音频过滤器框架同视频相似(可参考我的视频框架分析),有一个小小不同是音频在出口过滤器abuffersink前还需要一个接收过滤器aformat。具体见代码。
下面是实例代码,输入2段音频,将音频混合起来。
AudioMixer.h文件
#pragma once
#include<map>
#include<mutex>
#include<cstdio>
#include<cstdint>
#include<string>
#include<memory>
#include<sstream>
#include<iostream>
#include<fstream>
extern "C" {
#include<libavcodec/avcodec.h>
#include<libavformat/avformat.h>
#include<libavfilter/avfilter.h>
#include<libavfilter/buffersink.h>
#include<libavfilter/buffersrc.h>
#include<libavutil/opt.h>
}
class AudioMixer
{
public:
AudioMixer();
virtual ~AudioMixer();
int addAudioInput(uint32_t index, uint32_t samplerate, uint32_t channels,
uint32_t bitsPerSample, AVSampleFormat format);
int addAudioOutput(const uint32_t samplerate, const uint32_t channels,
const uint32_t bitsPerSample, const AVSampleFormat format);
int init(const char* duration = "longest");
int exit();
int addFrame(uint32_t index, uint8_t* inBuf, uint32_t size);
int getFrame(uint8_t* outBuf, uint32_t maxOutBufSize);
private:
struct AudioInfo
{
AudioInfo() {
filterCtx = nullptr;
}
uint32_t samplerate;
uint32_t channels;
uint32_t bitsPerSample;
AVSampleFormat format;
std::string name;
AVFilterContext* filterCtx;
};
bool initialized_ = false;
std::mutex mutex_;
std::map<uint32_t, AudioInfo> audio_input_info_;
std::shared_ptr<AudioInfo> audio_out_info_;
std::shared_ptr<AudioInfo> audio_mix_info_;
std::shared_ptr<AudioInfo> audio_sink_info_;
AVFilterGraph* filter_graph_ = nullptr;
};
AudioMixer.cpp文件
#include "AudioMixer.h"
AudioMixer::AudioMixer() :initialized_(false), filter_graph_(nullptr), audio_out_info_(nullptr)
{
audio_mix_info_.reset(new AudioInfo);
audio_mix_info_->name = "amix";
audio_sink_info_.reset(new AudioInfo);
audio_sink_info_->name = "sink";
}
AudioMixer::~AudioMixer()
{
if (initialized_) {
exit();
}
}
int AudioMixer::addAudioInput(uint32_t index, uint32_t samplerate, uint32_t channels,
uint32_t bitsPerSample, AVSampleFormat format)
{
std::lock_guard<std::mutex> locker(mutex_);
if (initialized_) {
return -1;
}
//判断是否存在当前输入
if (audio_input_info_.find(index) != audio_input_info_.end()) {
return -1;
}
//初始化输入,可以有多个输入流
auto& filterInfo = audio_input_info_[index];
filterInfo.samplerate = samplerate;
filterInfo.channels = channels;
filterInfo.bitsPerSample = bitsPerSample;
filterInfo.format = format;
filterInfo.name = std::string("input") + std::to_string(index);
return 0;
}
int AudioMixer::addAudioOutput(const uint32_t samplerate, const uint32_t channels,
const uint32_t bitsPerSample, const AVSampleFormat format)
{
std::lock_guard<std::mutex> locker(mutex_);
if (initialized_) {
return -1;
}
audio_out_info_.reset(new AudioInfo);
audio_out_info_->samplerate = samplerate;
audio_out_info_->channels = channels;
audio_out_info_->bitsPerSample = bitsPerSample;
audio_out_info_->format = format;
audio_out_info_->name = "output";
return 0;
}
int AudioMixer::init(const char* duration)
{
std::lock_guard<std::mutex> locker(mutex_);
if (initialized_) {
return -1;
}
if (audio_input_info_.size() == 0) {
return -1;
}
filter_graph_ = avfilter_graph_alloc();
if (filter_graph_ == nullptr) {
return -1;
}
/*char args[512] = { 0 };*/
//定义过滤器amix --混音
std::stringstream args;
const AVFilter* amix = avfilter_get_by_name("amix");
audio_mix_info_->filterCtx = avfilter_graph_alloc_filter(filter_graph_, amix, "amix");
/*inputs=输入流数量, duration=决定流的结束,
* dropout_transition= 输入流结束时,容量重整时间,
* duration (longest最长输入时间,shortest最短,first第一个输入持续的时间))*/
args << "inputs=" << audio_input_info_.size()
<< ":duration=" << duration
<< ":dropout_transition=0";
std::cout << args.str() << std::endl;
//使用字符串参数初始化过滤器上下文
