From 56f98e340fca894a76d1ddbe33118b8d8c4db34a Mon Sep 17 00:00:00 2001 From: Anton Khirnov Date: Thu, 24 Jul 2014 17:47:26 +0000 Subject: [PATCH] output example: convert audio to the format supported by the encoder --- doc/examples/output.c | 193 ++++++++++++++++++++++++++++++++---------- 1 file changed, 147 insertions(+), 46 deletions(-) diff --git a/doc/examples/output.c b/doc/examples/output.c index 239fe5b3eb..0534554360 100644 --- a/doc/examples/output.c +++ b/doc/examples/output.c @@ -36,7 +36,9 @@ #include "libavutil/channel_layout.h" #include "libavutil/mathematics.h" +#include "libavutil/opt.h" #include "libavformat/avformat.h" +#include "libavresample/avresample.h" #include "libswscale/swscale.h" /* 5 seconds stream duration */ @@ -60,6 +62,7 @@ typedef struct OutputStream { float t, tincr, tincr2; struct SwsContext *sws_ctx; + AVAudioResampleContext *avr; } OutputStream; /**************************************************************/ @@ -73,6 +76,7 @@ static void add_audio_stream(OutputStream *ost, AVFormatContext *oc, { AVCodecContext *c; AVCodec *codec; + int ret; /* find the audio encoder */ codec = avcodec_find_encoder(codec_id); @@ -90,23 +94,75 @@ static void add_audio_stream(OutputStream *ost, AVFormatContext *oc, c = ost->st->codec; /* put sample parameters */ - c->sample_fmt = AV_SAMPLE_FMT_S16; - c->bit_rate = 64000; - c->sample_rate = 44100; - c->channels = 2; - c->channel_layout = AV_CH_LAYOUT_STEREO; + c->sample_fmt = codec->sample_fmts ? codec->sample_fmts[0] : AV_SAMPLE_FMT_S16; + c->sample_rate = codec->supported_samplerates ? codec->supported_samplerates[0] : 44100; + c->channel_layout = codec->channel_layouts ? codec->channel_layouts[0] : AV_CH_LAYOUT_STEREO; + c->channels = av_get_channel_layout_nb_channels(c->channel_layout); + c->bit_rate = 64000; ost->st->time_base = (AVRational){ 1, c->sample_rate }; // some formats want stream headers to be separate if (oc->oformat->flags & AVFMT_GLOBALHEADER) c->flags |= CODEC_FLAG_GLOBAL_HEADER; + + /* initialize sample format conversion; + * to simplify the code, we always pass the data through lavr, even + * if the encoder supports the generated format directly -- the price is + * some extra data copying; + */ + ost->avr = avresample_alloc_context(); + if (!ost->avr) { + fprintf(stderr, "Error allocating the resampling context\n"); + exit(1); + } + + av_opt_set_int(ost->avr, "in_sample_fmt", AV_SAMPLE_FMT_S16, 0); + av_opt_set_int(ost->avr, "in_sample_rate", 44100, 0); + av_opt_set_int(ost->avr, "in_channel_layout", AV_CH_LAYOUT_STEREO, 0); + av_opt_set_int(ost->avr, "out_sample_fmt", c->sample_fmt, 0); + av_opt_set_int(ost->avr, "out_sample_rate", c->sample_rate, 0); + av_opt_set_int(ost->avr, "out_channel_layout", c->channel_layout, 0); + + ret = avresample_open(ost->avr); + if (ret < 0) { + fprintf(stderr, "Error opening the resampling context\n"); + exit(1); + } +} + +static AVFrame *alloc_audio_frame(enum AVSampleFormat sample_fmt, + uint64_t channel_layout, + int sample_rate, int nb_samples) +{ + AVFrame *frame = av_frame_alloc(); + int ret; + + if (!frame) { + fprintf(stderr, "Error allocating an audio frame\n"); + exit(1); + } + + frame->format = sample_fmt; + frame->channel_layout = channel_layout; + frame->sample_rate = sample_rate; + frame->nb_samples = nb_samples; + + if (nb_samples) { + ret = av_frame_get_buffer(frame, 0); + if (ret < 0) { + fprintf(stderr, "Error allocating an audio buffer\n"); + exit(1); + } + } + + return frame; } static void open_audio(AVFormatContext *oc, OutputStream *ost) { AVCodecContext *c; - int ret; + int nb_samples; c = ost->st->codec; @@ -122,47 +178,32 @@ static void open_audio(AVFormatContext *oc, OutputStream *ost) /* increment frequency by 110 Hz per second */ ost->tincr2 = 2 * M_PI * 110.0 / c->sample_rate / c->sample_rate; - ost->frame = av_frame_alloc(); - if (!ost->frame) - exit(1); - - ost->frame->sample_rate = c->sample_rate; - ost->frame->format = AV_SAMPLE_FMT_S16; - ost->frame->channel_layout = c->channel_layout; - if (c->codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE) - ost->frame->nb_samples = 10000; + nb_samples = 10000; else - ost->frame->nb_samples = c->frame_size; + nb_samples = c->frame_size; - ret = av_frame_get_buffer(ost->frame, 0); - if (ret < 0) { - fprintf(stderr, "Could not allocate an audio frame.