libavcodec/libvorbis.c

   1 /*
   2  * copyright (c) 2002 Mark Hills <mark@pogo.org.uk>
   3  *
   4  * This file is part of Libav.
   5  *
   6  * Libav is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * Libav is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with Libav; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 /**
  22  * @file
  23  * Vorbis encoding support via libvorbisenc.
  24  * @author Mark Hills <mark@pogo.org.uk>
  25  */
  26
  27 #include <vorbis/vorbisenc.h>
  28
  29 #include "libavutil/fifo.h"
  30 #include "libavutil/opt.h"
  31 #include "avcodec.h"
  32 #include "audio_frame_queue.h"
  33 #include "bytestream.h"
  34 #include "internal.h"
  35 #include "vorbis.h"
  36 #include "vorbis_parser.h"
  37
  38 #undef NDEBUG
  39 #include <assert.h>
  40
  41 /* Number of samples the user should send in each call.
  42  * This value is used because it is the LCD of all possible frame sizes, so
  43  * an output packet will always start at the same point as one of the input
  44  * packets.
  45  */
  46 #define LIBVORBIS_FRAME_SIZE 64
  47
  48 #define BUFFER_SIZE (1024 * 64)
  49
  50 typedef struct LibvorbisContext {
  51     AVClass *av_class;                  /**< class for AVOptions            */
  52     vorbis_info vi;                     /**< vorbis_info used during init   */
  53     vorbis_dsp_state vd;                /**< DSP state used for analysis    */
  54     vorbis_block vb;                    /**< vorbis_block used for analysis */
  55     AVFifoBuffer *pkt_fifo;             /**< output packet buffer           */
  56     int eof;                            /**< end-of-file flag               */
  57     int dsp_initialized;                /**< vd has been initialized        */
  58     vorbis_comment vc;                  /**< VorbisComment info             */
  59     ogg_packet op;                      /**< ogg packet                     */
  60     double iblock;                      /**< impulse block bias option      */
  61     AVVorbisParseContext *vp;           /**< parse context to get durations */
  62     AudioFrameQueue afq;                /**< frame queue for timestamps     */
  63 } LibvorbisContext;
  64
  65 static const AVOption options[] = {
  66     { "iblock", "Sets the impulse block bias", offsetof(LibvorbisContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
  67     { NULL }
  68 };
  69
  70 static const AVCodecDefault defaults[] = {
  71     { "b",  "0" },
  72     { NULL },
  73 };
  74
  75 static const AVClass class = {
  76     .class_name = "libvorbis",
  77     .item_name  = av_default_item_name,
  78     .option     = options,
  79     .version    = LIBAVUTIL_VERSION_INT,
  80 };
  81
  82
  83 static int vorbis_error_to_averror(int ov_err)
  84 {
  85     switch (ov_err) {
  86     case OV_EFAULT: return AVERROR_BUG;
  87     case OV_EINVAL: return AVERROR(EINVAL);
  88     case OV_EIMPL:  return AVERROR(EINVAL);
  89     default:        return AVERROR_UNKNOWN;
  90     }
  91 }
  92
  93 static av_cold int libvorbis_setup(vorbis_info *vi, AVCodecContext *avctx)
  94 {
  95     LibvorbisContext *s = avctx->priv_data;
  96     double cfreq;
  97     int ret;
  98
  99     if (avctx->flags & AV_CODEC_FLAG_QSCALE || !avctx->bit_rate) {
 100         /* variable bitrate
 101          * NOTE: we use the oggenc range of -1 to 10 for global_quality for
 102          *       user convenience, but libvorbis uses -0.1 to 1.0.
