Merge commit 'e6153f173a49e5bfa70b0c04d2f82930533597b9'
[ffmpeg.git] / 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
71     twolame_set_num_channels(s->glopts, avctx->channels);
72     twolame_set_in_samplerate(s->glopts, avctx->sample_rate);
73     twolame_set_out_samplerate(s->glopts, avctx->sample_rate);
74     if (avctx->flags & CODEC_FLAG_QSCALE || !avctx->bit_rate) {
75         twolame_set_VBR(s->glopts, TRUE);
76         twolame_set_VBR_level(s->glopts, avctx->global_quality);
77         av_log(avctx, AV_LOG_WARNING, "VBR mode is experimental!\n");
78     } else {
79         twolame_set_bitrate(s->glopts, avctx->bit_rate / 1000);
80     }
81
82     if ((ret = twolame_init_params(s->glopts)))
83         goto error;
84
85     return 0;
86 error:
87     twolame_encode_close(avctx);
88     return ret;
89 }
90
91 static int twolame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
92                                 const AVFrame *frame, int *got_packet_ptr)
93 {
94     TWOLAMEContext *s = avctx->priv_data;
95     int ret;
96
97     if ((ret = ff_alloc_packet2(avctx, avpkt, MPA_MAX_CODED_FRAME_SIZE)))
98         return ret;
99
100     if (frame) {
101         switch (avctx->sample_fmt) {
102         case AV_SAMPLE_FMT_FLT:
103             ret = twolame_encode_buffer_float32_interleaved(s->glopts,
104                                         frame->data[0],
105                                         frame->nb_samples,
106                                         avpkt->data, avpkt->size);
107             break;
108         case AV_SAMPLE_FMT_FLTP:
109             ret = twolame_encode_buffer_float32(s->glopts,
110                                         frame->data[0], frame->data[1],
111                                         frame->nb_samples,
112                                         avpkt->data, avpkt->size);
113             break;
114         case AV_SAMPLE_FMT_S16:
115             ret = twolame_encode_buffer_interleaved(s->glopts,
116                                         frame->data[0],
117                                         frame->nb_samples,
118                                         avpkt->data, avpkt->size);
119             break;
120         case AV_SAMPLE_FMT_S16P:
121             ret = twolame_encode_buffer(s->glopts,
122                                         frame->data[0], frame->data[1],
123                                         frame->nb_samples,
124                                         avpkt->data, avpkt->size);
125             break;
126         default:
127             return AVERROR_BUG;
128         }
129     } else {
130         ret = twolame_encode_flush(s->glopts, avpkt->data, avpkt->size);
131     }
132
133     if (ret > 0) {
134         avpkt->duration = ff_samples_to_time_base(avctx, avctx->frame_size);
135         if (frame) {
136             if (frame->pts != AV_NOPTS_VALUE)
137                 avpkt->pts  = frame->pts;
138         } else {
139             avpkt->pts = s->next_pts;
140         }
141         if (avpkt->pts != AV_NOPTS_VALUE)
142             s->next_pts = avpkt->pts + avpkt->duration;
143
144         avpkt->size     = ret;
145         *got_packet_ptr = 1;
146         return 0;
147     }
148
149     return ret;
150 }
151
152 #define OFFSET(x) offsetof(TWOLAMEContext, x)
153 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
154 static const AVOption options[] = {
155     { "mode",         "Mpeg Mode", OFFSET(mode), AV_OPT_TYPE_INT, { .i64 = TWOLAME_AUTO_MODE }, TWOLAME_AUTO_MODE, TWOLAME_MONO, AE, "mode"},
156     { "auto",         NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_AUTO_MODE },          0, 0, AE, "mode" },
157     { "stereo",       NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_STEREO },             0, 0, AE, "mode" },
158     { "joint_stereo", NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_JOINT_STEREO },       0, 0, AE, "mode" },
159     { "dual_channel", NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_DUAL_CHANNEL },       0, 0, AE, "mode" },
160     { "mono",         NULL,         0, AV_OPT_TYPE_CONST, { .i64 = TWOLAME_MONO },               0, 0, AE, "mode" },
161     { "psymodel",    "Psychoacoustic Model",  OFFSET(psymodel), AV_OPT_TYPE_INT, { .i64 = 3 }, -1, 4, AE},
162     { "energy_levels","enable energy levels", OFFSET(energy),   AV_OPT_TYPE_INT, { .i64 = 0 },  0, 1, AE},
163     { "error_protection","enable CRC error protection", OFFSET(error_protection), AV_OPT_TYPE_INT, { .i64 = 0 },  0, 1, AE},
164     { "copyright",    "set MPEG Audio Copyright flag",  OFFSET(copyright), AV_OPT_TYPE_INT, { .i64 = 0 },  0, 1, AE},
165     { "original",     "set MPEG Audio Original flag",   OFFSET(original),  AV_OPT_TYPE_INT, { .i64 = 0 },  0, 1, AE},
166     { NULL },
167 };
168
169 static const AVClass libtwolame_class = {
170     .class_name = "libtwolame encoder",
171     .item_name  = av_default_item_name,
172     .option     = options,
173     .version    = LIBAVUTIL_VERSION_INT,
174 };
175
176 AVCodec ff_libtwolame_encoder = {
177     .name                  = "libtwolame",
178     .type                  = AVMEDIA_TYPE_AUDIO,
179     .id                    = AV_CODEC_ID_MP2,
180     .priv_data_size        = sizeof(TWOLAMEContext),
181     .init                  = twolame_encode_init,
182     .encode2               = twolame_encode_frame,
183     .close                 = twolame_encode_close,
184     .capabilities          = CODEC_CAP_DELAY,
185     .sample_fmts           = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
186                                                              AV_SAMPLE_FMT_FLTP,
187                                                              AV_SAMPLE_FMT_S16,
188                                                              AV_SAMPLE_FMT_S16P,
189                                                              AV_SAMPLE_FMT_NONE },
190     .channel_layouts       = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
191                                                   AV_CH_LAYOUT_STEREO,
192                                                   0 },
193     .supported_samplerates = (const int[]){ 16000, 22050, 24000, 32000, 44100, 48000, 0 },
194     .long_name             = NULL_IF_CONFIG_SMALL("libtwolame MP2 (MPEG audio layer 2)"),
195     .priv_class            = &libtwolame_class,
196 };