FFmpeg: libavcodec/libvorbisenc.c Source File

00001 /*
00002  * Copyright (c) 2002 Mark Hills <mark@pogo.org.uk>
00003  *
00004  * This file is part of FFmpeg.
00005  *
00006  * FFmpeg is free software; you can redistribute it and/or
00007  * modify it under the terms of the GNU Lesser General Public
00008  * License as published by the Free Software Foundation; either
00009  * version 2.1 of the License, or (at your option) any later version.
00010  *
00011  * FFmpeg is distributed in the hope that it will be useful,
00012  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00013  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00014  * Lesser General Public License for more details.
00015  *
00016  * You should have received a copy of the GNU Lesser General Public
00017  * License along with FFmpeg; if not, write to the Free Software
00018  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00019  */
00020 
00021 #include <vorbis/vorbisenc.h>
00022 
00023 #include "libavutil/avassert.h"
00024 #include "libavutil/fifo.h"
00025 #include "libavutil/opt.h"
00026 #include "avcodec.h"
00027 #include "audio_frame_queue.h"
00028 #include "internal.h"
00029 #include "vorbis.h"
00030 #include "vorbis_parser.h"
00031 
00032 
00033 /* Number of samples the user should send in each call.
00034  * This value is used because it is the LCD of all possible frame sizes, so
00035  * an output packet will always start at the same point as one of the input
00036  * packets.
00037  */
00038 #define OGGVORBIS_FRAME_SIZE 64
00039 
00040 #define BUFFER_SIZE (1024 * 64)
00041 
00042 typedef struct OggVorbisEncContext {
00043     AVClass *av_class;                  
00044     AVFrame frame;
00045     vorbis_info vi;                     
00046     vorbis_dsp_state vd;                
00047     vorbis_block vb;                    
00048     AVFifoBuffer *pkt_fifo;             
00049     int eof;                            
00050     int dsp_initialized;                
00051     vorbis_comment vc;                  
00052     double iblock;                      
00053     VorbisParseContext vp;              
00054     AudioFrameQueue afq;                
00055 } OggVorbisEncContext;
00056 
00057 static const AVOption options[] = {
00058     { "iblock", "Sets the impulse block bias", offsetof(OggVorbisEncContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
00059     { NULL }
00060 };
00061 
00062 static const AVCodecDefault defaults[] = {
00063     { "b",  "0" },
00064     { NULL },
00065 };
00066 
00067 static const AVClass class = {
00068     .class_name = "libvorbis",
00069     .item_name  = av_default_item_name,
00070     .option     = options,
00071     .version    = LIBAVUTIL_VERSION_INT,
00072 };
00073 
00074 static int vorbis_error_to_averror(int ov_err)
00075 {
00076     switch (ov_err) {
00077     case OV_EFAULT: return AVERROR_BUG;
00078     case OV_EINVAL: return AVERROR(EINVAL);
00079     case OV_EIMPL:  return AVERROR(EINVAL);
00080     default:        return AVERROR_UNKNOWN;
00081     }
00082 }
00083 
00084 static av_cold int oggvorbis_init_encoder(vorbis_info *vi,
00085                                           AVCodecContext *avctx)
00086 {
00087     OggVorbisEncContext *s = avctx->priv_data;
00088     double cfreq;
00089     int ret;
00090 
00091     if (avctx->flags & CODEC_FLAG_QSCALE || !avctx->bit_rate) {
00092         /* variable bitrate
00093          * NOTE: we use the oggenc range of -1 to 10 for global_quality for
00094          *       user convenience, but libvorbis uses -0.1 to 1.0.
