Related
Context
Hello !
I'm currently working on the development of a small library allowing to cut an h.264 video on any frame, but without re-encoding (transcoding) the whole video. The idea is to re-encode only the GOP on which we want to cut, and to rewrite (remux) directly the others GOP.
The avcut project (https://github.com/anyc/avcut) allows to do that, but requires a systematic decoding of each package, and seems to not work with the recent versions of ffmpeg from the tests I could do and from the recent feedbacks in the github issues.
As a beginner, I started from the code examples provided in the ffmpeg documentation, in particular: transcoding.c and remuxing.c.
Problem encountered
The problem I'm having is that I can't get both transcoding and remuxing to work properly at the same time. In particular, depending on the method I use to initialize the AVCodecParameters of the output video stream, transcoding works, or remuxing works:
avcodec_parameters_copy works well for remuxing
avcodec_parameters_from_context works well for transcoding
In case I choose avcodec_parameters_from_context, the transcoded GOP are correctly read by my video player (parole), but the remuxed packets are not read, and ffprobe does not show/detect them.
In case I choose avcodec_parameters_from_context, the remuxing GOP are correctly read by my video player, but the transcoding key_frame are bugged (I have the impression that the b-frame and p-frame are ok), and ffprobe -i return an error about the NAL of the key-frames:
[h264 # 0x55ec8a079300] sps_id 32 out of range
[h264 # 0x55ec8a079300] Invalid NAL unit size (1677727148 > 735).
[h264 # 0x55ec8a079300] missing picture in access unit with size 744
I suspect that the problem is related to the extradata of the packets. Through some experiments on the different attributes of the output AVCodecParameters, it seems that it is the extradata and extradata_size attributes that are responsible for the functioning of one method or the other.
Version
ffmpeg development branch retrieved on 2022-05-17 from https://github.com/FFmpeg/FFmpeg.
Compiled with --enable-libx264 --enable-gpl --enable-decoder=png --enable-encoder=png
Code
My code is written in c++ and is based on two classes: a class defining the parameters and methods on the input file (InputContexts) and a class defining them for the output file (OutputContexts). The code of these two classes is defined in the following files:
contexts.h
contexts.cpp
The code normally involved in the problem is the following:
stream initialization
int OutputContexts::init(const char* out_filename, InputContexts* input_contexts){
int ret;
int stream_index = 0;
avformat_alloc_output_context2(&ofmt_ctx, NULL, NULL, out_filename);
if (!ofmt_ctx) {
fprintf(stderr, "Could not create output context\n");
ret = AVERROR_UNKNOWN;
return ret;
}
av_dump_format(ofmt_ctx, 0, out_filename, 1);
encoders.resize(input_contexts->ifmt_ctx->nb_streams, nullptr);
codecs.resize(input_contexts->ifmt_ctx->nb_streams, nullptr);
// stream mapping
for (int i = 0; i < input_contexts->ifmt_ctx->nb_streams; i++) {
AVStream *out_stream;
AVStream *in_stream = input_contexts->ifmt_ctx->streams[i];
AVCodecContext* decoder_ctx = input_contexts->decoders[i];
// add new stream to output context
out_stream = avformat_new_stream(ofmt_ctx, NULL);
if (!out_stream) {
fprintf(stderr, "Failed allocating output stream\n");
ret = AVERROR_UNKNOWN;
return ret;
}
// from avcut blog
av_dict_copy(&out_stream->metadata, in_stream->metadata, 0);
out_stream->time_base = in_stream->time_base;
// encoder
if (decoder_ctx->codec_type == AVMEDIA_TYPE_VIDEO){
ret = prepare_encoder_video(i, input_contexts);
if (ret < 0){
fprintf(stderr, "Error while preparing encoder for stream #%u\n", i);
return ret;
}
// from avcut
out_stream->sample_aspect_ratio = in_stream->sample_aspect_ratio;
// works well for remuxing
ret = avcodec_parameters_copy(out_stream->codecpar, in_stream->codecpar);
if (ret < 0) {
fprintf(stderr, "Failed to copy codec parameters\n");
return ret;
}
// works well for transcoding
// ret = avcodec_parameters_from_context(out_stream->codecpar, encoders[i]);
// if (ret < 0) {
// av_log(NULL, AV_LOG_ERROR, "Failed to copy encoder parameters to output stream #%u\n", i);
// return ret;
// }
} else if (decoder_ctx->codec_type == AVMEDIA_TYPE_AUDIO) {
...
} else {
...
}
// TODO useful ???
// set current stream position to 0
// out_stream->codecpar->codec_tag = 0;
}
// opening output file in write mode with the ouput context
if (!(ofmt_ctx->oformat->flags & AVFMT_NOFILE)) {
ret = avio_open(&ofmt_ctx->pb, out_filename, AVIO_FLAG_WRITE);
if (ret < 0) {
fprintf(stderr, "Could not open output file '%s'", out_filename);
return ret;
}
}
// write headers from output context in output file
ret = avformat_write_header(ofmt_ctx, NULL);
if (ret < 0) {
fprintf(stderr, "Error occurred when opening output file\n");
return ret;
}
return ret;
}
AVCodecContext initialization for encoder
int OutputContexts::prepare_encoder_video(int stream_index, InputContexts* input_contexts){
int ret;
const AVCodec* encoder;
AVCodecContext* decoder_ctx = input_contexts->decoders[stream_index];
AVCodecContext* encoder_ctx;
if (video_index >= 0){
fprintf(stderr, "Impossible to mark stream #%u as video, stream #%u is already registered as video stream.\n",
stream_index, video_index);
return -1; //TODO change this value for correct error code
}
video_index = stream_index;
if(decoder_ctx->codec_id == AV_CODEC_ID_H264){
encoder = avcodec_find_encoder_by_name("libx264");
if (!encoder) {
av_log(NULL, AV_LOG_FATAL, "Encoder libx264 not found\n");
return AVERROR_INVALIDDATA;
}
fmt::print("Encoder libx264 will be used for stream {}.\n", stream_index);
} else {
std::string s = fmt::format("No video encoder found for the given codec_id: {}\n", avcodec_get_name(decoder_ctx->codec_id));
av_log(NULL, AV_LOG_FATAL, s.c_str());
return AVERROR_INVALIDDATA;
}
encoder_ctx = avcodec_alloc_context3(encoder);
if (!encoder_ctx) {
av_log(NULL, AV_LOG_FATAL, "Failed to allocate the encoder context\n");
return AVERROR(ENOMEM);
}
// from avcut
encoder_ctx->time_base = decoder_ctx->time_base;
encoder_ctx->ticks_per_frame = decoder_ctx->ticks_per_frame;
encoder_ctx->delay = decoder_ctx->delay;
encoder_ctx->width = decoder_ctx->width;
encoder_ctx->height = decoder_ctx->height;
encoder_ctx->pix_fmt = decoder_ctx->pix_fmt;
encoder_ctx->sample_aspect_ratio = decoder_ctx->sample_aspect_ratio;
encoder_ctx->color_primaries = decoder_ctx->color_primaries;
encoder_ctx->color_trc = decoder_ctx->color_trc;
encoder_ctx->colorspace = decoder_ctx->colorspace;
encoder_ctx->color_range = decoder_ctx->color_range;
encoder_ctx->chroma_sample_location = decoder_ctx->chroma_sample_location;
encoder_ctx->profile = decoder_ctx->profile;
encoder_ctx->level = decoder_ctx->level;
encoder_ctx->thread_count = 1; // spawning more threads causes avcodec_close to free threads multiple times
encoder_ctx->codec_tag = 0;
// correct values ???
