--- /dev/null
+#include <stdio.h>
+
+extern "C" {
+
+#include <libavcodec/avcodec.h>
+#include <libavformat/avformat.h>
+#include <libswresample/swresample.h>
+
+}
+
+#include <memory>
+#include <vector>
+
+namespace {
+
+struct AVFormatCloserAndDeleter {
+ void operator() (AVFormatContext *ctx) {
+ avformat_close_input(&ctx);
+ avformat_free_context(ctx);
+ }
+};
+
+struct AVCodecContextDeleter {
+ void operator() (AVCodecContext *ctx) {
+ avcodec_free_context(&ctx);
+ }
+};
+
+struct SwrContextDeleter {
+ void operator() (SwrContext *swr) {
+ swr_free(&swr);
+ }
+};
+
+struct AVPacketDeleter {
+ void operator() (AVPacket *pkt) {
+ av_free_packet(pkt);
+ }
+};
+
+struct AVFrameDeleter {
+ void operator() (AVFrame *frame) {
+ av_frame_free(&frame);
+ }
+};
+
+struct AVSampleDeleter {
+ void operator() (uint8_t *data) {
+ av_freep(&data);
+ }
+};
+
+void convert_samples(SwrContext *swr, int sample_rate, const uint8_t **data, int nb_samples, std::vector<int16_t> *samples)
+{
+ int max_out_samples = nb_samples + swr_get_delay(swr, sample_rate);
+ if (max_out_samples == 0) {
+ return;
+ }
+ uint8_t *output;
+ av_samples_alloc(&output, nullptr, 1, max_out_samples, AV_SAMPLE_FMT_S16, 0);
+ std::unique_ptr<uint8_t, AVSampleDeleter> output_deleter(output);
+
+ int out_samples = swr_convert(swr, &output, max_out_samples, data, nb_samples);
+ if (out_samples > 0) {
+ const int16_t* start = reinterpret_cast<const int16_t *>(output);
+ const int16_t* end = start + out_samples;
+ samples->insert(samples->end(), start, end);
+ }
+}
+
+int decode_packet(const char *filename, AVCodecContext *codec_ctx, SwrContext *swr, AVFrame *audio_frame, AVPacket *packet, int *got_frame, std::vector<int16_t> *samples)
+{
+ *got_frame = 0;
+ int len1 = avcodec_decode_audio4(codec_ctx, audio_frame, got_frame, packet);
+ if (len1 < 0 || !*got_frame) {
+ return len1;
+ }
+
+ if (audio_frame->channel_layout != codec_ctx->channel_layout ||
+ audio_frame->sample_rate != codec_ctx->sample_rate) {
+ fprintf(stderr, "%s: Channel layout or sample rate changed mid-file\n", filename);
+ *got_frame = false;
+ return len1;
+ }
+ convert_samples(swr, codec_ctx->sample_rate, (const uint8_t **)audio_frame->data, audio_frame->nb_samples, samples);
+ return len1;
+}
+
+} // namespace
+
+bool read_audio_file(const char *filename, std::vector<int16_t> *samples)
+{
+ av_register_all();
+
+ AVFormatContext *format_ctx = nullptr;
+ if (avformat_open_input(&format_ctx, filename, nullptr, nullptr) != 0) {
+ fprintf(stderr, "Couldn't open %s\n", filename);
+ return false;
+ }
+ std::unique_ptr<AVFormatContext, AVFormatCloserAndDeleter> format_ctx_closer(format_ctx);
+
+ if (avformat_find_stream_info(format_ctx, nullptr) < 0) {
+ fprintf(stderr, "%s: Couldn't find stream information\n", filename);
+ return false;
+ }
+
+ // Find the first audio stream.
+ int audio_stream_index = -1;
+ for (unsigned i = 0; i < format_ctx->nb_streams; ++i) {
+ if (format_ctx->streams[i]->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
+ audio_stream_index = i;
+ break;
+ }
+ }
+ if (audio_stream_index == -1) {
+ fprintf(stderr, "%s: Couldn't find an audio stream\n", filename);
+ return false;
+ }
+
+ AVCodec *codec = avcodec_find_decoder(format_ctx->streams[audio_stream_index]->codec->codec_id);
+ if (codec == nullptr) {
+ fprintf(stderr, "%s: Unsupported codec\n", filename);
+ return false;
+ }
+
+ AVCodecContext *codec_ctx = avcodec_alloc_context3(codec);
+ std::unique_ptr<AVCodecContext, AVCodecContextDeleter> codec_ctx_deleter(codec_ctx);
+ if (avcodec_copy_context(codec_ctx, format_ctx->streams[audio_stream_index]->codec) != 0) {
+ fprintf(stderr, "%s: Couldn't copy codec context\n", filename);
+ return false;
+ }
+
+ if (avcodec_open2(codec_ctx, codec, nullptr) < 0) {
+ fprintf(stderr, "%s: Couldn't open codec\n", filename);
+ return false;
+ }
+
+ // Init resampler (to downmix to mono and convert to s16).
+ if (codec_ctx->channel_layout == 0) {
+ codec_ctx->channel_layout = av_get_default_channel_layout(codec_ctx->channels);
+ }
+ SwrContext *swr = swr_alloc_set_opts(
+ nullptr,
+ AV_CH_LAYOUT_MONO, AV_SAMPLE_FMT_S16, codec_ctx->sample_rate,
+ codec_ctx->channel_layout, codec_ctx->sample_fmt, codec_ctx->sample_rate,
+ 0, nullptr);
+ std::unique_ptr<SwrContext, SwrContextDeleter> swr_deleter(swr);
+ if (swr_init(swr) < 0) {
+ fprintf(stderr, "%s: Couldn't initialize resampler\n", filename);
+ return false;
+ }
+
+ AVPacket packet;
+ AVFrame* audio_frame = av_frame_alloc();
+ std::unique_ptr<AVFrame, AVFrameDeleter> audio_frame_deleter(audio_frame);
+ while (av_read_frame(format_ctx, &packet) >= 0) {
+ std::unique_ptr<AVPacket, AVPacketDeleter> av_packet_deleter(&packet);
+
+ if (packet.stream_index != audio_stream_index) {
+ continue;
+ }
+
+ while (packet.size > 0) {
+ int got_frame = 0;
+ int len1 = decode_packet(filename, codec_ctx, swr, audio_frame, &packet, &got_frame, samples);
+ if (len1 < 0) {
+ fprintf(stderr, "%s: Couldn't decode audio\n", filename);
+ return false;
+ }
+ if (!got_frame) {
+ break;
+ }
+ packet.data += len1;
+ packet.size -= len1;
+ }
+ }
+
+ // Flush any delayed data from the end.
+ packet.data = nullptr;
+ packet.size = 0;
+ int got_frame = 0;
+ do {
+ int len1 = decode_packet(filename, codec_ctx, swr, audio_frame, &packet, &got_frame, samples);
+ if (len1 < 0) {
+ fprintf(stderr, "%s: Couldn't decode audio\n", filename);
+ return false;
+ }
+ } while (got_frame);
+
+ // Convert any leftover samples from the converter.
+ convert_samples(swr, codec_ctx->sample_rate, nullptr, 0, samples);
+
+ return true;
+}