00095          */
00096         float q = avctx->global_quality / (float)FF_QP2LAMBDA;
00097         /* default to 3 if the user did not set quality or bitrate */
00098         if (!(avctx->flags & CODEC_FLAG_QSCALE))
00099             q = 3.0;
00100         if ((ret = vorbis_encode_setup_vbr(vi, avctx->channels,
00101                                            avctx->sample_rate,
00102                                            q / 10.0)))
00103             goto error;
00104     } else {
00105         int minrate = avctx->rc_min_rate > 0 ? avctx->rc_min_rate : -1;
00106         int maxrate = avctx->rc_max_rate > 0 ? avctx->rc_max_rate : -1;
00107 
00108         /* average bitrate */
00109         if ((ret = vorbis_encode_setup_managed(vi, avctx->channels,
00110                                                avctx->sample_rate, maxrate,
00111                                                avctx->bit_rate, minrate)))
00112             goto error;
00113 
00114         /* variable bitrate by estimate, disable slow rate management */
00115         if (minrate == -1 && maxrate == -1)
00116             if ((ret = vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL)))
00117                 goto error; /* should not happen */
00118     }
00119 
00120     /* cutoff frequency */
00121     if (avctx->cutoff > 0) {
00122         cfreq = avctx->cutoff / 1000.0;
00123         if ((ret = vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq)))
00124             goto error; /* should not happen */
00125     }
00126 
00127     /* impulse block bias */
00128     if (s->iblock) {
00129         if ((ret = vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &s->iblock)))
00130             goto error;
00131     }
00132 
00133     if (avctx->channels == 3 &&
00134             avctx->channel_layout != (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER) ||
00135         avctx->channels == 4 &&
00136             avctx->channel_layout != AV_CH_LAYOUT_2_2 &&
00137             avctx->channel_layout != AV_CH_LAYOUT_QUAD ||
00138         avctx->channels == 5 &&
00139             avctx->channel_layout != AV_CH_LAYOUT_5POINT0 &&
00140             avctx->channel_layout != AV_CH_LAYOUT_5POINT0_BACK ||
00141         avctx->channels == 6 &&
00142             avctx->channel_layout != AV_CH_LAYOUT_5POINT1 &&
00143             avctx->channel_layout != AV_CH_LAYOUT_5POINT1_BACK ||
00144         avctx->channels == 7 &&
00145             avctx->channel_layout != (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER) ||
00146         avctx->channels == 8 &&
00147             avctx->channel_layout != AV_CH_LAYOUT_7POINT1) {
00148         if (avctx->channel_layout) {
00149             char name[32];
00150             av_get_channel_layout_string(name, sizeof(name), avctx->channels,
00151                                          avctx->channel_layout);
00152             av_log(avctx, AV_LOG_ERROR, "%s not supported by Vorbis: "
00153                                              "output stream will have incorrect "
00154                                              "channel layout.\n", name);
00155         } else {
00156             av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The encoder "
00157                                                "will use Vorbis channel layout for "
00158                                                "%d channels.\n", avctx->channels);
00159         }
00160     }
00161 
00162     if ((ret = vorbis_encode_setup_init(vi)))
00163         goto error;
00164 
00165     return 0;
00166 error:
00167     return vorbis_error_to_averror(ret);
00168 }
00169 
00170 /* How many bytes are needed for a buffer of length 'l' */
00171 static int xiph_len(int l)
00172 {
00173     return 1 + l / 255 + l;
00174 }
00175 
00176 static av_cold int oggvorbis_encode_close(AVCodecContext *avctx)
00177 {
00178     OggVorbisEncContext *s = avctx->priv_data;
00179 
00180     /* notify vorbisenc this is EOF */
00181     if (s->dsp_initialized)
00182         vorbis_analysis_wrote(&s->vd, 0);
00183 
00184     vorbis_block_clear(&s->vb);