encoder_ctx->qmin = 16;
encoder_ctx->qmax = 26;
encoder_ctx->max_qdiff = 4;
// end from avcut
// according to avcut, should not be set
// if (ofmt_ctx->oformat->flags & AVFMT_GLOBALHEADER){
// encoder_ctx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
// }
ret = avcodec_open2(encoder_ctx, encoder, NULL);
if (ret < 0) {
av_log(NULL, AV_LOG_ERROR, "Cannot open video encoder for stream #%u\n", stream_index);
return ret;
}
codecs[stream_index] = encoder;
encoders[stream_index] = encoder_ctx;
return ret;
}
Example
To illustrate my problem, I provide here a test code using the two classes that alternates between transcoding and remuxing at each key-frame encountered in the file using my classes.
trans_remux.cpp
To compile the code:
g++ -o trans_remux trans_remux.cpp contexts.cpp -D__STDC_CONSTANT_MACROS `pkg-config --libs libavfilter` -lfmt -g
Currently the code is using avcodec_parameters_copy (contexts.cpp:333), so it works well for remuxing. If you want to test the version with avcodec_parameters_from_context, pls comment from line 333 to 337 in contexts.cpp and uncomment from line 340 to 344 and recompile.
As it stands, my project correctly uses libavcodec to decode a video, where each frame is manipulated (it doesn't matter how) and output to a new video. I've cobbled this together from examples found online, and it works. The result is a perfect .mp4 of the manipulated frames, minus the audio.
My problem is, when I try to add an audio stream to the output container, I get a crash in mux.c that I can't explain. It's in static int compute_muxer_pkt_fields(AVFormatContext *s, AVStream *st, AVPacket *pkt). Where st->internal->priv_pts->val = pkt->dts; is attempted, priv_pts is nullptr.
I don't recall the version number, but this is from a November 4, 2020 ffmpeg build from git.
My MediaContentMgr is much bigger than what I have here. I'm stripping out everything to do with the frame manipulation, so if I'm missing anything, please let me know and I'll edit.
The code that, when added, triggers the nullptr exception, is called out inline
The .h:
#ifndef _API_EXAMPLE_H
#define _API_EXAMPLE_H
#include <glad/glad.h>
#include <GLFW/glfw3.h>
#include "glm/glm.hpp"
extern "C" {
#include <libavcodec/avcodec.h>
#include <libavformat/avformat.h>
#include <libavutil/avutil.h>
#include <libavutil/opt.h>
#include <libswscale/swscale.h>
}
#include "shader_s.h"
class MediaContainerMgr {
public:
MediaContainerMgr(const std::string& infile, const std::string& vert, const std::string& frag,
const glm::vec3* extents);
~MediaContainerMgr();
void render();
bool recording() { return m_recording; }
// Major thanks to "shi-yan" who helped make this possible:
// https://github.com/shi-yan/videosamples/blob/master/libavmp4encoding/main.cpp
bool init_video_output(const std::string& video_file_name, unsigned int width, unsigned int height);
bool output_video_frame(uint8_t* buf);
bool finalize_output();
private:
AVFormatContext* m_format_context;
AVCodec* m_video_codec;
AVCodec* m_audio_codec;
AVCodecParameters* m_video_codec_parameters;
AVCodecParameters* m_audio_codec_parameters;
AVCodecContext* m_codec_context;
AVFrame* m_frame;
AVPacket* m_packet;
uint32_t m_video_stream_index;
uint32_t m_audio_stream_index;
void init_rendering(const glm::vec3* extents);
int decode_packet();
// For writing the output video:
void free_output_assets();
bool m_recording;
AVOutputFormat* m_output_format;
AVFormatContext* m_output_format_context;
AVCodec* m_output_video_codec;
AVCodecContext* m_output_video_codec_context;
AVFrame* m_output_video_frame;
SwsContext* m_output_scale_context;
AVStream* m_output_video_stream;
AVCodec* m_output_audio_codec;
AVStream* m_output_audio_stream;
AVCodecContext* m_output_audio_codec_context;
};
#endif
And, the hellish .cpp:
#include <stdio.h>
#include <stdarg.h>
#include <stdlib.h>
#include <string.h>
#include <inttypes.h>
#include "media_container_manager.h"
MediaContainerMgr::MediaContainerMgr(const std::string& infile, const std::string& vert, const std::string& frag,
const glm::vec3* extents) :
m_video_stream_index(-1),
m_audio_stream_index(-1),
m_recording(false),
m_output_format(nullptr),
m_output_format_context(nullptr),
m_output_video_codec(nullptr),
m_output_video_codec_context(nullptr),
m_output_video_frame(nullptr),
m_output_scale_context(nullptr),
m_output_video_stream(nullptr)
{
// AVFormatContext holds header info from the format specified in the container:
m_format_context = avformat_alloc_context();
if (!m_format_context) {
throw "ERROR could not allocate memory for Format Context";
}
// open the file and read its header. Codecs are not opened here.
if (avformat_open_input(&m_format_context, infile.c_str(), NULL, NULL) != 0) {
throw "ERROR could not open input file for reading";
}
printf("format %s, duration %lldus, bit_rate %lld\n", m_format_context->iformat->name, m_format_context->duration, m_format_context->bit_rate);
//read avPackets (?) from the avFormat (?) to get stream info. This populates format_context->streams.
if (avformat_find_stream_info(m_format_context, NULL) < 0) {
throw "ERROR could not get stream info";
}
for (unsigned int i = 0; i < m_format_context->nb_streams; i++) {
AVCodecParameters* local_codec_parameters = NULL;
local_codec_parameters = m_format_context->streams[i]->codecpar;
printf("AVStream->time base before open coded %d/%d\n", m_format_context->streams[i]->time_base.num, m_format_context->streams[i]->time_base.den);
printf("AVStream->r_frame_rate before open coded %d/%d\n", m_format_context->streams[i]->r_frame_rate.num, m_format_context->streams[i]->r_frame_rate.den);
printf("AVStream->start_time %" PRId64 "\n", m_format_context->streams[i]->start_time);
printf("AVStream->duration %" PRId64 "\n", m_format_context->streams[i]->duration);
printf("duration(s): %lf\n", (float)m_format_context->streams[i]->duration / m_format_context->streams[i]->time_base.den * m_format_context->streams[i]->time_base.num);
AVCodec* local_codec = NULL;
local_codec = avcodec_find_decoder(local_codec_parameters->codec_id);
if (local_codec == NULL) {
throw "ERROR unsupported codec!";
}
if (local_codec_parameters->codec_type == AVMEDIA_TYPE_VIDEO) {
if (m_video_stream_index == -1) {
m_video_stream_index = i;
m_video_codec = local_codec;
m_video_codec_parameters = local_codec_parameters;
}
m_height = local_codec_parameters->height;
m_width = local_codec_parameters->width;
printf("Video Codec: resolution %dx%d\n", m_width, m_height);
}
else if (local_codec_parameters->codec_type == AVMEDIA_TYPE_AUDIO) {
if (m_audio_stream_index == -1) {
m_audio_stream_index = i;
m_audio_codec = local_codec;
m_audio_codec_parameters = local_codec_parameters;
}
printf("Audio Codec: %d channels, sample rate %d\n", local_codec_parameters->channels, local_codec_parameters->sample_rate);
}
printf("\tCodec %s ID %d bit_rate %lld\n", local_codec->name, local_codec->id, local_codec_parameters->bit_rate);
}
m_codec_context = avcodec_alloc_context3(m_video_codec);
if (!m_codec_context) {
throw "ERROR failed to allocate memory for AVCodecContext";
}
if (avcodec_parameters_to_context(m_codec_context, m_video_codec_parameters) < 0) {
throw "ERROR failed to copy codec params to codec context";
}
if (avcodec_open2(m_codec_context, m_video_codec, NULL) < 0) {
throw "ERROR avcodec_open2 failed to open codec";
}
m_frame = av_frame_alloc();
if (!m_frame) {
throw "ERROR failed to allocate AVFrame memory";
}
m_packet = av_packet_alloc();
if (!m_packet) {
throw "ERROR failed to allocate AVPacket memory";
}
}
MediaContainerMgr::~MediaContainerMgr() {
avformat_close_input(&m_format_context);
av_packet_free(&m_packet);
av_frame_free(&m_frame);
avcodec_free_context(&m_codec_context);
glDeleteVertexArrays(1, &m_VAO);
glDeleteBuffers(1, &m_VBO);
}
bool MediaContainerMgr::advance_frame() {
while (true) {
if (av_read_frame(m_format_context, m_packet) < 0) {
// Do we actually need to unref the packet if it failed?
