-import androidx.media3.exoplayer.audio.SilenceSkippingAudioProcessor
-import androidx.media3.exoplayer.audio.SilenceSkippingAudioProcessor.DEFAULT_PADDING_SILENCE_US
-import androidx.media3.exoplayer.audio.SilenceSkippingAudioProcessor.DEFAULT_SILENCE_THRESHOLD_LEVEL
-import androidx.media3.exoplayer.audio.TeeAudioProcessor
 import androidx.media3.exoplayer.source.MediaSource
 import com.drake.net.utils.withMain
 import com.github.jing332.tts_server_android.help.audio.AudioDecoderException
@@ -37,10 +31,10 @@ class ExoAudioDecoder(val context: Context) {
                 context: Context,
                 enableFloatOutput: Boolean,
                 enableAudioTrackPlaybackParams: Boolean
-            ): AudioSink? {
+            ): AudioSink {
                 return DecoderAudioSink { callback?.onReadPcmAudio(it) }
             }
-        }.apply { forceEnableMediaCodecAsynchronousQueueing() }
+        }
 
         ExoPlayer.Builder(context, rendererFactory).build().apply {
             addListener(object : Player.Listener {