libvorbisenc.c 13.6 KB
Newer Older
1
/*
2
 * Copyright (c) 2002 Mark Hills <mark@pogo.org.uk>
3
 *
4 5 6
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
7 8
 * modify it under the terms of the GNU Lesser General Public
 * 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
 * FFmpeg is distributed in the hope that it will be useful,
12 13 14 15 16
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with FFmpeg; if not, write to the Free Software
18
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 20
 */

21 22
#include <vorbis/vorbisenc.h>

23
#include "libavutil/avassert.h"
24
#include "libavutil/fifo.h"
25
#include "libavutil/opt.h"
26
#include "avcodec.h"
27
#include "audio_frame_queue.h"
28
#include "internal.h"
29
#include "vorbis.h"
30
#include "vorbis_parser.h"
31

32

33 34 35 36 37
/* Number of samples the user should send in each call.
 * This value is used because it is the LCD of all possible frame sizes, so
 * an output packet will always start at the same point as one of the input
 * packets.
 */
38
#define LIBVORBIS_FRAME_SIZE 64
39

40
#define BUFFER_SIZE (1024 * 64)
41

42
typedef struct LibvorbisEncContext {
43 44 45 46
    AVClass *av_class;                  /**< class for AVOptions            */
    vorbis_info vi;                     /**< vorbis_info used during init   */
    vorbis_dsp_state vd;                /**< DSP state used for analysis    */
    vorbis_block vb;                    /**< vorbis_block used for analysis */
47
    AVFifoBuffer *pkt_fifo;             /**< output packet buffer           */
48
    int eof;                            /**< end-of-file flag               */
49
    int dsp_initialized;                /**< vd has been initialized        */
50 51
    vorbis_comment vc;                  /**< VorbisComment info             */
    double iblock;                      /**< impulse block bias option      */
52
    AVVorbisParseContext *vp;           /**< parse context to get durations */
53
    AudioFrameQueue afq;                /**< frame queue for timestamps     */
54
} LibvorbisEncContext;
55

56
static const AVOption options[] = {
57
    { "iblock", "Sets the impulse block bias", offsetof(LibvorbisEncContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
58
    { NULL }
59
};
60 61 62 63 64 65

static const AVCodecDefault defaults[] = {
    { "b",  "0" },
    { NULL },
};

66
static const AVClass vorbis_class = {
67 68 69 70 71
    .class_name = "libvorbis",
    .item_name  = av_default_item_name,
    .option     = options,
    .version    = LIBAVUTIL_VERSION_INT,
};
72

73
static int vorbis_error_to_averror(int ov_err)
74
{
75 76 77 78 79
    switch (ov_err) {
    case OV_EFAULT: return AVERROR_BUG;
    case OV_EINVAL: return AVERROR(EINVAL);
    case OV_EIMPL:  return AVERROR(EINVAL);
    default:        return AVERROR_UNKNOWN;
80 81 82
    }
}

83
static av_cold int libvorbis_setup(vorbis_info *vi, AVCodecContext *avctx)
84
{
85
    LibvorbisEncContext *s = avctx->priv_data;
86
    double cfreq;
87
    int ret;
88

89
    if (avctx->flags & AV_CODEC_FLAG_QSCALE || !avctx->bit_rate) {
90 91
        /* variable bitrate
         * NOTE: we use the oggenc range of -1 to 10 for global_quality for
92
         *       user convenience, but libvorbis uses -0.1 to 1.0.
93 94
         */
        float q = avctx->global_quality / (float)FF_QP2LAMBDA;
95
        /* default to 3 if the user did not set quality or bitrate */
96
        if (!(avctx->flags & AV_CODEC_FLAG_QSCALE))
97
            q = 3.0;
98 99
        if ((ret = vorbis_encode_setup_vbr(vi, avctx->channels,
                                           avctx->sample_rate,
100 101
                                           q / 10.0)))
            goto error;
102
    } else {
103
        int minrate = avctx->rc_min_rate > 0 ? avctx->rc_min_rate : -1;
104
        int maxrate = avctx->rc_max_rate > 0 ? avctx->rc_max_rate : -1;
105