\n"); - exit(1); - } + ost->frame = alloc_audio_frame(c->sample_fmt, c->channel_layout, + c->sample_rate, nb_samples); + ost->tmp_frame = alloc_audio_frame(AV_SAMPLE_FMT_S16, AV_CH_LAYOUT_STEREO, + 44100, nb_samples); } /* Prepare a 16 bit dummy audio frame of 'frame_size' samples and * 'nb_channels' channels. */ static AVFrame *get_audio_frame(OutputStream *ost) { - int j, i, v, ret; - int16_t *q = (int16_t*)ost->frame->data[0]; + AVFrame *frame = ost->tmp_frame; + int j, i, v; + int16_t *q = (int16_t*)frame->data[0]; /* check if we want to generate more frames */ if (av_compare_ts(ost->next_pts, ost->st->codec->time_base, STREAM_DURATION, (AVRational){ 1, 1 }) >= 0) return NULL; - /* when we pass a frame to the encoder, it may keep a reference to it - * internally; - * make sure we do not overwrite it here - */ - ret = av_frame_make_writable(ost->frame); - if (ret < 0) - exit(1); - for (j = 0; j < ost->frame->nb_samples; j++) { + for (j = 0; j < frame->nb_samples; j++) { v = (int)(sin(ost->t) * 10000); for (i = 0; i < ost->st->codec->channels; i++) *q++ = v; @@ -170,33 +211,26 @@ static AVFrame *get_audio_frame(OutputStream *ost) ost->tincr += ost->tincr2; } - ost->frame->pts = ost->next_pts; - ost->next_pts += ost->frame->nb_samples; - - return ost->frame; + return frame; } -/* - * encode one audio frame and send it to the muxer +/* if a frame is provided, send it to the encoder, otherwise flush the encoder; * return 1 when encoding is finished, 0 otherwise */ -static int write_audio_frame(AVFormatContext *oc, OutputStream *ost) +static int encode_audio_frame(AVFormatContext *oc, OutputStream *ost, + AVFrame *frame) { - AVCodecContext *c; AVPacket pkt = { 0 }; // data and size must be 0; - AVFrame *frame; int got_packet; av_init_packet(&pkt); - c = ost->st->codec; - - frame = get_audio_frame(ost); - - avcodec_encode_audio2(c, &pkt, frame, &got_packet); + avcodec_encode_audio2(ost->st->codec, &pkt, frame, &got_packet); if (got_packet) { pkt.stream_index = ost->st->index; + av_packet_rescale_ts(&pkt, ost->st->codec->time_base, ost->st->time_base); + /* Write the compressed frame to the media file. */ if (av_interleaved_write_frame(oc, &pkt) != 0) { fprintf(stderr, "Error while writing audio frame\n"); @@ -207,6 +241,72 @@ static int write_audio_frame(AVFormatContext *oc, OutputStream *ost) return (frame || got_packet) ? 0 : 1; } +/* + * encode one audio frame and send it to the muxer + * return 1 when encoding is finished, 0 otherwise + */ +static int process_audio_stream(AVFormatContext *oc, OutputStream *ost) +{ + AVFrame *frame; + int got_output = 0; + int ret; + + frame = get_audio_frame(ost); + got_output |= !!frame; + + /* feed the data to lavr */ + if (frame) { + ret = avresample_convert(ost->avr, NULL, 0, 0, + frame->extended_data, frame->linesize[0], + frame->nb_samples); + if (ret < 0) { + fprintf(stderr, "Error feeding audio data to the resampler\n"); + exit(1); + } + } + + while ((frame && avresample_available(ost->avr) >= ost->frame->nb_samples) || + (!frame && avresample_get_out_samples(ost->avr, 0))) { + /* when we pass a frame to the encoder, it may keep a reference to it + * internally; + * make sure we do not overwrite it here + */ + ret = av_frame_make_writable(ost->frame); + if (ret < 0) + exit(1); + + /* the difference between the two avresample calls here is that the + * first one just reads the already converted data that is buffered in + * the lavr output buffer, while the second one also flushes the + * resampler */ + if (frame) { + ret = avresample_read(ost->avr, ost->frame->extended_data, + ost->frame->nb_samples); + } else { + ret = avresample_convert(ost->avr, ost->frame->extended_data, + ost->frame->linesize[0], ost->frame->nb_samples, + NULL, 0, 0); + } + + if (ret < 0) { + fprintf(stderr, "Error while resampling\n"); + exit(1); + } else if (frame && ret != ost->frame->nb_samples) { + fprintf(stderr, "Too few samples returned from lavr\n"); + exit(1); + } + + ost->frame->nb_samples = ret; + + ost->frame->pts = ost->next_pts; + ost->next_pts += ost->frame->nb_samples; + + got_output |= encode_audio_frame(oc, ost, ret ? ost->frame : NULL); + } + + return !got_output; +} + /**************************************************************/ /* video output */ @@ -447,6 +547,7 @@ static void close_stream(AVFormatContext *oc, OutputStream *ost) av_frame_free(&ost->frame); av_frame_free(&ost->tmp_frame); sws_freeContext(ost->sws_ctx); + avresample_free(&ost->avr); } /**************************************************************/ @@ -535,7 +636,7 @@ int main(int argc, char **argv) audio_st.next_pts, audio_st.st->codec->time_base) <= 0)) { encode_video = !write_video_frame(oc, &video_st); } else { - encode_audio = !write_audio_frame(oc, &audio_st); + encode_audio = !process_audio_stream(oc, &audio_st); } }