问题
I am trying to adapt the code found in ExtractDecodeEditEncodeMuxTest.java in order to extract audio and video from a mp4 recorded via Cordova's device.capture.captureVideo, decode the audio, edit the decoded audio samples, encode the audio, and mux the audio back with the video and save as an mp4 again.
My first attempt is simply to extract, decode, encode and mux audio without trying to edit any of the audio samples - if I can do this I am fairly certain that I can edit the decoded samples as desired. I don't need to edit the video, so I assume I can simply use MediaExtractor to extract and mux the video track.
However, the problem I am having is that I cannot seem to get the audio decoding/encoding process right. What keeps happening is that the muxer creates the mp4 from the extracted video track and the extracted -> decoded -> encoded audio track, but while the video plays fine, the audio starts with a short burst of noise, then what seems like the last couple seconds of audio data playing normally (but at the beginning of the video), then silence for the rest of the video.
Some of the relevant fields:
private MediaFormat audioFormat;
private MediaFormat videoFormat;
private int videoTrackIndex = -1;
private int audioTrackIndex = -1;
private static final int MAX_BUFFER_SIZE = 256 * 1024;
// parameters for the audio encoder
private static final String OUTPUT_AUDIO_MIME_TYPE = "audio/mp4a-latm"; // Advanced Audio Coding
private static final int OUTPUT_AUDIO_CHANNEL_COUNT = 2; // Must match the input stream. not using this, getting from input format
private static final int OUTPUT_AUDIO_BIT_RATE = 128 * 1024;
private static final int OUTPUT_AUDIO_AAC_PROFILE = MediaCodecInfo.CodecProfileLevel.AACObjectHE; //not using this, getting from input format
private static final int OUTPUT_AUDIO_SAMPLE_RATE_HZ = 44100; // Must match the input stream
private static final String TAG = "vvsLog";
private static final Boolean DEBUG = false;
private static final Boolean INFO = true;
/** How long to wait for the next buffer to become available. */
private static final int TIMEOUT_USEC = 10000;
private String videoPath;
The code configuring the decoder, encoder and muxer:
MediaCodecInfo audioCodecInfo = selectCodec(OUTPUT_AUDIO_MIME_TYPE);
if (audioCodecInfo == null) {
// Don't fail CTS if they don't have an AAC codec (not here, anyway).
Log.e(TAG, "Unable to find an appropriate codec for " + OUTPUT_AUDIO_MIME_TYPE);
return;
}
MediaExtractor videoExtractor = null;
MediaExtractor audioExtractor = null;
MediaCodec audioDecoder = null;
MediaCodec audioEncoder = null;
MediaMuxer muxer = null;
try {
/**
* Video
* just need to configure the extractor, no codec processing required
*/
videoExtractor = createExtractor(originalAssetPath);
String vidMimeStartsWith = "video/";
int videoInputTrack = getAndSelectTrackIndex(videoExtractor, vidMimeStartsWith);
videoFormat = videoExtractor.getTrackFormat(videoInputTrack);
/**
* Audio
* needs an extractor plus an audio decoder and encoder
*/
audioExtractor = createExtractor(originalAssetPath);
String audMimeStartsWith = "audio/";
int audioInputTrack = getAndSelectTrackIndex(audioExtractor, audMimeStartsWith);
audioFormat = audioExtractor.getTrackFormat(audioInputTrack);
audioFormat.setInteger(MediaFormat.KEY_SAMPLE_RATE,OUTPUT_AUDIO_SAMPLE_RATE_HZ);
MediaFormat outputAudioFormat = MediaFormat.createAudioFormat(OUTPUT_AUDIO_MIME_TYPE,
audioFormat.getInteger(MediaFormat.KEY_SAMPLE_RATE),
audioFormat.getInteger(MediaFormat.KEY_CHANNEL_COUNT));
outputAudioFormat.setInteger(MediaFormat.KEY_AAC_PROFILE, audioFormat.getInteger(MediaFormat.KEY_AAC_PROFILE));
outputAudioFormat.setInteger(MediaFormat.KEY_BIT_RATE, OUTPUT_AUDIO_BIT_RATE);
// Create a MediaCodec for the decoder, based on the extractor's format, configure and start it.
audioDecoder = createAudioDecoder(audioFormat);
// Create a MediaCodec for the desired codec, then configure it as an encoder and start it.