106 107
        /* average bitrate */
        if ((ret = vorbis_encode_setup_managed(vi, avctx->channels,
108 109
                                               avctx->sample_rate, maxrate,
                                               avctx->bit_rate, minrate)))
110
            goto error;
111

112
        /* variable bitrate by estimate, disable slow rate management */
113
        if (minrate == -1 && maxrate == -1)
114
            if ((ret = vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL)))
115
                goto error; /* should not happen */
116
    }
117

118
    /* cutoff frequency */
119 120
    if (avctx->cutoff > 0) {
        cfreq = avctx->cutoff / 1000.0;
121
        if ((ret = vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq)))
122
            goto error; /* should not happen */
123
    }
124

125 126 127
    /* impulse block bias */
    if (s->iblock) {
        if ((ret = vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &s->iblock)))
128
            goto error;
129 130
    }

131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146
    if (avctx->channels == 3 &&
            avctx->channel_layout != (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER) ||
        avctx->channels == 4 &&
            avctx->channel_layout != AV_CH_LAYOUT_2_2 &&
            avctx->channel_layout != AV_CH_LAYOUT_QUAD ||
        avctx->channels == 5 &&
            avctx->channel_layout != AV_CH_LAYOUT_5POINT0 &&
            avctx->channel_layout != AV_CH_LAYOUT_5POINT0_BACK ||
        avctx->channels == 6 &&
            avctx->channel_layout != AV_CH_LAYOUT_5POINT1 &&
            avctx->channel_layout != AV_CH_LAYOUT_5POINT1_BACK ||
        avctx->channels == 7 &&
            avctx->channel_layout != (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER) ||
        avctx->channels == 8 &&
            avctx->channel_layout != AV_CH_LAYOUT_7POINT1) {
        if (avctx->channel_layout) {
147
            char name[32];
148 149 150
            av_get_channel_layout_string(name, sizeof(name), avctx->channels,
                                         avctx->channel_layout);
            av_log(avctx, AV_LOG_ERROR, "%s not supported by Vorbis: "
151 152 153
                                             "output stream will have incorrect "
                                             "channel layout.\n", name);
        } else {
154
            av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The encoder "
155
                                               "will use Vorbis channel layout for "
156
                                               "%d channels.\n", avctx->channels);
157 158 159
        }
    }

160 161 162 163 164 165
    if ((ret = vorbis_encode_setup_init(vi)))
        goto error;

    return 0;
error:
    return vorbis_error_to_averror(ret);
166 167
}

168
/* How many bytes are needed for a buffer of length 'l' */
169 170
static int xiph_len(int l)
{
171
    return 1 + l / 255 + l;
172
}
173

174
static av_cold int libvorbis_encode_close(AVCodecContext *avctx)
175
{
176
    LibvorbisEncContext *s = avctx->priv_data;
177

178
    /* notify vorbisenc this is EOF */
179 180
    if (s->dsp_initialized)
        vorbis_analysis_wrote(&s->vd, 0);
181

182 183 184
    vorbis_block_clear(&s->vb);
    vorbis_dsp_clear(&s->vd);
    vorbis_info_clear(&s->vi);
185

Lukasz Marek's avatar
Lukasz Marek committed
186
    av_fifo_freep(&s->pkt_fifo);
187
    ff_af_queue_close(&s->afq);
188
    av_freep(&avctx->extradata);
189

190 191
    av_vorbis_parse_free(&s->vp);

192 193 194
    return 0;
}

195
static av_cold int libvorbis_encode_init(AVCodecContext *avctx)
196
{
197
    LibvorbisEncContext *s = avctx->priv_data;
198 199
    ogg_packet header, header_comm, header_code;
    uint8_t *p;
200
    unsigned int offset;
201
    int ret;
202