00185     vorbis_dsp_clear(&s->vd);
00186     vorbis_info_clear(&s->vi);
00187 
00188     av_fifo_free(s->pkt_fifo);
00189     ff_af_queue_close(&s->afq);
00190 #if FF_API_OLD_ENCODE_AUDIO
00191     av_freep(&avctx->coded_frame);
00192 #endif
00193     av_freep(&avctx->extradata);
00194 
00195     return 0;
00196 }
00197 
00198 static av_cold int oggvorbis_encode_init(AVCodecContext *avctx)
00199 {
00200     OggVorbisEncContext *s = avctx->priv_data;
00201     ogg_packet header, header_comm, header_code;
00202     uint8_t *p;
00203     unsigned int offset;
00204     int ret;
00205 
00206     vorbis_info_init(&s->vi);
00207     if ((ret = oggvorbis_init_encoder(&s->vi, avctx))) {
00208         av_log(avctx, AV_LOG_ERROR, "encoder setup failed\n");
00209         goto error;
00210     }
00211     if ((ret = vorbis_analysis_init(&s->vd, &s->vi))) {
00212         av_log(avctx, AV_LOG_ERROR, "analysis init failed\n");
00213         ret = vorbis_error_to_averror(ret);
00214         goto error;
00215     }
00216     s->dsp_initialized = 1;
00217     if ((ret = vorbis_block_init(&s->vd, &s->vb))) {
00218         av_log(avctx, AV_LOG_ERROR, "dsp init failed\n");
00219         ret = vorbis_error_to_averror(ret);
00220         goto error;
00221     }
00222 
00223     vorbis_comment_init(&s->vc);
00224     if (!(avctx->flags & CODEC_FLAG_BITEXACT))
00225         vorbis_comment_add_tag(&s->vc, "encoder", LIBAVCODEC_IDENT);
00226 
00227     if ((ret = vorbis_analysis_headerout(&s->vd, &s->vc, &header, &header_comm,
00228                                          &header_code))) {
00229         ret = vorbis_error_to_averror(ret);
00230         goto error;
00231     }
00232 
00233     avctx->extradata_size = 1 + xiph_len(header.bytes)      +
00234                                 xiph_len(header_comm.bytes) +
00235                                 header_code.bytes;
00236     p = avctx->extradata = av_malloc(avctx->extradata_size +
00237                                      FF_INPUT_BUFFER_PADDING_SIZE);
00238     if (!p) {
00239         ret = AVERROR(ENOMEM);
00240         goto error;
00241     }
00242     p[0]    = 2;
00243     offset  = 1;
00244     offset += av_xiphlacing(&p[offset], header.bytes);
00245     offset += av_xiphlacing(&p[offset], header_comm.bytes);
00246     memcpy(&p[offset], header.packet, header.bytes);
00247     offset += header.bytes;
00248     memcpy(&p[offset], header_comm.packet, header_comm.bytes);
00249     offset += header_comm.bytes;
00250     memcpy(&p[offset], header_code.packet, header_code.bytes);
00251     offset += header_code.bytes;
00252     av_assert0(offset == avctx->extradata_size);
00253 
00254     if ((ret = avpriv_vorbis_parse_extradata(avctx, &s->vp)) < 0) {
00255         av_log(avctx, AV_LOG_ERROR, "invalid extradata\n");
00256         return ret;
00257     }
00258 
00259     vorbis_comment_clear(&s->vc);
00260 
00261     avctx->frame_size = OGGVORBIS_FRAME_SIZE;
00262     ff_af_queue_init(avctx, &s->afq);
00263 
00264     s->pkt_fifo = av_fifo_alloc(BUFFER_SIZE);
00265     if (!s->pkt_fifo) {
00266         ret = AVERROR(ENOMEM);
00267         goto error;
00268     }
00269 
00270 #if FF_API_OLD_ENCODE_AUDIO
00271     avctx->coded_frame = avcodec_alloc_frame();
00272     if (!avctx->coded_frame) {
00273         ret = AVERROR(ENOMEM);
00274         goto error;
00275     }
00276 #endif
00277 
00278     return 0;
00279 error:
00280     oggvorbis_encode_close(avctx);
00281     return ret;
00282 }
00283 
00284 static int oggvorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
00285                                   const AVFrame *frame, int *got_packet_ptr)
00286 {
00287     OggVorbisEncContext *s = avctx->priv_data;
00288     ogg_packet op;
00289     int ret, duration;
00290 
00291     /* send samples to libvorbis */
00292     if (frame) {
00293         const int samples = frame->nb_samples;
00294         float **buffer;
00295         int c, channels = s->vi.channels;
00296 
00297         buffer = vorbis_analysis_buffer(&s->vd, samples);
00298         for (c = 0; c < channels; c++) {