audioEncoder = createAudioEncoder(audioCodecInfo, outputAudioFormat);
//create muxer to overwrite original asset path
muxer = createMuxer(originalAssetPath);
//add the video and audio tracks
/**
* need to wait to add the audio track until after the first encoder output buffer is created
* since the encoder changes the MediaFormat at that time
* and the muxer needs the correct format, including the correct Coded Specific Data (CSD) ByteBuffer
*/
doExtractDecodeEditEncodeMux(
videoExtractor,
audioExtractor,
audioDecoder,
audioEncoder,
muxer);
}
The monster doExtractDecodeEditEncodeMux method:
private void doExtractDecodeEditEncodeMux(
MediaExtractor videoExtractor,
MediaExtractor audioExtractor,
MediaCodec audioDecoder,
MediaCodec audioEncoder,
MediaMuxer muxer) {
ByteBuffer videoInputBuffer = ByteBuffer.allocate(MAX_BUFFER_SIZE);
MediaCodec.BufferInfo videoBufferInfo = new MediaCodec.BufferInfo();
ByteBuffer[] audioDecoderInputBuffers = null;
ByteBuffer[] audioDecoderOutputBuffers = null;
ByteBuffer[] audioEncoderInputBuffers = null;
ByteBuffer[] audioEncoderOutputBuffers = null;
MediaCodec.BufferInfo audioDecoderOutputBufferInfo = null;
MediaCodec.BufferInfo audioEncoderOutputBufferInfo = null;
audioDecoderInputBuffers = audioDecoder.getInputBuffers();
audioDecoderOutputBuffers = audioDecoder.getOutputBuffers();
audioEncoderInputBuffers = audioEncoder.getInputBuffers();
audioEncoderOutputBuffers = audioEncoder.getOutputBuffers();
audioDecoderOutputBufferInfo = new MediaCodec.BufferInfo();
audioEncoderOutputBufferInfo = new MediaCodec.BufferInfo();
/**
* sanity checks
*/
//frames
int videoExtractedFrameCount = 0;
int audioExtractedFrameCount = 0;
int audioDecodedFrameCount = 0;
int audioEncodedFrameCount = 0;
//times
long lastPresentationTimeVideoExtractor = 0;
long lastPresentationTimeAudioExtractor = 0;
long lastPresentationTimeAudioDecoder = 0;
long lastPresentationTimeAudioEncoder = 0;
// We will get these from the decoders when notified of a format change.
MediaFormat decoderOutputAudioFormat = null;
// We will get these from the encoders when notified of a format change.
MediaFormat encoderOutputAudioFormat = null;
// We will determine these once we have the output format.
int outputAudioTrack = -1;
// Whether things are done on the video side.
boolean videoExtractorDone = false;
// Whether things are done on the audio side.
boolean audioExtractorDone = false;
boolean audioDecoderDone = false;
boolean audioEncoderDone = false;
// The audio decoder output buffer to process, -1 if none.
int pendingAudioDecoderOutputBufferIndex = -1;
boolean muxing = false;
/**
* need to wait to add the audio track until after the first encoder output buffer is created
* since the encoder changes the MediaFormat at that time
* and the muxer needs the correct format, including the correct Coded Specific Data (CSD) ByteBuffer
* muxer.start();
* muxing = true;
*/
MediaMetadataRetriever retrieverTest = new MediaMetadataRetriever();
retrieverTest.setDataSource(videoPath);
String degreesStr = retrieverTest.extractMetadata(MediaMetadataRetriever.METADATA_KEY_VIDEO_ROTATION);
if (degreesStr != null) {
Integer degrees = Integer.parseInt(degreesStr);
if (degrees >= 0) {
muxer.setOrientationHint(degrees);
}
}
while (!videoExtractorDone || !audioEncoderDone) {
if (INFO) {
Log.d(TAG, String.format("ex:%d at %d | de:%d at %d | en:%d at %d ",
audioExtractedFrameCount, lastPresentationTimeAudioExtractor,
audioDecodedFrameCount, lastPresentationTimeAudioDecoder,
audioEncodedFrameCount, lastPresentationTimeAudioEncoder
));
}
/**
* Extract and mux video
*/
while (!videoExtractorDone && muxing) {
try {