203
    vorbis_info_init(&s->vi);
204
    if ((ret = libvorbis_setup(&s->vi, avctx))) {
205
        av_log(avctx, AV_LOG_ERROR, "encoder setup failed\n");
206 207
        goto error;
    }
208
    if ((ret = vorbis_analysis_init(&s->vd, &s->vi))) {
209
        av_log(avctx, AV_LOG_ERROR, "analysis init failed\n");
210 211 212
        ret = vorbis_error_to_averror(ret);
        goto error;
    }
213
    s->dsp_initialized = 1;
214
    if ((ret = vorbis_block_init(&s->vd, &s->vb))) {
215
        av_log(avctx, AV_LOG_ERROR, "dsp init failed\n");
216 217
        ret = vorbis_error_to_averror(ret);
        goto error;
218 219
    }

220
    vorbis_comment_init(&s->vc);
221
    if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT))
222
        vorbis_comment_add_tag(&s->vc, "encoder", LIBAVCODEC_IDENT);
223

224 225
    if ((ret = vorbis_analysis_headerout(&s->vd, &s->vc, &header, &header_comm,
                                         &header_code))) {
226 227 228
        ret = vorbis_error_to_averror(ret);
        goto error;
    }
229

230 231 232 233
    avctx->extradata_size = 1 + xiph_len(header.bytes)      +
                                xiph_len(header_comm.bytes) +
                                header_code.bytes;
    p = avctx->extradata = av_malloc(avctx->extradata_size +
234
                                     AV_INPUT_BUFFER_PADDING_SIZE);
235 236 237 238
    if (!p) {
        ret = AVERROR(ENOMEM);
        goto error;
    }
239 240
    p[0]    = 2;
    offset  = 1;
241 242 243 244 245 246 247 248
    offset += av_xiphlacing(&p[offset], header.bytes);
    offset += av_xiphlacing(&p[offset], header_comm.bytes);
    memcpy(&p[offset], header.packet, header.bytes);
    offset += header.bytes;
    memcpy(&p[offset], header_comm.packet, header_comm.bytes);
    offset += header_comm.bytes;
    memcpy(&p[offset], header_code.packet, header_code.bytes);
    offset += header_code.bytes;
249
    av_assert0(offset == avctx->extradata_size);
250

251 252
    s->vp = av_vorbis_parse_init(avctx->extradata, avctx->extradata_size);
    if (!s->vp) {
253 254 255 256
        av_log(avctx, AV_LOG_ERROR, "invalid extradata\n");
        return ret;
    }

257
    vorbis_comment_clear(&s->vc);
258

259
    avctx->frame_size = LIBVORBIS_FRAME_SIZE;
260
    ff_af_queue_init(avctx, &s->afq);
261

262 263 264 265 266
    s->pkt_fifo = av_fifo_alloc(BUFFER_SIZE);
    if (!s->pkt_fifo) {
        ret = AVERROR(ENOMEM);
        goto error;
    }
267

268
    return 0;
269
error:
270
    libvorbis_encode_close(avctx);
271
    return ret;
272 273
}

274
static int libvorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
275
                                  const AVFrame *frame, int *got_packet_ptr)
276
{
277
    LibvorbisEncContext *s = avctx->priv_data;
278
    ogg_packet op;
279
    int ret, duration;
280

281
    /* send samples to libvorbis */
282 283
    if (frame) {
        const int samples = frame->nb_samples;
284
        float **buffer;
285
        int c, channels = s->vi.channels;
286