 103          */
 104         float q = avctx->global_quality / (float)FF_QP2LAMBDA;
 105         /* default to 3 if the user did not set quality or bitrate */
 106         if (!(avctx->flags & AV_CODEC_FLAG_QSCALE))
 107             q = 3.0;
 108         if ((ret = vorbis_encode_setup_vbr(vi, avctx->channels,
 109                                            avctx->sample_rate,
 110                                            q / 10.0)))
 111             goto error;
 112     } else {
 113         int minrate = avctx->rc_min_rate > 0 ? avctx->rc_min_rate : -1;
 114         int maxrate = avctx->rc_max_rate > 0 ? avctx->rc_max_rate : -1;
 115
 116         /* average bitrate */
 117         if ((ret = vorbis_encode_setup_managed(vi, avctx->channels,
 118                                                avctx->sample_rate, maxrate,
 119                                                avctx->bit_rate, minrate)))
 120             goto error;
 121
 122         /* variable bitrate by estimate, disable slow rate management */
 123         if (minrate == -1 && maxrate == -1)
 124             if ((ret = vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL)))
 125                 goto error;
 126     }
 127
 128     /* cutoff frequency */
 129     if (avctx->cutoff > 0) {
 130         cfreq = avctx->cutoff / 1000.0;
 131         if ((ret = vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq)))
 132             goto error;
 133     }
 134
 135     /* impulse block bias */
 136     if (s->iblock) {
 137         if ((ret = vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &s->iblock)))
 138             goto error;
 139     }
 140
 141     if ((ret = vorbis_encode_setup_init(vi)))
 142         goto error;
 143
 144     return 0;
 145 error:
 146     return vorbis_error_to_averror(ret);
 147 }
 148
 149 /* How many bytes are needed for a buffer of length 'l' */
 150 static int xiph_len(int l)
 151 {
 152     return 1 + l / 255 + l;
 153 }
 154
 155 static av_cold int libvorbis_encode_close(AVCodecContext *avctx)
 156 {
 157     LibvorbisContext *s = avctx->priv_data;
 158
 159     /* notify vorbisenc this is EOF */
 160     if (s->dsp_initialized)
 161         vorbis_analysis_wrote(&s->vd, 0);
 162
 163     vorbis_block_clear(&s->vb);
 164     vorbis_dsp_clear(&s->vd);
 165     vorbis_info_clear(&s->vi);
 166
 167     av_fifo_free(s->pkt_fifo);
 168     ff_af_queue_close(&s->afq);
 169     av_freep(&avctx->extradata);
 170
 171     av_vorbis_parse_free(&s->vp);
 172
 173     return 0;
 174 }
 175
 176 static av_cold int libvorbis_encode_init(AVCodecContext *avctx)
 177 {
 178     LibvorbisContext *s = avctx->priv_data;
 179     ogg_packet header, header_comm, header_code;
 180     uint8_t *p;
 181     unsigned int offset;
 182     int ret;
 183
 184     vorbis_info_init(&s->vi);
 185     if ((ret = libvorbis_setup(&s->vi, avctx))) {
 186         av_log(avctx, AV_LOG_ERROR, "encoder setup failed\n");
 187         goto error;
 188     }
 189     if ((ret = vorbis_analysis_init(&s->vd, &s->vi))) {
 190         av_log(avctx, AV_LOG_ERROR, "analysis init failed\n");
 191         ret = vorbis_error_to_averror(ret);
 192         goto error;
 193     }
 194     s->dsp_initialized = 1;
 195     if ((ret = vorbis_block_init(&s->vd, &s->vb))) {
 196         av_log(avctx, AV_LOG_ERROR, "dsp init failed\n");
 197         ret = vorbis_error_to_averror(ret);
 198         goto error;
 199     }
 200
 201     vorbis_comment_init(&s->vc);
 202     vorbis_comment_add_tag(&s->vc, "encoder", LIBAVCODEC_IDENT);
 203
 204     if ((ret = vorbis_analysis_headerout(&s->vd, &s->vc, &header, &header_comm,
 205                                          &header_code))) {
 206         ret = vorbis_error_to_averror(ret);
 207         goto error;
 208     }
 209
 210     avctx->extradata_size = 1 + xiph_len(header.bytes)      +
 211                                 xiph_len(header_comm.bytes) +
 212                                 header_code.bytes;
 213     p = avctx->extradata = av_malloc(avctx->extradata_size +
 214                                      AV_INPUT_BUFFER_PADDING_SIZE);
 215     if (!p) {
 216         ret = AVERROR(ENOMEM);
 217         goto error;
 218     }
 219     p[0]    = 2;
 220     offset  = 1;
 221     offset += av_xiphlacing(&p[offset], header.bytes);
 222     offset += av_xiphlacing(&p[offset], header_comm.bytes);
 223     memcpy(&p[offset], header.packet, header.bytes);
 224     offset += header.bytes;
 225     memcpy(&p[offset], header_comm.packet, header_comm.bytes);
 226     offset += header_comm.bytes;
 227     memcpy(&p[offset], header_code.packet, header_code.bytes);
 228     offset += header_code.bytes;
 229     assert(offset == avctx->extradata_size);
 230
 231     s->vp = av_vorbis_parse_init(avctx->extradata, avctx->extradata_size);
 232     if (!s->vp) {
 233         av_log(avctx, AV_LOG_ERROR, "invalid extradata\n");
 234         return ret;
 235     }
 236