av_packet_unref(m_packet);
continue;
//return false;
}
else {
if (m_packet->stream_index == m_video_stream_index) {
//printf("AVPacket->pts %" PRId64 "\n", m_packet->pts);
int response = decode_packet();
av_packet_unref(m_packet);
if (response != 0) {
continue;
//return false;
}
return true;
}
else {
printf("m_packet->stream_index: %d\n", m_packet->stream_index);
printf(" m_packet->pts: %lld\n", m_packet->pts);
printf(" mpacket->size: %d\n", m_packet->size);
if (m_recording) {
int err = 0;
//err = avcodec_send_packet(m_output_video_codec_context, m_packet);
printf(" encoding error: %d\n", err);
}
}
}
// We're done with the packet (it's been unpacked to a frame), so deallocate & reset to defaults:
/*
if (m_frame == NULL)
return false;
if (m_frame->data[0] == NULL || m_frame->data[1] == NULL || m_frame->data[2] == NULL) {
printf("WARNING: null frame data");
continue;
}
*/
}
}
int MediaContainerMgr::decode_packet() {
// Supply raw packet data as input to a decoder
// https://ffmpeg.org/doxygen/trunk/group__lavc__decoding.html#ga58bc4bf1e0ac59e27362597e467efff3
int response = avcodec_send_packet(m_codec_context, m_packet);
if (response < 0) {
char buf[256];
av_strerror(response, buf, 256);
printf("Error while receiving a frame from the decoder: %s\n", buf);
return response;
}
// Return decoded output data (into a frame) from a decoder
// https://ffmpeg.org/doxygen/trunk/group__lavc__decoding.html#ga11e6542c4e66d3028668788a1a74217c
response = avcodec_receive_frame(m_codec_context, m_frame);
if (response == AVERROR(EAGAIN) || response == AVERROR_EOF) {
return response;
} else if (response < 0) {
char buf[256];
av_strerror(response, buf, 256);
printf("Error while receiving a frame from the decoder: %s\n", buf);
return response;
} else {
printf(
"Frame %d (type=%c, size=%d bytes) pts %lld key_frame %d [DTS %d]\n",
m_codec_context->frame_number,
av_get_picture_type_char(m_frame->pict_type),
m_frame->pkt_size,
m_frame->pts,
m_frame->key_frame,
m_frame->coded_picture_number
);
}
return 0;
}
bool MediaContainerMgr::init_video_output(const std::string& video_file_name, unsigned int width, unsigned int height) {
if (m_recording)
return true;
m_recording = true;
advance_to(0L); // I've deleted the implmentation. Just seeks to beginning of vid. Works fine.
if (!(m_output_format = av_guess_format(nullptr, video_file_name.c_str(), nullptr))) {
printf("Cannot guess output format.\n");
return false;
}
int err = avformat_alloc_output_context2(&m_output_format_context, m_output_format, nullptr, video_file_name.c_str());
if (err < 0) {
printf("Failed to allocate output context.\n");
return false;
}
//TODO(P0): Break out the video and audio inits into their own methods.
m_output_video_codec = avcodec_find_encoder(m_output_format->video_codec);
if (!m_output_video_codec) {
printf("Failed to create video codec.\n");
return false;
}
m_output_video_stream = avformat_new_stream(m_output_format_context, m_output_video_codec);
if (!m_output_video_stream) {
printf("Failed to find video format.\n");
return false;
}
m_output_video_codec_context = avcodec_alloc_context3(m_output_video_codec);
if (!m_output_video_codec_context) {
printf("Failed to create video codec context.\n");
return(false);
}
m_output_video_stream->codecpar->codec_id = m_output_format->video_codec;
m_output_video_stream->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
m_output_video_stream->codecpar->width = width;
m_output_video_stream->codecpar->height = height;
m_output_video_stream->codecpar->format = AV_PIX_FMT_YUV420P;
// Use the same bit rate as the input stream.
m_output_video_stream->codecpar->bit_rate = m_format_context->streams[m_video_stream_index]->codecpar->bit_rate;
m_output_video_stream->avg_frame_rate = m_format_context->streams[m_video_stream_index]->avg_frame_rate;
avcodec_parameters_to_context(m_output_video_codec_context, m_output_video_stream->codecpar);
m_output_video_codec_context->time_base = m_format_context->streams[m_video_stream_index]->time_base;
//TODO(P1): Set these to match the input stream?
m_output_video_codec_context->max_b_frames = 2;
m_output_video_codec_context->gop_size = 12;
m_output_video_codec_context->framerate = m_format_context->streams[m_video_stream_index]->r_frame_rate;
//m_output_codec_context->refcounted_frames = 0;
if (m_output_video_stream->codecpar->codec_id == AV_CODEC_ID_H264) {
av_opt_set(m_output_video_codec_context, "preset", "ultrafast", 0);
} else if (m_output_video_stream->codecpar->codec_id == AV_CODEC_ID_H265) {
av_opt_set(m_output_video_codec_context, "preset", "ultrafast", 0);
} else {
av_opt_set_int(m_output_video_codec_context, "lossless", 1, 0);
}
avcodec_parameters_from_context(m_output_video_stream->codecpar, m_output_video_codec_context);
m_output_audio_codec = avcodec_find_encoder(m_output_format->audio_codec);
if (!m_output_audio_codec) {
printf("Failed to create audio codec.\n");
return false;
}
I've commented out all of the audio stream init beyond this next line, because this is where
the trouble begins. Creating this output stream causes the null reference I mentioned. If I
uncomment everything below here, I still get the null deref. If I comment out this line, the
deref exception vanishes. (IOW, I commented out more and more code until I found that this
was the trigger that caused the problem.)