videoBufferInfo.size = videoExtractor.readSampleData(videoInputBuffer, 0);
} catch (Exception e) {
e.printStackTrace();
}
if (videoBufferInfo.size < 0) {
videoBufferInfo.size = 0;
videoExtractorDone = true;
} else {
videoBufferInfo.presentationTimeUs = videoExtractor.getSampleTime();
lastPresentationTimeVideoExtractor = videoBufferInfo.presentationTimeUs;
videoBufferInfo.flags = videoExtractor.getSampleFlags();
muxer.writeSampleData(videoTrackIndex, videoInputBuffer, videoBufferInfo);
videoExtractor.advance();
videoExtractedFrameCount++;
}
}
/**
* Extract, decode, watermark, encode and mux audio
*/
/** Extract audio from file and feed to decoder. **/
while (!audioExtractorDone && (encoderOutputAudioFormat == null || muxing)) {
int decoderInputBufferIndex = audioDecoder.dequeueInputBuffer(TIMEOUT_USEC);
if (decoderInputBufferIndex == MediaCodec.INFO_TRY_AGAIN_LATER) {
break;
}
if (DEBUG) {
Log.d(TAG, "audio decoder: returned input buffer: " + decoderInputBufferIndex);
}
ByteBuffer decoderInputBuffer = audioDecoderInputBuffers[decoderInputBufferIndex];
int size = audioExtractor.readSampleData(decoderInputBuffer, 0);
long presentationTime = audioExtractor.getSampleTime();
lastPresentationTimeAudioExtractor = presentationTime;
if (DEBUG) {
Log.d(TAG, "audio extractor: returned buffer of size " + size);
Log.d(TAG, "audio extractor: returned buffer for time " + presentationTime);
}
if (size >= 0) {
audioDecoder.queueInputBuffer(
decoderInputBufferIndex,
0,
size,
presentationTime,
audioExtractor.getSampleFlags());
}
audioExtractorDone = !audioExtractor.advance();
if (audioExtractorDone) {
if (DEBUG) Log.d(TAG, "audio extractor: EOS");
audioDecoder.queueInputBuffer(
decoderInputBufferIndex,
0,
0,
0,
MediaCodec.BUFFER_FLAG_END_OF_STREAM);
}
audioExtractedFrameCount++;
// We extracted a frame, let's try something else next.
break;
}
/**
* Poll output frames from the audio decoder.
* Do not poll if we already have a pending buffer to feed to the encoder.
*/
while (!audioDecoderDone && pendingAudioDecoderOutputBufferIndex == -1 && (encoderOutputAudioFormat == null || muxing)) {
int decoderOutputBufferIndex =
audioDecoder.dequeueOutputBuffer(
audioDecoderOutputBufferInfo, TIMEOUT_USEC);
if (decoderOutputBufferIndex == MediaCodec.INFO_TRY_AGAIN_LATER) {
if (DEBUG) Log.d(TAG, "no audio decoder output buffer");
break;
}
if (decoderOutputBufferIndex == MediaCodec.INFO_OUTPUT_BUFFERS_CHANGED) {
if (DEBUG) Log.d(TAG, "audio decoder: output buffers changed");
audioDecoderOutputBuffers = audioDecoder.getOutputBuffers();
break;
}
if (decoderOutputBufferIndex == MediaCodec.INFO_OUTPUT_FORMAT_CHANGED) {
decoderOutputAudioFormat = audioDecoder.getOutputFormat();
if (DEBUG) {
Log.d(TAG, "audio decoder: output format changed: "
+ decoderOutputAudioFormat);
}
break;
}
if (DEBUG) {
Log.d(TAG, "audio decoder: returned output buffer: "
+ decoderOutputBufferIndex);
}
if (DEBUG) {
Log.d(TAG, "audio decoder: returned buffer of size "
+ audioDecoderOutputBufferInfo.size);
}
ByteBuffer decoderOutputBuffer =
audioDecoderOutputBuffers[decoderOutputBufferIndex];
if ((audioDecoderOutputBufferInfo.flags & MediaCodec.BUFFER_FLAG_CODEC_CONFIG)
!= 0) {
if (DEBUG) Log.d(TAG, "audio decoder: codec config buffer");
audioDecoder.releaseOutputBuffer(decoderOutputBufferIndex, false);
break;
}
if (DEBUG) {
Log.d(TAG, "audio decoder: returned buffer for time "
+ audioDecoderOutputBufferInfo.presentationTimeUs);
}
if (DEBUG) {
Log.d(TAG, "audio decoder: output buffer is now pending: "
+ pendingAudioDecoderOutputBufferIndex);
}
pendingAudioDecoderOutputBufferIndex = decoderOutputBufferIndex;
audioDecodedFrameCount++;
// We extracted a pending frame, let's try something else next.
