1 #include "audio_encoder.h"
4 #include <libavcodec/avcodec.h>
5 #include <libavformat/avformat.h>
6 #include <libswresample/swresample.h>
7 #include <libavutil/channel_layout.h>
8 #include <libavutil/error.h>
9 #include <libavutil/frame.h>
10 #include <libavutil/mem.h>
11 #include <libavutil/opt.h>
12 #include <libavutil/rational.h>
13 #include <libavutil/samplefmt.h>
25 #include "shared/mux.h"
26 #include "shared/shared_defs.h"
27 #include "shared/timebase.h"
31 AudioEncoder::AudioEncoder(const string &codec_name, int bit_rate, const AVOutputFormat *oformat)
33 const AVCodec *codec = avcodec_find_encoder_by_name(codec_name.c_str());
34 if (codec == nullptr) {
35 fprintf(stderr, "ERROR: Could not find codec '%s'\n", codec_name.c_str());
39 ctx = avcodec_alloc_context3(codec);
40 ctx->bit_rate = bit_rate;
41 ctx->sample_rate = OUTPUT_FREQUENCY;
42 ctx->sample_fmt = codec->sample_fmts[0];
43 ctx->ch_layout.order = AV_CHANNEL_ORDER_NATIVE;
44 ctx->ch_layout.nb_channels = 2;
45 ctx->ch_layout.u.mask = AV_CH_LAYOUT_STEREO;
46 ctx->time_base = AVRational{1, TIMEBASE};
47 if (oformat->flags & AVFMT_GLOBALHEADER) {
48 ctx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
50 if (avcodec_open2(ctx, codec, NULL) < 0) {
51 fprintf(stderr, "Could not open codec '%s'\n", codec_name.c_str());
56 int ok = swr_alloc_set_opts2(&resampler,
57 /*out_ch_layout=*/&ctx->ch_layout,
58 /*out_sample_fmt=*/ctx->sample_fmt,
59 /*out_sample_rate=*/OUTPUT_FREQUENCY,
60 /*in_ch_layout=*/&ctx->ch_layout,
61 /*in_sample_fmt=*/AV_SAMPLE_FMT_FLT,
62 /*in_sample_rate=*/OUTPUT_FREQUENCY,
66 fprintf(stderr, "Allocating resampler failed.\n");
70 if (swr_init(resampler) < 0) {
71 fprintf(stderr, "Could not open resample context.\n");
75 audio_frame = av_frame_alloc();
78 AudioEncoder::~AudioEncoder()
80 av_frame_free(&audio_frame);
82 avcodec_free_context(&ctx);
85 void AudioEncoder::encode_audio(const vector<float> &audio, int64_t audio_pts)
87 if (ctx->frame_size == 0) {
88 // No queueing needed.
89 assert(audio_queue.empty());
90 assert(audio.size() % 2 == 0);
91 encode_audio_one_frame(&audio[0], audio.size() / 2, audio_pts);
95 int64_t sample_offset = audio_queue.size();
97 audio_queue.insert(audio_queue.end(), audio.begin(), audio.end());
100 sample_num + ctx->frame_size * 2 <= audio_queue.size();
101 sample_num += ctx->frame_size * 2) {
102 int64_t adjusted_audio_pts = audio_pts + (int64_t(sample_num) - sample_offset) * TIMEBASE / (OUTPUT_FREQUENCY * 2);
103 encode_audio_one_frame(&audio_queue[sample_num],
107 audio_queue.erase(audio_queue.begin(), audio_queue.begin() + sample_num);
109 last_pts = audio_pts + audio.size() * TIMEBASE / (OUTPUT_FREQUENCY * 2);
112 void AudioEncoder::encode_audio_one_frame(const float *audio, size_t num_samples, int64_t audio_pts)
114 audio_frame->pts = audio_pts;
115 audio_frame->nb_samples = num_samples;
116 audio_frame->ch_layout.order = AV_CHANNEL_ORDER_NATIVE;
117 audio_frame->ch_layout.nb_channels = 2;
118 audio_frame->ch_layout.u.mask = AV_CH_LAYOUT_STEREO;
119 audio_frame->format = ctx->sample_fmt;
120 audio_frame->sample_rate = OUTPUT_FREQUENCY;
122 if (av_samples_alloc(audio_frame->data, nullptr, 2, num_samples, ctx->sample_fmt, 0) < 0) {
123 fprintf(stderr, "Could not allocate %zu samples.\n", num_samples);
127 if (swr_convert(resampler, audio_frame->data, num_samples, reinterpret_cast<const uint8_t **>(&audio), num_samples) < 0) {
128 fprintf(stderr, "Audio conversion failed.\n");
132 int err = avcodec_send_frame(ctx, audio_frame);
134 fprintf(stderr, "avcodec_send_frame() failed with error %d\n", err);
138 for ( ;; ) { // Termination condition within loop.
139 AVPacketWithDeleter pkt = av_packet_alloc_unique();
142 int err = avcodec_receive_packet(ctx, pkt.get());
144 pkt->stream_index = 1;
146 for (Mux *mux : muxes) {
147 mux->add_packet(*pkt, pkt->pts, pkt->dts);
149 } else if (err == AVERROR(EAGAIN)) {
152 fprintf(stderr, "avcodec_receive_frame() failed with error %d\n", err);
157 av_freep(&audio_frame->data[0]);
158 av_frame_unref(audio_frame);
161 void AudioEncoder::encode_last_audio()
163 if (!audio_queue.empty()) {
164 // Last frame can be whatever size we want.
165 assert(audio_queue.size() % 2 == 0);
166 encode_audio_one_frame(&audio_queue[0], audio_queue.size() / 2, last_pts);
170 if (ctx->codec->capabilities & AV_CODEC_CAP_DELAY) {
171 // Collect any delayed frames.
173 AVPacketWithDeleter pkt = av_packet_alloc_unique();
176 int err = avcodec_receive_packet(ctx, pkt.get());
178 pkt->stream_index = 1;
180 for (Mux *mux : muxes) {
181 mux->add_packet(*pkt, pkt->pts, pkt->dts);
183 } else if (err == AVERROR_EOF) {
186 fprintf(stderr, "avcodec_receive_frame() failed with error %d\n", err);
193 AVCodecParametersWithDeleter AudioEncoder::get_codec_parameters()
195 AVCodecParameters *codecpar = avcodec_parameters_alloc();
196 avcodec_parameters_from_context(codecpar, ctx);
197 return AVCodecParametersWithDeleter(codecpar);