I assume that there's something I'm doing wrong in the rest of the commented out code, that,
when fixed, will fix the nullptr and give me a working audio stream.
m_output_audio_stream = avformat_new_stream(m_output_format_context, m_output_audio_codec);
if (!m_output_audio_stream) {
printf("Failed to find audio format.\n");
return false;
}
/*
m_output_audio_codec_context = avcodec_alloc_context3(m_output_audio_codec);
if (!m_output_audio_codec_context) {
printf("Failed to create audio codec context.\n");
return(false);
}
m_output_audio_stream->codecpar->codec_id = m_output_format->audio_codec;
m_output_audio_stream->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
m_output_audio_stream->codecpar->format = m_format_context->streams[m_audio_stream_index]->codecpar->format;
m_output_audio_stream->codecpar->bit_rate = m_format_context->streams[m_audio_stream_index]->codecpar->bit_rate;
m_output_audio_stream->avg_frame_rate = m_format_context->streams[m_audio_stream_index]->avg_frame_rate;
avcodec_parameters_to_context(m_output_audio_codec_context, m_output_audio_stream->codecpar);
m_output_audio_codec_context->time_base = m_format_context->streams[m_audio_stream_index]->time_base;
*/
//TODO(P2): Free assets that have been allocated.
err = avcodec_open2(m_output_video_codec_context, m_output_video_codec, nullptr);
if (err < 0) {
printf("Failed to open codec.\n");
return false;
}
if (!(m_output_format->flags & AVFMT_NOFILE)) {
err = avio_open(&m_output_format_context->pb, video_file_name.c_str(), AVIO_FLAG_WRITE);
if (err < 0) {
printf("Failed to open output file.");
return false;
}
}
err = avformat_write_header(m_output_format_context, NULL);
if (err < 0) {
printf("Failed to write header.\n");
return false;
}
av_dump_format(m_output_format_context, 0, video_file_name.c_str(), 1);
return true;
}
//TODO(P2): make this a member. (Thanks to https://emvlo.wordpress.com/2016/03/10/sws_scale/)
void PrepareFlipFrameJ420(AVFrame* pFrame) {
for (int i = 0; i < 4; i++) {
if (i)
pFrame->data[i] += pFrame->linesize[i] * ((pFrame->height >> 1) - 1);
else
pFrame->data[i] += pFrame->linesize[i] * (pFrame->height - 1);
pFrame->linesize[i] = -pFrame->linesize[i];
}
}
This is where we take an altered frame and write it to the output container. This works fine
as long as we haven't set up an audio stream in the output container.
bool MediaContainerMgr::output_video_frame(uint8_t* buf) {
int err;
if (!m_output_video_frame) {
m_output_video_frame = av_frame_alloc();
m_output_video_frame->format = AV_PIX_FMT_YUV420P;
m_output_video_frame->width = m_output_video_codec_context->width;
m_output_video_frame->height = m_output_video_codec_context->height;
err = av_frame_get_buffer(m_output_video_frame, 32);
if (err < 0) {
printf("Failed to allocate output frame.\n");
return false;
}
}
if (!m_output_scale_context) {
m_output_scale_context = sws_getContext(m_output_video_codec_context->width, m_output_video_codec_context->height,
AV_PIX_FMT_RGB24,
m_output_video_codec_context->width, m_output_video_codec_context->height,
AV_PIX_FMT_YUV420P, SWS_BICUBIC, nullptr, nullptr, nullptr);
}
int inLinesize[1] = { 3 * m_output_video_codec_context->width };
sws_scale(m_output_scale_context, (const uint8_t* const*)&buf, inLinesize, 0, m_output_video_codec_context->height,
m_output_video_frame->data, m_output_video_frame->linesize);
PrepareFlipFrameJ420(m_output_video_frame);
//TODO(P0): Switch m_frame to be m_input_video_frame so I don't end up using the presentation timestamp from
// an audio frame if I threadify the frame reading.
m_output_video_frame->pts = m_frame->pts;
printf("Output PTS: %d, time_base: %d/%d\n", m_output_video_frame->pts,
m_output_video_codec_context->time_base.num, m_output_video_codec_context->time_base.den);
err = avcodec_send_frame(m_output_video_codec_context, m_output_video_frame);
if (err < 0) {
printf(" ERROR sending new video frame output: ");
switch (err) {
case AVERROR(EAGAIN):
printf("AVERROR(EAGAIN): %d\n", err);
break;
case AVERROR_EOF:
printf("AVERROR_EOF: %d\n", err);
break;
case AVERROR(EINVAL):
printf("AVERROR(EINVAL): %d\n", err);
break;
case AVERROR(ENOMEM):
printf("AVERROR(ENOMEM): %d\n", err);
break;
}
return false;
}
AVPacket pkt;
av_init_packet(&pkt);
pkt.data = nullptr;
pkt.size = 0;
pkt.flags |= AV_PKT_FLAG_KEY;
int ret = 0;
if ((ret = avcodec_receive_packet(m_output_video_codec_context, &pkt)) == 0) {
static int counter = 0;
printf("pkt.key: 0x%08x, pkt.size: %d, counter:\n", pkt.flags & AV_PKT_FLAG_KEY, pkt.size, counter++);
uint8_t* size = ((uint8_t*)pkt.data);
printf("sizes: %d %d %d %d %d %d %d %d %d\n", size[0], size[1], size[2], size[2], size[3], size[4], size[5], size[6], size[7]);
av_interleaved_write_frame(m_output_format_context, &pkt);
}
printf("push: %d\n", ret);
av_packet_unref(&pkt);
return true;
}
bool MediaContainerMgr::finalize_output() {
if (!m_recording)
return true;
AVPacket pkt;
av_init_packet(&pkt);
pkt.data = nullptr;
pkt.size = 0;
for (;;) {
avcodec_send_frame(m_output_video_codec_context, nullptr);
if (avcodec_receive_packet(m_output_video_codec_context, &pkt) == 0) {
av_interleaved_write_frame(m_output_format_context, &pkt);
printf("final push:\n");
} else {
break;
}
}
av_packet_unref(&pkt);
av_write_trailer(m_output_format_context);
if (!(m_output_format->flags & AVFMT_NOFILE)) {
int err = avio_close(m_output_format_context->pb);
if (err < 0) {
printf("Failed to close file. err: %d\n", err);
return false;
}
}
return true;
}
EDIT
The call stack on the crash (which I should have included in the original question):
avformat-58.dll!compute_muxer_pkt_fields(AVFormatContext * s, AVStream * st, AVPacket * pkt) Line 630 C
avformat-58.dll!write_packet_common(AVFormatContext * s, AVStream * st, AVPacket * pkt, int interleaved) Line 1122 C
avformat-58.dll!write_packets_common(AVFormatContext * s, AVPacket * pkt, int interleaved) Line 1186 C
avformat-58.dll!av_interleaved_write_frame(AVFormatContext * s, AVPacket * pkt) Line 1241 C
CamBot.exe!MediaContainerMgr::output_video_frame(unsigned char * buf) Line 553 C++
CamBot.exe!main() Line 240 C++
If I move the call to avformat_write_header so it's immediately before the audio stream initialization, I still get a crash, but in a different place. The crash happens on line 6459 of movenc.c, where we have:
/* Non-seekable output is ok if using fragmentation. If ism_lookahead
* is enabled, we don't support non-seekable output at all. */
if (!(s->pb->seekable & AVIO_SEEKABLE_NORMAL) && // CRASH IS HERE
(!(mov->flags & FF_MOV_FLAG_FRAGMENT) || mov->ism_lookahead)) {
av_log(s, AV_LOG_ERROR, "muxer does not support non seekable output\n");
return AVERROR(EINVAL);
}
The exception is a nullptr exception, where s->pb is NULL. The call stack is:
avformat-58.dll!mov_init(AVFormatContext * s) Line 6459 C
avformat-58.dll!init_muxer(AVFormatContext * s, AVDictionary * * options) Line 407 C
[Inline Frame] avformat-58.dll!avformat_init_output(AVFormatContext *) Line 489 C
avformat-58.dll!avformat_write_header(AVFormatContext * s, AVDictionary * * options) Line 512 C
CamBot.exe!MediaContainerMgr::init_video_output(const std::string & video_file_name, unsigned int width, unsigned int height) Line 424 C++
CamBot.exe!main() Line 183 C++
Please note that you should always try to provide a self-contained minimal working example to make it easier for others to help. With the actual code, the matching FFmpeg version, and an input video that triggers the segmentation fault (to be sure), the issue would be a matter of analyzing the control flow to identify why st->internal->priv_pts was not allocated. Without the full scenario, I have to report to making assumptions that may or may not correspond to your actual code.