break;
}
// Feed the pending decoded audio buffer to the audio encoder.
while (pendingAudioDecoderOutputBufferIndex != -1) {
if (DEBUG) {
Log.d(TAG, "audio decoder: attempting to process pending buffer: "
+ pendingAudioDecoderOutputBufferIndex);
}
int encoderInputBufferIndex = audioEncoder.dequeueInputBuffer(TIMEOUT_USEC);
if (encoderInputBufferIndex == MediaCodec.INFO_TRY_AGAIN_LATER) {
if (DEBUG) Log.d(TAG, "no audio encoder input buffer");
break;
}
if (DEBUG) {
Log.d(TAG, "audio encoder: returned input buffer: " + encoderInputBufferIndex);
}
ByteBuffer encoderInputBuffer = audioEncoderInputBuffers[encoderInputBufferIndex];
int size = audioDecoderOutputBufferInfo.size;
long presentationTime = audioDecoderOutputBufferInfo.presentationTimeUs;
lastPresentationTimeAudioDecoder = presentationTime;
if (DEBUG) {
Log.d(TAG, "audio decoder: processing pending buffer: "
+ pendingAudioDecoderOutputBufferIndex);
}
if (DEBUG) {
Log.d(TAG, "audio decoder: pending buffer of size " + size);
Log.d(TAG, "audio decoder: pending buffer for time " + presentationTime);
}
if (size >= 0) {
ByteBuffer decoderOutputBuffer =
audioDecoderOutputBuffers[pendingAudioDecoderOutputBufferIndex]
.duplicate();
decoderOutputBuffer.position(audioDecoderOutputBufferInfo.offset);
decoderOutputBuffer.limit(audioDecoderOutputBufferInfo.offset + size);
encoderInputBuffer.position(0);
encoderInputBuffer.put(decoderOutputBuffer);
audioEncoder.queueInputBuffer(
encoderInputBufferIndex,
0,
size,
presentationTime,
audioDecoderOutputBufferInfo.flags);
}
audioDecoder.releaseOutputBuffer(pendingAudioDecoderOutputBufferIndex, false);
pendingAudioDecoderOutputBufferIndex = -1;
if ((audioDecoderOutputBufferInfo.flags
& MediaCodec.BUFFER_FLAG_END_OF_STREAM) != 0) {
if (DEBUG) Log.d(TAG, "audio decoder: EOS");
audioDecoderDone = true;
}
// We enqueued a pending frame, let's try something else next.
break;
}
// Poll frames from the audio encoder and send them to the muxer.
while (!audioEncoderDone && (encoderOutputAudioFormat == null || muxing)) {
int encoderOutputBufferIndex = audioEncoder.dequeueOutputBuffer(
audioEncoderOutputBufferInfo, TIMEOUT_USEC);
if (encoderOutputBufferIndex == MediaCodec.INFO_TRY_AGAIN_LATER) {
if (DEBUG) Log.d(TAG, "no audio encoder output buffer");
break;
}
if (encoderOutputBufferIndex == MediaCodec.INFO_OUTPUT_BUFFERS_CHANGED) {
if (DEBUG) Log.d(TAG, "audio encoder: output buffers changed");
audioEncoderOutputBuffers = audioEncoder.getOutputBuffers();
break;
}
if (encoderOutputBufferIndex == MediaCodec.INFO_OUTPUT_FORMAT_CHANGED) {
encoderOutputAudioFormat = audioEncoder.getOutputFormat();
if (DEBUG) {
Log.d(TAG, "audio encoder: output format changed");
}
if (outputAudioTrack >= 0) {
Log.e(TAG,"audio encoder changed its output format again?");
}
break;
}
if (DEBUG) {
Log.d(TAG, "audio encoder: returned output buffer: "
+ encoderOutputBufferIndex);
Log.d(TAG, "audio encoder: returned buffer of size "
+ audioEncoderOutputBufferInfo.size);
}
ByteBuffer encoderOutputBuffer =
audioEncoderOutputBuffers[encoderOutputBufferIndex];
if ((audioEncoderOutputBufferInfo.flags & MediaCodec.BUFFER_FLAG_CODEC_CONFIG)
!= 0) {
if (DEBUG) Log.d(TAG, "audio encoder: codec config buffer");
// Simply ignore codec config buffers.
