git.sesse.net Git - ffmpeg/blob - libavcodec/libtwolame.c

   1 /*
   2  * Interface to libtwolame for mp2 encoding
   3  * Copyright (c) 2012 Paul B Mahol
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 /**
  23  * @file
  24  * Interface to libtwolame for mp2 encoding.
  25  */
  26
  27 #include <twolame.h>
  28
  29 #include "libavutil/opt.h"
  30 #include "avcodec.h"
  31 #include "internal.h"
  32 #include "mpegaudio.h"
  33
  34 typedef struct TWOLAMEContext {
  35     AVClass  *class;
  36     int      mode;
  37     int      psymodel;
  38     int      energy;
  39     int      error_protection;
  40     int      copyright;
  41     int      original;
  42
  43     twolame_options *glopts;
  44     int64_t  next_pts;
  45 } TWOLAMEContext;
  46
  47 static av_cold int twolame_encode_close(AVCodecContext *avctx)
  48 {
  49     TWOLAMEContext *s = avctx->priv_data;
  50     twolame_close(&s->glopts);
  51     return 0;
  52 }
  53
  54 static av_cold int twolame_encode_init(AVCodecContext *avctx)
  55 {
  56     TWOLAMEContext *s = avctx->priv_data;
  57     int ret;
  58
  59     avctx->frame_size = TWOLAME_SAMPLES_PER_FRAME;
  60
  61     s->glopts = twolame_init();
  62     if (!s->glopts)
  63         return AVERROR(ENOMEM);
  64
  65     twolame_set_verbosity(s->glopts, 0);
  66     twolame_set_mode(s->glopts, s->mode);
  67     twolame_set_psymodel(s->glopts, s->psymodel);
  68     twolame_set_energy_levels(s->glopts, s->energy);
  69     twolame_set_error_protection(s->glopts, s->error_protection);
  70     twolame_set_copyright(s->glopts, s->copyright);
  71     twolame_set_original(s->glopts, s->original);
  72
  73     twolame_set_num_channels(s->glopts, avctx->channels);
  74     twolame_set_in_samplerate(s->glopts, avctx->sample_rate);
  75     twolame_set_out_samplerate(s->glopts, avctx->sample_rate);
  76     if (avctx->flags & CODEC_FLAG_QSCALE || !avctx->bit_rate) {
  77         twolame_set_VBR(s->glopts, TRUE);
  78         twolame_set_VBR_level(s->glopts, avctx->global_quality);
  79         av_log(avctx, AV_LOG_WARNING, "VBR mode is experimental!\n");
  80     } else {
  81         twolame_set_bitrate(s->glopts, avctx->bit_rate / 1000);
  82     }
  83
  84     if ((ret = twolame_init_params(s->glopts)))
  85         goto error;
  86
  87     return 0;
  88 error:
  89     twolame_encode_close(avctx);
  90     return ret;
  91 }
  92
  93 static int twolame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
  94                                 const AVFrame *frame, int *got_packet_ptr)
  95 {
  96     TWOLAMEContext *s = avctx->priv_data;
  97     int ret;
  98
  99     if ((ret = ff_alloc_packet2(avctx, avpkt, MPA_MAX_CODED_FRAME_SIZE)) < 0)
 100         return ret;
 101
 102     if (frame) {
 103         switch (avctx->sample_fmt) {
 104         case AV_SAMPLE_FMT_FLT:
 105             ret = twolame_encode_buffer_float32_interleaved(s->glopts,
 106                                         frame->data[0],
 107                                         frame->nb_samples,
 108                                         avpkt->data, avpkt->size);
 109             break;
 110         case AV_SAMPLE_FMT_FLTP:
 111             ret = twolame_encode_buffer_float32(s->glopts,
 112                                         frame->data[0], frame->data[1],
 113                                         frame->nb_samples,
 114                                         avpkt->data, avpkt->size);
 115             break;
 116         case AV_SAMPLE_FMT_S16:
 117             ret = twolame_encode_buffer_interleaved(s->glopts,
 118                                         frame->data[0],
 119                                         frame->nb_samples,
 120                                         avpkt->data, avpkt->size);
 121             break;
 122         case AV_SAMPLE_FMT_S16P:
 123             ret = twolame_encode_buffer(s->glopts,
 124                                         frame->data[0], frame->data[1],
 125                                         frame->nb_samples,