Based on your description, I attempted to reproduce the issue by cloning https://github.com/FFmpeg/FFmpeg.git and creating a new branch from commit b52e0d95 (November 4, 2020) to approximate your FFmpeg version.
I recreated your scenario using the provided code snippets by
including the avformat_new_stream() call for the audio stream
keeping the remaining audio initialization commented out
including the original avformat_write_header() call site (unchanged order)
With that scenario, the video write with MP4 video/audio input fails in avformat_write_header():
[mp4 # 0x2b39f40] sample rate not set 0
The call stack of the error location:
#0 0x00007ffff75253d7 in raise () from /lib64/libc.so.6
#1 0x00007ffff7526ac8 in abort () from /lib64/libc.so.6
#2 0x000000000094feca in init_muxer (s=0x2b39f40, options=0x0) at libavformat/mux.c:309
#3 0x00000000009508f4 in avformat_init_output (s=0x2b39f40, options=0x0) at libavformat/mux.c:490
#4 0x0000000000950a10 in avformat_write_header (s=0x2b39f40, options=0x0) at libavformat/mux.c:514
[...]
In init_muxer(), the sample rate in the stream parameters is checked unconditionally:
case AVMEDIA_TYPE_AUDIO:
if (par->sample_rate <= 0) {
av_log(s, AV_LOG_ERROR, "sample rate not set %d\n", par->sample_rate); abort();
ret = AVERROR(EINVAL);
goto fail;
}
That condition has been in effect since 2014-06-18 at the very least (didn't go back any further) and still exists. With a version from November 2020, the check must be active and the parameter must be set accordingly.
If I uncomment the remaining audio initialization, the situation remains unchanged (as expected). So, satisfy the condition, I added the missing parameter as follows:
m_output_audio_stream->codecpar->sample_rate =
m_format_context->streams[m_audio_stream_index]->codecpar->sample_rate;
With that, the check succeeds, avformat_write_header() succeeds, and the actual video write succeeds.
As you indicated in your question, the segmentation fault is caused by st->internal->priv_pts being NULL at this location:
#0 0x00000000009516db in compute_muxer_pkt_fields (s=0x2b39f40, st=0x2b3a580, pkt=0x7fffffffe2d0) at libavformat/mux.c:632
#1 0x0000000000953128 in write_packet_common (s=0x2b39f40, st=0x2b3a580, pkt=0x7fffffffe2d0, interleaved=1) at libavformat/mux.c:1125
#2 0x0000000000953473 in write_packets_common (s=0x2b39f40, pkt=0x7fffffffe2d0, interleaved=1) at libavformat/mux.c:1188
#3 0x0000000000953634 in av_interleaved_write_frame (s=0x2b39f40, pkt=0x7fffffffe2d0) at libavformat/mux.c:1243
[...]
In the FFmpeg code base, the allocation of priv_pts is handled by init_pts() for all streams referenced by the context. init_pts() has two call sites:
libavformat/mux.c:496:
if (s->oformat->init && ret) {
if ((ret = init_pts(s)) < 0)
return ret;
return AVSTREAM_INIT_IN_INIT_OUTPUT;
}
libavformat/mux.c:530:
if (!s->internal->streams_initialized) {
if ((ret = init_pts(s)) < 0)
goto fail;
}
In both cases, the calls are triggered by avformat_write_header() (indirectly via avformat_init_output() for the first, directly for the second). According to control flow analysis, there's no success case that would leave priv_pts unallocated.
Considering a high probability that our versions of FFmpeg are compatible in terms of behavior, I have to assume that 1) the sample rate must be provided for audio streams and 1) priv_pts is always allocated by avformat_write_header() in the absence of errors. Therefore, two possible root causes come to mind:
Your stream is not an audio stream (unlikely; the type is based on the codec, which in turn is based on the output file extension - assuming mp4)
You do not call avformat_write_header() (unlikely) or do not handle the error in the caller of your C++ member function (the return value of avformat_write_header() is checked but I do not have code corresponding to the caller of the C++ member function; your actual code might differ significantly from the code provided, so it's possible and the only plausible conclusion that can be drawn from available data)
The solution: Ensure that processing does not continue if avformat_write_header() fails. By adding the audio stream, avformat_write_header() starts to fail unless you set the stream sample rate. If the error is ignored, av_interleaved_write_frame() triggers a segmentation fault by accessing the unallocated st->internal->priv_pts.
As mentioned initially, scenario is incomplete. If you do call avformat_write_header() and stop processing in case of an error (meaning you do not call av_interleaved_write_frame()), more information is needed. As it stands now, that is unlikely. For further analysis, the executable output (stdout, stderr) is required to see your traces and FFmpeg log messages. If that does not reveal new information, a self-contained minimal working example and the video input are needed to get all the full picture.
I'm using FFmpeg and try to encode and decode a raw PCM sound to Opus using a built-in FFmpeg "opus" codec. My input samples are raw PCM 8000 Hz 16 bit mono, in AV_SAMPLE_FMT_S16 format. Since Opus requires sample format AV_SAMPLE_FMT_FLTP and sample rate 48000 Hz only, so I resample my samples before encode them.
I have two instances of ResamplerAudio class that does the work of resampling audio samples and has a member of SwrContext, I use the first instance of ResamplerAudio for resampling a raw PCM input audio before encoding and the second for resampling decoded audio to get it's format and sample rate the same as source values of input raw audio.