 237     vorbis_comment_clear(&s->vc);
 238
 239     avctx->frame_size = LIBVORBIS_FRAME_SIZE;
 240     ff_af_queue_init(avctx, &s->afq);
 241
 242     s->pkt_fifo = av_fifo_alloc(BUFFER_SIZE);
 243     if (!s->pkt_fifo) {
 244         ret = AVERROR(ENOMEM);
 245         goto error;
 246     }
 247
 248     return 0;
 249 error:
 250     libvorbis_encode_close(avctx);
 251     return ret;
 252 }
 253
 254 static int libvorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
 255                                   const AVFrame *frame, int *got_packet_ptr)
 256 {
 257     LibvorbisContext *s = avctx->priv_data;
 258     ogg_packet op;
 259     int ret, duration;
 260
 261     /* send samples to libvorbis */
 262     if (frame) {
 263         const int samples = frame->nb_samples;
 264         float **buffer;
 265         int c, channels = s->vi.channels;
 266
 267         buffer = vorbis_analysis_buffer(&s->vd, samples);
 268         for (c = 0; c < channels; c++) {
 269             int co = (channels > 8) ? c :
 270                      ff_vorbis_encoding_channel_layout_offsets[channels - 1][c];
 271             memcpy(buffer[c], frame->extended_data[co],
 272                    samples * sizeof(*buffer[c]));
 273         }
 274         if ((ret = vorbis_analysis_wrote(&s->vd, samples)) < 0) {
 275             av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
 276             return vorbis_error_to_averror(ret);
 277         }
 278         if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
 279             return ret;
 280     } else {
 281         if (!s->eof)
 282             if ((ret = vorbis_analysis_wrote(&s->vd, 0)) < 0) {
 283                 av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
 284                 return vorbis_error_to_averror(ret);
 285             }
 286         s->eof = 1;
 287     }
 288
 289     /* retrieve available packets from libvorbis */
 290     while ((ret = vorbis_analysis_blockout(&s->vd, &s->vb)) == 1) {
 291         if ((ret = vorbis_analysis(&s->vb, NULL)) < 0)
 292             break;
 293         if ((ret = vorbis_bitrate_addblock(&s->vb)) < 0)
 294             break;
 295
 296         /* add any available packets to the output packet buffer */
 297         while ((ret = vorbis_bitrate_flushpacket(&s->vd, &op)) == 1) {
 298             if (av_fifo_space(s->pkt_fifo) < sizeof(ogg_packet) + op.bytes) {
 299                 av_log(avctx, AV_LOG_ERROR, "packet buffer is too small");
 300                 return AVERROR_BUG;
 301             }
 302             av_fifo_generic_write(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
 303             av_fifo_generic_write(s->pkt_fifo, op.packet, op.bytes, NULL);
 304         }
 305         if (ret < 0) {
 306             av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
 307             break;
 308         }
 309     }
 310     if (ret < 0) {
 311         av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
 312         return vorbis_error_to_averror(ret);
 313     }
 314
 315     /* check for available packets */
 316     if (av_fifo_size(s->pkt_fifo) < sizeof(ogg_packet))
 317         return 0;
 318
 319     av_fifo_generic_read(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
 320
 321     if ((ret = ff_alloc_packet(avpkt, op.bytes))) {
 322         av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
 323         return ret;
 324     }
 325     av_fifo_generic_read(s->pkt_fifo, avpkt->data, op.bytes, NULL);
 326
 327     avpkt->pts = ff_samples_to_time_base(avctx, op.granulepos);
 328
 329     duration = av_vorbis_parse_frame(s->vp, avpkt->data, avpkt->size);
 330     if (duration > 0) {
 331         /* we do not know encoder delay until we get the first packet from
 332          * libvorbis, so we have to update the AudioFrameQueue counts */
 333         if (!avctx->initial_padding) {
 334             avctx->initial_padding    = duration;
 335             s->afq.remaining_delay   += duration;
 336             s->afq.remaining_samples += duration;
 337         }
 338         ff_af_queue_remove(&s->afq, duration, &avpkt->pts, &avpkt->duration);
 339     }
 340
 341     *got_packet_ptr = 1;
 342     return 0;
 343 }
 344
 345 AVCodec ff_libvorbis_encoder = {
 346     .name           = "libvorbis",
 347     .long_name      = NULL_IF_CONFIG_SMALL("libvorbis Vorbis"),
 348     .type           = AVMEDIA_TYPE_AUDIO,
 349     .id             = AV_CODEC_ID_VORBIS,
 350     .priv_data_size = sizeof(LibvorbisContext),
 351     .init           = libvorbis_encode_init,
 352     .encode2        = libvorbis_encode_frame,
 353     .close          = libvorbis_encode_close,
 354     .capabilities   = AV_CODEC_CAP_DELAY,
 355     .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
 356                                                       AV_SAMPLE_FMT_NONE },
 357     .priv_class     = &class,
 358     .defaults       = defaults,
 359     .wrapper_name   = "libvorbis",
 360 };