 126                                         avpkt->data, avpkt->size);
 127             break;
 128         default:
 129             return AVERROR_BUG;
 130         }
 131     } else {
 132         ret = twolame_encode_flush(s->glopts, avpkt->data, avpkt->size);
 133     }
 134
 135     if (ret > 0) {
 136         avpkt->duration = ff_samples_to_time_base(avctx, avctx->frame_size);
 137         if (frame) {
 138             if (frame->pts != AV_NOPTS_VALUE)
 139                 avpkt->pts  = frame->pts;
 140         } else {
 141             avpkt->pts = s->next_pts;
 142         }
 143         if (avpkt->pts != AV_NOPTS_VALUE)
 144             s->next_pts = avpkt->pts + avpkt->duration;
 145
 146         avpkt->size     = ret;
 147         *got_packet_ptr = 1;
 148         return 0;
 149     }
 150
 151     return ret;
 152 }
 153
 154 #define OFFSET(x) offsetof(TWOLAMEContext, x)
 155 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
 156 static const AVOption options[] = {
 157     { "mode",         "Mpeg Mode", OFFSET(mode), AV_OPT_TYPE_INT, { .i64 = TWOLAME_AUTO_MODE }, TWOLAME_AUTO_MODE, TWOLAME_MONO, AE, "mode"},
 158     { "auto",         NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_AUTO_MODE },          0, 0, AE, "mode" },
 159     { "stereo",       NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_STEREO },             0, 0, AE, "mode" },
 160     { "joint_stereo", NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_JOINT_STEREO },       0, 0, AE, "mode" },
 161     { "dual_channel", NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_DUAL_CHANNEL },       0, 0, AE, "mode" },
 162     { "mono",         NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_MONO },               0, 0, AE, "mode" },
 163     { "psymodel",    "Psychoacoustic Model",  OFFSET(psymodel), AV_OPT_TYPE_INT, { .i64 = 3 }, -1, 4, AE},
 164     { "energy_levels","enable energy levels", OFFSET(energy),   AV_OPT_TYPE_INT, { .i64 = 0 },  0, 1, AE},
 165     { "error_protection","enable CRC error protection", OFFSET(error_protection), AV_OPT_TYPE_INT, { .i64 = 0 },  0, 1, AE},
 166     { "copyright",    "set MPEG Audio Copyright flag",  OFFSET(copyright), AV_OPT_TYPE_INT, { .i64 = 0 },  0, 1, AE},
 167     { "original",     "set MPEG Audio Original flag",   OFFSET(original),  AV_OPT_TYPE_INT, { .i64 = 0 },  0, 1, AE},
 168     { NULL },
 169 };
 170
 171 static const AVClass libtwolame_class = {
 172     .class_name = "libtwolame encoder",
 173     .item_name  = av_default_item_name,
 174     .option     = options,
 175     .version    = LIBAVUTIL_VERSION_INT,
 176 };
 177
 178 AVCodec ff_libtwolame_encoder = {
 179     .name                  = "libtwolame",
 180     .type                  = AVMEDIA_TYPE_AUDIO,
 181     .id                    = AV_CODEC_ID_MP2,
 182     .priv_data_size        = sizeof(TWOLAMEContext),
 183     .init                  = twolame_encode_init,
 184     .encode2               = twolame_encode_frame,
 185     .close                 = twolame_encode_close,
 186     .capabilities          = CODEC_CAP_DELAY,
 187     .sample_fmts           = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
 188                                                              AV_SAMPLE_FMT_FLTP,
 189                                                              AV_SAMPLE_FMT_S16,
 190                                                              AV_SAMPLE_FMT_S16P,
 191                                                              AV_SAMPLE_FMT_NONE },
 192     .channel_layouts       = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
 193                                                   AV_CH_LAYOUT_STEREO,
 194                                                   0 },
 195     .supported_samplerates = (const int[]){ 16000, 22050, 24000, 32000, 44100, 48000, 0 },
 196     .long_name             = NULL_IF_CONFIG_SMALL("libtwolame MP2 (MPEG audio layer 2)"),
 197     .priv_class            = &libtwolame_class,
 198 };