ResamplerAudio class has a function that init it's SwrContext member like this:
void ResamplerAudio::init(AVCodecContext *codecContext, int inSampleRate, int outSampleRate, AVSampleFormat inSampleFmt, AVSampleFormat outSampleFmt)
{
swrContext = swr_alloc();
if (!swrContext)
{
LOGE(TAG, "[init] Couldn't allocate swr context");
return;
}
av_opt_set_int(swrContext, "in_channel_layout", (int64_t) codecContext->channel_layout, 0);
av_opt_set_int(swrContext, "out_channel_layout", (int64_t) codecContext->channel_layout, 0);
av_opt_set_int(swrContext, "in_channel_count", codecContext->channels, 0);
av_opt_set_int(swrContext, "out_channel_count", codecContext->channels, 0);
av_opt_set_int(swrContext, "in_sample_rate", inSampleRate, 0);
av_opt_set_int(swrContext, "out_sample_rate", outSampleRate, 0);
av_opt_set_sample_fmt(swrContext, "in_sample_fmt", inSampleFmt, 0);
av_opt_set_sample_fmt(swrContext, "out_sample_fmt", outSampleFmt, 0);
int ret = swr_init(swrContext);
if (ret < 0)
{
LOGE(TAG, "[init] swr_init error: %s", av_err2str(ret));
return;
}
LOGD(TAG, "[init] success codecContext->channel_layout: %d; inSampleRate: %d; outSampleRate: %d; inSampleFmt: %d; outSampleFmt: %d", (int) codecContext->channel_layout, inSampleRate, outSampleRate, inSampleFmt, outSampleFmt);
}
And I call ResamplerAudio::init function for the first instance of ResamplerAudio (this instance do resamping a raw PCM input audio before encoding and I called it resamplerEncoder) with the following args:
resamplerEncoder->init(contextEncoder, 8000, 48000, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP);
The second instance of ResamplerAudio (this instance do resamping after decoding audio from Opus and I called it resamplerDecoder) I init with the following args:
resamplerDecoder->init(contextDecoder, 48000, 8000, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_S16);
The function of ResamplerAudio that does resampling looks like this:
std::vector<uint8_t> ResamplerAudio::convert(uint8_t **inData, int inSamplesCount, int outChannels, int outFormat)
{
std::vector<uint8_t> result;
uint8_t *dstData = NULL;
const int dstNbSamples = swr_get_out_samples(swrContext, inSamplesCount);
av_samples_alloc(&dstData, NULL, outChannels, dstNbSamples, AVSampleFormat(outFormat), 1);
int resampledSize = swr_convert(swrContext, &dstData, dstNbSamples, (const uint8_t **)inData, inSamplesCount);
int dstBufSize = av_samples_get_buffer_size(NULL, outChannels, resampledSize, AVSampleFormat(outFormat), 1);
if (dstBufSize <= 0) return result;
std::copy(&dstData[0], &dstData[dstBufSize], std::back_inserter(result));
return result;
}
And I call ResamplerAudio::convert function before encoding with the following args:
// data - an array of raw pcm audio
// dataLength - the length of data array
// getSamplesCount() - function that calculates samples count
// frameEncode - AVFrame that using for encode audio
std::vector<uint8_t> resampledData = resamplerEncoder->convert(&data, getSamplesCount(dataLength, frameEncode->channels, AV_SAMPLE_FMT_S16), frameEncode->channels, frameEncode->format);
getSamplesCount() function looks like this:
getSamplesCount(int bytesCount, int channels, AVSampleFormat format)
{
return bytesCount / av_get_bytes_per_sample(format) / channels;
}
After that I fill my frameEncode with resampled samples:
memcpy(&frame->data[0][0], &resampledData[0], sizeof(uint8_t) * resampledDataLength);
And pass frameEncode to encoding like this encodeFrame(resampledDataLength):
void encodeFrame(int dataLength)
{
/* send the frame for encoding */
int ret = avcodec_send_frame(contextEncoder, frameEncode);
if (ret < 0)
{
LOGE(TAG, "[encodeFrame] avcodec_send_frame error: %s", av_err2str(ret));
return;
}
/* read all the available output packets (in general there may be any number of them */
while (ret >= 0)
{
ret = avcodec_receive_packet(contextEncoder, packetEncode);
if (ret < 0 && ret != AVERROR(EAGAIN)) LOGE(TAG, "[encodeFrame] error in avcodec_receive_packet: %s", av_err2str(ret));
if (ret < 0) break;
// encodedData - std::vector<uint8_t> that stores encoded data
std::copy(&packetEncode->data[0], &packetEncode->data[dataLength], std::back_inserter(encodedData));
av_packet_unref(packetEncode);
}
}
Then I decode my encoded samples and do resampling to get back them in source sample format and sample rate so I call ResamplerAudio::convert function for resamplerDecoder with the following args:
// frameDecode - AVFrame that holds decoded audio
std::vector<uint8_t> resampledData = resamplerDecoder->convert(frameDecode->data, frameDecode->nb_samples, frameDecode->channels, AV_SAMPLE_FMT_S16);
And result sound is choppy and I also noticed that the decoded array size is bigger than the source array size with raw pcm audio.
Please any ideas what I'm doing wrong?
UPD 18.05.2020
I tested my resampling logic, I did resampling of raw pcm sound without any encoding and decoding routines. First I tried to convert the sample rate of input sound from 8000 Hz to 48000 Hz than I took resampled samples from step above and convert it's sample rate from 48000 Hz to 8000 Hz and the result sound is perfect and clean, also I did the same steps but I converted not a sample rate but a sample format from AV_SAMPLE_FMT_S16 to AV_SAMPLE_FMT_FLTP and vice versa and again the result sound is perfect and clean, also I got the same result when I coverted both a sample rate and a sample format.
So I assume that the problem of distorted and choppy sound is in my encoding or decoding routine, I think most likely in decoding routine because after decoding I ALWAYS get AVFrame with 960 nb_samples despite what was the size of input sound.
My decoding routine looks like this:
std::vector<uint8_t> decode(uint8_t *data, unsigned int dataLength)
{
decodedData.clear();
int dataSize = dataLength;
while (dataSize > 0)
{
if (!frameDecode)
{
frameDecode = av_frame_alloc();
if (!frameDecode)
{
LOGE(TAG, "[decode] Couldn't allocate the frame");
return EMPTY_DATA;
}
}
ret = av_parser_parse2(parser, contextDecoder, &packetDecode->data, &packetDecode->size, &data[0], dataSize, AV_NOPTS_VALUE, AV_NOPTS_VALUE, 0);
if (ret < 0) {
LOGE(TAG, "[decode] av_parser_parse2 error: %s", av_err2str(ret));
return EMPTY_DATA;
}
data += ret;
dataSize -= ret;
doDecode();
}
return decodedData;
}
void doDecode()
{
if (packetDecode->size) {
/* send the packet with the compressed data to the decoder */
int ret = avcodec_send_packet(contextDecoder, packetDecode);
if (ret < 0) LOGE(TAG, "[decode] avcodec_send_packet error: %s", av_err2str(ret));
/* read all the output frames (in general there may be any number of them */
while (ret >= 0)
{
ret = avcodec_receive_frame(contextDecoder, frameDecode);
if (ret < 0 && ret != AVERROR(EAGAIN) && ret != AVERROR_EOF) LOGE(TAG, "[decode] avcodec_receive_frame error: %s", av_err2str(ret));
if (ret < 0) break;
std::vector<uint8_t> resampledData = resamplerDecoder->convert(frameDecode->data, frameDecode->nb_samples, frameDecode->channels, AV_SAMPLE_FMT_S16);
if (!resampledData.size()) continue;
std::copy(&resampledData.data()[0], &resampledData.data()[resampledData.size()], std::back_inserter(decodedData));
}
}
}
UPD 30.05.2020
I decided to refuse to use FFmpeg in my project and use libopus 1.3.1 instead, so I made a wrapper around it and it works fine.
I'm using FFMPEG to decode H264 (or H265) RTSP Stream.
My system have 2 software: Server and Client
Server: Read frames from RTSP stream --> Forward frames to Client
Client: Receive frames from Server --> Decode --> Render
I have implemented and it worked ok, but there is a case make my system work not good. That is when internet from Server - Client is slow, frames can not transfer real-time to Client.
In present, I deal with this issue by Skip some frames (not send to Client) when the Queue is reached limit of count. The following is my summary code
//At Server Software (include 2 threads A and B)
//Thread A: Read AVPacket and forward to Client
while(true)
{
AVPacket packet;
av_init_packet(&packet);
packet.size = 0;
packet.data = NULL;
int ret = AVERROR(EAGAIN);
while (AVERROR(EAGAIN) == ret)
ret = av_read_frame(pFormatCtx, &packet);
if(packet.size > 0)
{
if(mySendQueue.count < 120) //limit 120 packet in queue
mySendQueue.Enqueue(packet); ////Thread B will read from this queue, to send packets to Client via TCP socket
else
;//SkipThisFrame ***: No send
}
}
//Thread B: Send To Client via TCP Socket
While(true)
{
AVPacket packet;
if(mySendQueue.Dequeue(packet))
{
SendPacketToClient(packet);
}
}
//At Server Software : Receive AVPacket from Server --> Decode --> Render
While(true)
{
AVPacket packet;
AVFrame frame;
ReadPacketFromServer(packet);
if (av_decode_asyn(pCodecCtx, &frame, &frameFinished, &packet) == RS_OK)
{
if (frameFinished)
{
RenderFrame(frame);
}
}
}
UINT32 __clrcall av_decode_asyn(AVCodecContext *pCodecCtx, AVFrame *frame, int *frameFinished, AVPacket *packet)
{
int ret = -1;
*frameFinished = 0;
if (packet)
{
ret = avcodec_send_packet(pCodecCtx, packet);
// In particular, we don't expect AVERROR(EAGAIN), because we read all
// decoded frames with avcodec_receive_frame() until done.
if (ret < 0 && ret != AVERROR_EOF)
return RS_NOT_OK;
}
ret = avcodec_receive_frame(pCodecCtx, frame);
if (ret < 0 && ret != AVERROR(EAGAIN))
{
return RS_NOT_OK;
}
if (ret >= 0)
*frameFinished = 1;
return RS_OK;
}
My question is focus in line of code SkipThisFrame ***, this algorithm skip frame continuously, so it maybe make the decoder on Client occur unexpectedly error or Crash?