if (avfilter_init_str(audio_mix_info_->filterCtx, args.str().c_str()) != 0) {
std::cout << "[AudioMixer] avfilter_init_str(amix) failed." << std::endl;
return -1;
}
//定义出口过滤器abuffersink
const AVFilter* abuffersink = avfilter_get_by_name("abuffersink");
audio_sink_info_->filterCtx = avfilter_graph_alloc_filter(filter_graph_, abuffersink, "sink");
if (avfilter_init_str(audio_sink_info_->filterCtx, nullptr) != 0) {
std::cout << "[AudioMixer] avfilter_init_str(sink) failed." << std::endl;
return -1;
}
//定义入口过滤器abuffer,可能有多个输入
for (auto& iter : audio_input_info_) {
const AVFilter* abuffer = avfilter_get_by_name("abuffer");
args.str("");
args.clear();
args << "sample_rate=" << iter.second.samplerate
<< ":sample_fmt=" << av_get_sample_fmt_name(iter.second.format)
<< ":channel_layout=0x" <<av_get_default_channel_layout(iter.second.channels);
std::cout << "[AudioMixer] input(" << iter.first << ") args:" << args.str() << std::endl;
iter.second.filterCtx = avfilter_graph_alloc_filter(filter_graph_, abuffer,
audio_out_info_->name.c_str());
if (avfilter_init_str(iter.second.filterCtx, args.str().c_str()) != 0) {
std::cout << "[AudioMixer] avfilter_init_str(abuffer) failed." << std::endl;
return -1;
}
//输入过滤器链接到过滤器amix
if (avfilter_link(iter.second.filterCtx, 0, audio_mix_info_->filterCtx, iter.first) != 0) {
std::cout << "[AudioMixer] avfilter_link(abuffer(" << iter.first << "),amix failed." << std::endl;
return -1;
}
}
if (audio_out_info_ != nullptr) {
//音频在出口过滤器abuffersink前还需要一个接收过滤器aformat
const AVFilter* aformat = avfilter_get_by_name("aformat");
args.str("");
args.clear();
args << "sample_rates=" << audio_out_info_->samplerate
<< ":sample_fmts=" << av_get_sample_fmt_name(audio_out_info_->format)
<< ":channel_layouts=0x" <<av_get_default_channel_layout(audio_out_info_->channels);
std::cout << "[AudioMixer] output args:" << args.str() << std::endl;
audio_out_info_->filterCtx = avfilter_graph_alloc_filter(filter_graph_, aformat, "aformat");
if (avfilter_init_str(audio_out_info_->filterCtx, args.str().c_str()) != 0) {
std::cout << "[AudioMixer] avfilter_init_str(aformat) failed." << std::endl;
return -1;
}
//链接amix->aformat
if (avfilter_link(audio_mix_info_->filterCtx, 0, audio_out_info_->filterCtx, 0) != 0) {
std::cout << "[AudioMixer] avfilter_link(amix,aformat) failed." << std::endl;
return -1;
}
//链接aformat->buffersink
if (avfilter_link(audio_out_info_->filterCtx, 0, audio_sink_info_->filterCtx, 0) != 0) {
std::cout << "[AudioMixer] avfilter_link(aformat,abuffersink) failed." << std::endl;
return -1;
}
}
if (avfilter_graph_config(filter_graph_, nullptr) < 0) {
std::cout << "[AudioMixer] avfilter_graph_config() failed." << std::endl;
return -1;
}
char* graph_str = avfilter_graph_dump(filter_graph_, nullptr);
std::ofstream graphFile("graphFile.txt");
graphFile.write(graph_str, strlen(graph_str));
initialized_ = true;
return 0;
}
int AudioMixer::exit()
{
std::lock_guard<std::mutex> locker(mutex_);
if (initialized_)
{
for (auto iter : audio_input_info_)
{
if (iter.second.filterCtx != nullptr)
{
avfilter_free(iter.second.filterCtx);
}
}
audio_input_info_.clear();
if (audio_out_info_ && audio_out_info_->filterCtx)
{
avfilter_free(audio_out_info_->filterCtx);
audio_out_info_->filterCtx = nullptr;
}
if (audio_mix_info_->filterCtx)
{
avfilter_free(audio_mix_info_->filterCtx);
audio_mix_info_->filterCtx = nullptr;
}
if (audio_sink_info_->filterCtx)
{
avfilter_free(audio_sink_info_->filterCtx);
audio_sink_info_->filterCtx = nullptr;
}
avfilter_graph_free(&filter_graph_);
filter_graph_ = nullptr;
initialized_ = false;
}
return 0;
}
//向流程中输入帧
int AudioMixer::addFrame(uint32_t index, uint8_t* inBuf, uint32_t size)