287
        buffer = vorbis_analysis_buffer(&s->vd, samples);
288 289
        for (c = 0; c < channels; c++) {
            int co = (channels > 8) ? c :
290
                     ff_vorbis_encoding_channel_layout_offsets[channels - 1][c];
291 292
            memcpy(buffer[c], frame->extended_data[co],
                   samples * sizeof(*buffer[c]));
293
        }
294 295
        if ((ret = vorbis_analysis_wrote(&s->vd, samples)) < 0) {
            av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
296
            return vorbis_error_to_averror(ret);
297
        }
298
        if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
299
            return ret;
300
    } else {
301
        if (!s->eof && s->afq.frame_alloc)
302 303
            if ((ret = vorbis_analysis_wrote(&s->vd, 0)) < 0) {
                av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
304
                return vorbis_error_to_averror(ret);
305
            }
306
        s->eof = 1;
307
    }
308

309
    /* retrieve available packets from libvorbis */
310 311 312 313 314
    while ((ret = vorbis_analysis_blockout(&s->vd, &s->vb)) == 1) {
        if ((ret = vorbis_analysis(&s->vb, NULL)) < 0)
            break;
        if ((ret = vorbis_bitrate_addblock(&s->vb)) < 0)
            break;
315

316
        /* add any available packets to the output packet buffer */
317
        while ((ret = vorbis_bitrate_flushpacket(&s->vd, &op)) == 1) {
318
            if (av_fifo_space(s->pkt_fifo) < sizeof(ogg_packet) + op.bytes) {
319
                av_log(avctx, AV_LOG_ERROR, "packet buffer is too small\n");
320
                return AVERROR_BUG;
321
            }
322 323
            av_fifo_generic_write(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
            av_fifo_generic_write(s->pkt_fifo, op.packet, op.bytes, NULL);
324
        }
325 326
        if (ret < 0) {
            av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
327
            break;
328
        }
329
    }
330 331
    if (ret < 0) {
        av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
332
        return vorbis_error_to_averror(ret);
333
    }
334

335 336 337 338 339 340
    /* check for available packets */
    if (av_fifo_size(s->pkt_fifo) < sizeof(ogg_packet))
        return 0;

    av_fifo_generic_read(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);

341
    if ((ret = ff_alloc_packet2(avctx, avpkt, op.bytes, 0)) < 0)
342 343 344 345 346
        return ret;
    av_fifo_generic_read(s->pkt_fifo, avpkt->data, op.bytes, NULL);

    avpkt->pts = ff_samples_to_time_base(avctx, op.granulepos);

347
    duration = av_vorbis_parse_frame(s->vp, avpkt->data, avpkt->size);
348 349 350
    if (duration > 0) {
        /* we do not know encoder delay until we get the first packet from
         * libvorbis, so we have to update the AudioFrameQueue counts */
351
        if (!avctx->initial_padding && s->afq.frames) {
352
            avctx->initial_padding    = duration;
353 354
            av_assert0(!s->afq.remaining_delay);
            s->afq.frames->duration  += duration;
355 356
            if (s->afq.frames->pts != AV_NOPTS_VALUE)
                s->afq.frames->pts       -= duration;
357
            s->afq.remaining_samples += duration;
358
        }
359
        ff_af_queue_remove(&s->afq, duration, &avpkt->pts, &avpkt->duration);
360 361
    }

362 363
    *got_packet_ptr = 1;
    return 0;
364 365
}

366
AVCodec ff_libvorbis_encoder = {
367
    .name           = "libvorbis",
368
    .long_name      = NULL_IF_CONFIG_SMALL("libvorbis"),
369
    .type           = AVMEDIA_TYPE_AUDIO,
370
    .id             = AV_CODEC_ID_VORBIS,
371
    .priv_data_size = sizeof(LibvorbisEncContext),
372 373 374
    .init           = libvorbis_encode_init,
    .encode2        = libvorbis_encode_frame,
    .close          = libvorbis_encode_close,
375
    .capabilities   = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_SMALL_LAST_FRAME,
376
    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
377
                                                      AV_SAMPLE_FMT_NONE },
378
    .priv_class     = &vorbis_class,
379
    .defaults       = defaults,
380
    .wrapper_name   = "libvorbis",
381
};