audioEncoder.releaseOutputBuffer(encoderOutputBufferIndex, false);
break;
}
if (DEBUG) {
Log.d(TAG, "audio encoder: returned buffer for time "
+ audioEncoderOutputBufferInfo.presentationTimeUs);
}
if (audioEncoderOutputBufferInfo.size != 0) {
lastPresentationTimeAudioEncoder = audioEncoderOutputBufferInfo.presentationTimeUs;
muxer.writeSampleData(
audioTrackIndex, encoderOutputBuffer, audioEncoderOutputBufferInfo);
}
if ((audioEncoderOutputBufferInfo.flags & MediaCodec.BUFFER_FLAG_END_OF_STREAM)
!= 0) {
if (DEBUG) Log.d(TAG, "audio encoder: EOS");
audioEncoderDone = true;
}
audioEncoder.releaseOutputBuffer(encoderOutputBufferIndex, false);
audioEncodedFrameCount++;
// We enqueued an encoded frame, let's try something else next.
break;
}
if (!muxing && (encoderOutputAudioFormat != null)) {
Log.d(TAG, "muxer: adding video track.");
videoTrackIndex = muxer.addTrack(videoFormat);
Log.d(TAG, "muxer: adding audio track.");
audioTrackIndex = muxer.addTrack(encoderOutputAudioFormat);
Log.d(TAG, "muxer: starting");
muxer.start();
muxing = true;
}
}
/**
* Done processing audio and video
*/
Log.d(TAG,"encoded and decoded audio frame counts should match. decoded:"+audioDecodedFrameCount+" encoded:"+audioEncodedFrameCount);
Log.d(TAG,"decoded frame count should be less than extracted frame coun. decoded:"+audioDecodedFrameCount+" extracted:"+audioExtractedFrameCount);
Log.d(TAG,"no audio frame should be pending "+pendingAudioDecoderOutputBufferIndex);
PluginResult result = new PluginResult(PluginResult.Status.OK, videoPath);
result.setKeepCallback(false);
callbackContext.sendPluginResult(result);
}
I am seeing this ACodec error for the first several hundred audio frames extracted:
11-25 20:49:58.497 9807-13101/com.vvs.VVS430011 E/ACodec﹕ OMXCodec::onEvent, OMX_ErrorStreamCorrupt
11-25 20:49:58.497 9807-13101/com.vvs.VVS430011 W/AHierarchicalStateMachine﹕ Warning message AMessage(what = 'omx ', target = 8) = {
int32_t type = 0
int32_t node = 7115
int32_t event = 1
int32_t data1 = -2147479541
int32_t data2 = 0
} unhandled in root state.
Here's a pastebin of the entire logcat, which includes some sanity check logs in the format of:
D/vvsLog﹕ ex:{extracted frame #} at {presentationTime} | de:{decoded frame #} at {presentationTime} | en:{encoded frame #} at {presentationTime}
The presentationTime of encoded and decoded frames seems to be incrementing too quickly while those OMX_ErrorStreamCorrupt messages are appearing. When they stop, the presentationTime for the decoded and encoded frames seems to return to "normal", and also seems to match up with the actual "good" audio I hear at the beginning of the video - the "good" audio being from the end of the original audio track.
I am hoping someone with a lot more experience with these low-level Android multimedia APIs than I have can help me understand why this is happening. Keep in mind I am well aware that this code is not optimized, running in separate threads, etc.. - I will refactor to clean things up once I have a working example of the basic extract->decode->edit->encode->mux process.
Thanks!
回答1:
Turns out the above code works fine - as long as you're not trying to mux the same file you're extracting, at the same time.
:-)
I had a previous version of this that extracted, then muxed tracks to the same file, and forgot to change that in this version.
This little method saved the day lol.
private String getMuxedAssetPath() {
String muxedAssetPath = Environment.getExternalStoragePublicDirectory(Environment.DIRECTORY_DCIM) + "/" + CAMERA_DIRECTORY + "/muxedAudioVideo.mp4";
File file = new File(muxedAssetPath);
if (!file.exists()) {
try {
file.createNewFile();
} catch (IOException e) {
e.printStackTrace();
muxedAssetPath = null;
}
}
return muxedAssetPath;
}
来源：https://stackoverflow.com/questions/33930260/android-extract-decode-encode-mux-audio