00299             int co = (channels > 8) ? c :
00300                      ff_vorbis_encoding_channel_layout_offsets[channels - 1][c];
00301             memcpy(buffer[c], frame->extended_data[co],
00302                    samples * sizeof(*buffer[c]));
00303         }
00304         if ((ret = vorbis_analysis_wrote(&s->vd, samples)) < 0) {
00305             av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
00306             return vorbis_error_to_averror(ret);
00307         }
00308         if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
00309             return ret;
00310     } else {
00311         if (!s->eof)
00312             if ((ret = vorbis_analysis_wrote(&s->vd, 0)) < 0) {
00313                 av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
00314                 return vorbis_error_to_averror(ret);
00315             }
00316         s->eof = 1;
00317     }
00318 
00319     /* retrieve available packets from libvorbis */
00320     while ((ret = vorbis_analysis_blockout(&s->vd, &s->vb)) == 1) {
00321         if ((ret = vorbis_analysis(&s->vb, NULL)) < 0)
00322             break;
00323         if ((ret = vorbis_bitrate_addblock(&s->vb)) < 0)
00324             break;
00325 
00326         /* add any available packets to the output packet buffer */
00327         while ((ret = vorbis_bitrate_flushpacket(&s->vd, &op)) == 1) {
00328             if (av_fifo_space(s->pkt_fifo) < sizeof(ogg_packet) + op.bytes) {
00329                 av_log(avctx, AV_LOG_ERROR, "packet buffer is too small\n");
00330                 return AVERROR_BUG;
00331             }
00332             av_fifo_generic_write(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
00333             av_fifo_generic_write(s->pkt_fifo, op.packet, op.bytes, NULL);
00334         }
00335         if (ret < 0) {
00336             av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
00337             break;
00338         }
00339     }
00340     if (ret < 0) {
00341         av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
00342         return vorbis_error_to_averror(ret);
00343     }
00344 
00345     /* check for available packets */
00346     if (av_fifo_size(s->pkt_fifo) < sizeof(ogg_packet))
00347         return 0;
00348 
00349     av_fifo_generic_read(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
00350 
00351     if ((ret = ff_alloc_packet2(avctx, avpkt, op.bytes)))
00352         return ret;
00353     av_fifo_generic_read(s->pkt_fifo, avpkt->data, op.bytes, NULL);
00354 
00355     avpkt->pts = ff_samples_to_time_base(avctx, op.granulepos);
00356 
00357     duration = avpriv_vorbis_parse_frame(&s->vp, avpkt->data, avpkt->size);
00358     if (duration > 0) {
00359         /* we do not know encoder delay until we get the first packet from
00360          * libvorbis, so we have to update the AudioFrameQueue counts */
00361         if (!avctx->delay && s->afq.frames) {
00362             avctx->delay              = duration;
00363             av_assert0(!s->afq.remaining_delay);
00364             s->afq.frames->duration  += duration;
00365             s->afq.frames->pts       -= duration;
00366             s->afq.remaining_samples += duration;
00367         }
00368         ff_af_queue_remove(&s->afq, duration, &avpkt->pts, &avpkt->duration);
00369     }
00370 
00371     *got_packet_ptr = 1;
00372     return 0;
00373 }
00374 
00375 AVCodec ff_libvorbis_encoder = {
00376     .name           = "libvorbis",
00377     .type           = AVMEDIA_TYPE_AUDIO,
00378     .id             = AV_CODEC_ID_VORBIS,
00379     .priv_data_size = sizeof(OggVorbisEncContext),
00380     .init           = oggvorbis_encode_init,
00381     .encode2        = oggvorbis_encode_frame,
00382     .close          = oggvorbis_encode_close,
00383     .capabilities   = CODEC_CAP_DELAY,
00384     .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
00385                                                       AV_SAMPLE_FMT_NONE },
00386     .long_name      = NULL_IF_CONFIG_SMALL("libvorbis"),
00387     .priv_class     = &class,
00388     .defaults       = defaults,
00389 };