And when skip frame like that, make Client Render frames is not normally?
And someone call show me the proper algorithm to skip frames in my case?
Thank you very much!
I have a brief read on doc of AVPacket, it says:
For video, it should typically contain one compressed frame.
Theoretically you cannot skip frames for a compressed video stream, as most frames do not contain complete information about that frame's image, but only contain changes compared with some previous frames. So if you skip a frame, it is probable that many trailing decoded frames won't contain correct result (until next key frame flushes whole image).
"My question is focus in line of code SkipThisFrame ***, this algorithm
skip frame continuously, so it maybe make the decoder on Client occur
unexpectedly error or Crash?"
One thing I notice is wrong...
Your While(true) statements also need a break; to stop, otherwise they will run forever, blocking other functions and causing the system to crash. Think about it, you say "While the loop is true do X-Y-Z instructions" but you never say when to stop (eg: break out of this While loop to do next instructions). Computer is stuck doing first While loop only and also repeating that to infinity...
Try setting up like this:
//At Server Software (include 2 threads A and B)
//Thread A: Read AVPacket and forward to Client
while(true)
{
AVPacket packet;
av_init_packet(&packet);
packet.size = 0;
packet.data = NULL;
int ret = AVERROR(EAGAIN);
while (AVERROR(EAGAIN) == ret) { ret = av_read_frame(pFormatCtx, &packet); }
if(packet.size > 0)
{
if(mySendQueue.count < 120) //limit 120 packet in queue
{
mySendQueue.Enqueue(packet); ////Thread B will read from this queue, to send packets to Client via TCP socket
}
//else { } //no need for ELSE if doing nothing... //SkipThisFrame ***: No send
}
break; //stop this part and move to "Thead B"
}
//Thread B: Send To Client via TCP Socket
While(true)
{
AVPacket packet;
if( mySendQueue.Dequeue(packet) )
{ SendPacketToClient(packet); break; }
}
//At Server Software : Receive AVPacket from Server --> Decode --> Render
While(true)
{
AVPacket packet; AVFrame frame;
ReadPacketFromServer(packet);
if (av_decode_asyn(pCodecCtx, &frame, &frameFinished, &packet) == RS_OK)
{
if (frameFinished) { RenderFrame(frame); break; }
}
}
UINT32 __clrcall av_decode_asyn(AVCodecContext *pCodecCtx, AVFrame *frame, int *frameFinished, AVPacket *packet)
{
int ret = -1;
*frameFinished = 0;
if (packet)
{
ret = avcodec_send_packet(pCodecCtx, packet);
// In particular, we don't expect AVERROR(EAGAIN), because we read all
// decoded frames with avcodec_receive_frame() until done.
if (ret < 0 && ret != AVERROR_EOF)
return RS_NOT_OK;
}
ret = avcodec_receive_frame(pCodecCtx, frame);
if (ret < 0 && ret != AVERROR(EAGAIN))
{
return RS_NOT_OK;
}
if (ret >= 0)
*frameFinished = 1;
return RS_OK;
}
Hope it helps. Let me know of results / errors.
I am decoding an OGG video (theora & vorbis as codecs) and want to show it on the screen (using Ogre 3D) while playing its sound. I can decode the image stream just fine and the video plays perfectly with the correct frame rate, etc.
However, I cannot get the sound to play at all with OpenAL.
Edit: I managed to make the playing sound resemble the actual audio in the video at least somewhat. Updated sample code.
Edit 2: I was able to get "almost" correct sound now. I had to set OpenAL to use AL_FORMAT_STEREO_FLOAT32 (after initializing the extension) instead of just STEREO16. Now the sound is "only" extremely high pitched and stuttering, but at the correct speed.
Here is how I decode audio packets (in a background thread, the equivalent works just fine for the image stream of the video file):
//------------------------------------------------------------------------------
int decodeAudioPacket( AVPacket& p_packet, AVCodecContext* p_audioCodecContext, AVFrame* p_frame,
FFmpegVideoPlayer* p_player, VideoInfo& p_videoInfo)
{
// Decode audio frame
int got_frame = 0;
int decoded = avcodec_decode_audio4(p_audioCodecContext, p_frame, &got_frame, &p_packet);
if (decoded < 0)
{
p_videoInfo.error = "Error decoding audio frame.";
return decoded;
}
// Frame is complete, store it in audio frame queue
if (got_frame)
{
int bufferSize = av_samples_get_buffer_size(NULL, p_audioCodecContext->channels, p_frame->nb_samples,
p_audioCodecContext->sample_fmt, 0);
int64_t duration = p_frame->pkt_duration;
int64_t dts = p_frame->pkt_dts;
if (staticOgreLog)
{
staticOgreLog->logMessage("Audio frame bufferSize / duration / dts: "
+ boost::lexical_cast<std::string>(bufferSize) + " / "
+ boost::lexical_cast<std::string>(duration) + " / "
+ boost::lexical_cast<std::string>(dts), Ogre::LML_NORMAL);
}
// Create the audio frame
AudioFrame* frame = new AudioFrame();
frame->dataSize = bufferSize;
frame->data = new uint8_t[bufferSize];
if (p_frame->channels == 2)
{
memcpy(frame->data, p_frame->data[0], bufferSize >> 1);
memcpy(frame->data + (bufferSize >> 1), p_frame->data[1], bufferSize >> 1);
}
else
{
memcpy(frame->data, p_frame->data, bufferSize);
}
double timeBase = ((double)p_audioCodecContext->time_base.num) / (double)p_audioCodecContext->time_base.den;
frame->lifeTime = duration * timeBase;
p_player->addAudioFrame(frame);
}
return decoded;
}
So, as you can see, I decode the frame, memcpy it to my own struct, AudioFrame. Now, when the sound is played, I use these audio frame like this:
int numBuffers = 4;
ALuint buffers[4];
alGenBuffers(numBuffers, buffers);
ALenum success = alGetError();
if(success != AL_NO_ERROR)
{
CONSOLE_LOG("Error on alGenBuffers : " + Ogre::StringConverter::toString(success) + alGetString(success));
return;
}
// Fill a number of data buffers with audio from the stream
std::vector<AudioFrame*> audioBuffers;
std::vector<unsigned int> audioBufferSizes;
unsigned int numReturned = FFMPEG_PLAYER->getDecodedAudioFrames(numBuffers, audioBuffers, audioBufferSizes);
// Assign the data buffers to the OpenAL buffers
for (unsigned int i = 0; i < numReturned; ++i)
{
alBufferData(buffers[i], _streamingFormat, audioBuffers[i]->data, audioBufferSizes[i], _streamingFrequency);
success = alGetError();
if(success != AL_NO_ERROR)
{
CONSOLE_LOG("Error on alBufferData : " + Ogre::StringConverter::toString(success) + alGetString(success)
+ " size: " + Ogre::StringConverter::toString(audioBufferSizes[i]));
return;
}
}
// Queue the buffers into OpenAL
alSourceQueueBuffers(_source, numReturned, buffers);
success = alGetError();
if(success != AL_NO_ERROR)
{
CONSOLE_LOG("Error queuing streaming buffers: " + Ogre::StringConverter::toString(success) + alGetString(success));
return;
}
}
alSourcePlay(_source);
The format and frequency I give to OpenAL are AL_FORMAT_STEREO_FLOAT32 (it is a stereo sound stream, and I did initialize the FLOAT32 extension) and 48000 (which is the sample rate of the AVCodecContext of the audio stream).