{
std::lock_guard<std::mutex> locker(mutex_);
if (!initialized_) {
return -1;
}
auto iter = audio_input_info_.find(index);
if (iter == audio_input_info_.end()) {
return -1;
}
if (inBuf && size > 0) {
std::shared_ptr<AVFrame> avFrame(av_frame_alloc(), [](AVFrame* ptr) {av_frame_free(&ptr); });
avFrame->sample_rate = iter->second.samplerate;
avFrame->format = iter->second.format;
avFrame->channels = iter->second.channels;
avFrame->nb_samples = size * 8 / iter->second.bitsPerSample / iter->second.channels;
av_frame_get_buffer(avFrame.get(), 1);
memcpy(avFrame->extended_data[0], inBuf, size);
if (av_buffersrc_add_frame(iter->second.filterCtx, avFrame.get()) != 0) {
return -1;
}
}
else {
if (av_buffersrc_add_frame(iter->second.filterCtx,NULL) != 0) { //冲刷缓冲区
return -1;
}
}
return 0;
}
//获取处理后的帧
int AudioMixer::getFrame(uint8_t* outBuf, uint32_t maxOutBufSize)
{
std::lock_guard<std::mutex> locker(mutex_);
if (!initialized_)
{
return -1;
}
std::shared_ptr<AVFrame> avFrame(av_frame_alloc(), [](AVFrame* ptr) { av_frame_free(&ptr); });
int ret = av_buffersink_get_frame(audio_sink_info_->filterCtx, avFrame.get());
if (ret < 0)
{
// printf("ret = %d, %d\n", ret, AVERROR(EAGAIN));
return -1;
}
int size = av_samples_get_buffer_size(NULL, avFrame->channels, avFrame->nb_samples, (AVSampleFormat)avFrame->format, 1);
if (size > (int)maxOutBufSize)
{
return 0;
}
memcpy(outBuf, avFrame->extended_data[0], size);
return size;
}
main.cpp文件
#include"AudioMixer.h"
using namespace std;
#define PCM1_FRAME_SZIE (4096*2)
#define PCM2_FRAME_SZIE (4096)
#define PCM_OUT_FRAME_SIZE (40000)
int main(int argc, char** argv) {
AudioMixer amix;
//初始化参数
amix.addAudioInput(0, 48000, 2, 32, AV_SAMPLE_FMT_FLT);
amix.addAudioInput(1, 48000, 2, 16, AV_SAMPLE_FMT_S16);
amix.addAudioOutput(48000, 2, 16, AV_SAMPLE_FMT_S16);
if (amix.init("longest") < 0) {
return -1;
}
int len1 = 0, len2 = 0;
uint8_t buf1[PCM1_FRAME_SZIE];
uint8_t buf2[PCM2_FRAME_SZIE];
ifstream file1("48000_2_f32le.pcm", ios::binary);
if (!file1) {
cout << "open 48000_2_f32le.pcm failed." << endl;
return -1;
}
ifstream file2("48000_2_s16le.pcm", ios::binary);
if (!file2) {
cout << "open 48000_2_s16le.pcm failed." << endl;
return -1;
}
ofstream file_out("output.pcm", ios::binary);
if (!file_out) {
cout << "open output.pcm failed." << endl;
return -1;
}
uint8_t out_buf[PCM_OUT_FRAME_SIZE];
uint32_t out_size = 0;
int file1_finish=0;
int file2_finish=0;
//读取数据
while (1)
{
file1.read((char*)buf1, PCM1_FRAME_SZIE);
file2.read((char*)buf2, PCM2_FRAME_SZIE);
len1 = file1.gcount();
len2 = file2.gcount();
if (len1 > 0 || len2 > 0) {
if (len1 > 0) {
if (amix.addFrame(0, buf1, len1) < 0) {
cout << "amix.addFrame(0, buf1, len1) failed." << endl;
break;
}
}
else {
if (file1_finish == 0) {
file1_finish = 1;
if (amix.addFrame(0, nullptr, 0) < 0) {
cout << "amix.addFrame(0, nullptr, 0) failed." << endl;
}
}
}
if (len2 > 0) {
if (amix.addFrame(1, buf2, len2) < 0) {
cout << "amix.addFrame(1, buf2, len2) failed." << endl;
break;
}
}
else {
if (file2_finish == 0) {
file2_finish = 1;
if (amix.addFrame(1, nullptr, 0) < 0) {
cout << "amix.addFrame(1, nullptr, 0) failed." << endl;
}
}
}
int ret = 0;
while ((ret=amix.getFrame(out_buf,10240))>=0)
{
out_size += ret;
if (out_size % (1024 * 1024) == 0) {
cout << "mix audio: " << ret
<< " ,out_size: " << out_size << endl;
}
file_out.write((char*)out_buf, ret);
}
}
else {
cout << "two file finish." << endl;
break;
}
}
cout << "end, out_size: " << out_size << endl;
amix.exit();
if (file_out) {
file_out.close();
}
if (file1) {
file1.close();
}
if (file2) {
file2.close();
}
system("pause");
return 0;
}