And during playback, I do the following to refill OpenAL's buffers:
ALint numBuffersProcessed;
// Check if OpenAL is done with any of the queued buffers
alGetSourcei(_source, AL_BUFFERS_PROCESSED, &numBuffersProcessed);
if(numBuffersProcessed <= 0)
return;
// Fill a number of data buffers with audio from the stream
std::vector<AudiFrame*> audioBuffers;
std::vector<unsigned int> audioBufferSizes;
unsigned int numFilled = FFMPEG_PLAYER->getDecodedAudioFrames(numBuffersProcessed, audioBuffers, audioBufferSizes);
// Assign the data buffers to the OpenAL buffers
ALuint buffer;
for (unsigned int i = 0; i < numFilled; ++i)
{
// Pop the oldest queued buffer from the source,
// fill it with the new data, then re-queue it
alSourceUnqueueBuffers(_source, 1, &buffer);
ALenum success = alGetError();
if(success != AL_NO_ERROR)
{
CONSOLE_LOG("Error Unqueuing streaming buffers: " + Ogre::StringConverter::toString(success));
return;
}
alBufferData(buffer, _streamingFormat, audioBuffers[i]->data, audioBufferSizes[i], _streamingFrequency);
success = alGetError();
if(success != AL_NO_ERROR)
{
CONSOLE_LOG("Error on re- alBufferData: " + Ogre::StringConverter::toString(success));
return;
}
alSourceQueueBuffers(_source, 1, &buffer);
success = alGetError();
if(success != AL_NO_ERROR)
{
CONSOLE_LOG("Error re-queuing streaming buffers: " + Ogre::StringConverter::toString(success) + " "
+ alGetString(success));
return;
}
}
// Make sure the source is still playing,
// and restart it if needed.
ALint playStatus;
alGetSourcei(_source, AL_SOURCE_STATE, &playStatus);
if(playStatus != AL_PLAYING)
alSourcePlay(_source);
As you can see, I do quite heavy error checking. But I do not get any errors, neither from OpenAL nor from FFmpeg.
Edit: What I hear somewhat resembles the actual audio from the video, but VERY high pitched and stuttering VERY much. Also, it seems to be playing on top of TV noise. Very strange. Plus, it is playing much slower than the correct audio would.
Edit: 2 After using AL_FORMAT_STEREO_FLOAT32, the sound plays at the correct speed, but is still very high pitched and stuttering (though less than before).
The video itself is not broken, it can be played fine on any player. OpenAL can also play *.way files just fine in the same application, so it is also working.
Any ideas what could be wrong here or how to do this correctly?
My only guess is that somehow, FFmpeg's decode function does not produce data OpenGL can read. But this is as far as the FFmpeg decode example goes, so I don't know what's missing. As I understand it, the decode_audio4 function decodes the frame to raw data. And OpenAL should be able to work with RAW data (or rather, doesn't work with anything else).
So, I finally figured out how to do it. Gee, what a mess. It was a hint from a user on the libav-users mailing list that put me on the correct path.
Here are my mistakes:
Using the wrong format in the alBufferData function. I used AL_FORMAT_STEREO16 (as that is what every single streaming example with OpenAL uses). I should have used AL_FORMAT_STEREO_FLOAT32, as the video I stream is Ogg and vorbis is stored in floating points. And using swr_convert to convert from AV_SAMPLE_FMT_FLTP to AV_SAMPLE_FMT_S16 just crashes. No idea why.
Not using swr_convert to convert the decoded audio frame to the target format. After I was trying to use swr_convert to convert from FLTP to S16, and it would simply crash without a reason given, I assumed it was broken. But after figuring out my first mistake, I tried again, converting from FLTP to FLT (non-planar) and then it worked! So OpenAL uses interleaved format, not planar. Good to know.
So here is the decodeAudioPacket function that is working for me with Ogg video, vorbis audio stream:
int decodeAudioPacket( AVPacket& p_packet, AVCodecContext* p_audioCodecContext, AVFrame* p_frame,
SwrContext* p_swrContext, uint8_t** p_destBuffer, int p_destLinesize,
FFmpegVideoPlayer* p_player, VideoInfo& p_videoInfo)
{
// Decode audio frame
int got_frame = 0;
int decoded = avcodec_decode_audio4(p_audioCodecContext, p_frame, &got_frame, &p_packet);
if (decoded < 0)
{
p_videoInfo.error = "Error decoding audio frame.";
return decoded;
}
if(decoded <= p_packet.size)
{
/* Move the unread data to the front and clear the end bits */
int remaining = p_packet.size - decoded;
memmove(p_packet.data, &p_packet.data[decoded], remaining);
av_shrink_packet(&p_packet, remaining);
}
// Frame is complete, store it in audio frame queue
if (got_frame)
{
int outputSamples = swr_convert(p_swrContext,
p_destBuffer, p_destLinesize,
(const uint8_t**)p_frame->extended_data, p_frame->nb_samples);
int bufferSize = av_get_bytes_per_sample(AV_SAMPLE_FMT_FLT) * p_videoInfo.audioNumChannels
* outputSamples;
int64_t duration = p_frame->pkt_duration;
int64_t dts = p_frame->pkt_dts;
if (staticOgreLog)
{
staticOgreLog->logMessage("Audio frame bufferSize / duration / dts: "
+ boost::lexical_cast<std::string>(bufferSize) + " / "
+ boost::lexical_cast<std::string>(duration) + " / "
+ boost::lexical_cast<std::string>(dts), Ogre::LML_NORMAL);
}
// Create the audio frame
AudioFrame* frame = new AudioFrame();
frame->dataSize = bufferSize;
frame->data = new uint8_t[bufferSize];
memcpy(frame->data, p_destBuffer[0], bufferSize);
double timeBase = ((double)p_audioCodecContext->time_base.num) / (double)p_audioCodecContext->time_base.den;
frame->lifeTime = duration * timeBase;
p_player->addAudioFrame(frame);
}
return decoded;
}
And here is how I initialize the context and the destination buffer:
// Initialize SWR context
SwrContext* swrContext = swr_alloc_set_opts(NULL,
audioCodecContext->channel_layout, AV_SAMPLE_FMT_FLT, audioCodecContext->sample_rate,
audioCodecContext->channel_layout, audioCodecContext->sample_fmt, audioCodecContext->sample_rate,
0, NULL);
int result = swr_init(swrContext);
// Create destination sample buffer
uint8_t** destBuffer = NULL;
int destBufferLinesize;
av_samples_alloc_array_and_samples( &destBuffer,
&destBufferLinesize,
videoInfo.audioNumChannels,
2048,
AV_SAMPLE_FMT_FLT,
0);