2 * copyright (c) 2002 Mark Hills <mark@pogo.org.uk>
4 * This file is part of FFmpeg.
6 * FFmpeg 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.
11 * FFmpeg 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.
16 * 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
23 * Vorbis encoding support via libvorbisenc.
24 * @author Mark Hills <mark@pogo.org.uk>
27 #include <vorbis/vorbisenc.h>
29 #include "libavutil/fifo.h"
30 #include "libavutil/opt.h"
32 #include "audio_frame_queue.h"
33 #include "bytestream.h"
36 #include "vorbis_parser.h"
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
46 #define OGGVORBIS_FRAME_SIZE 64
48 #define BUFFER_SIZE (1024 * 64)
50 typedef struct OggVorbisContext {
51 AVClass *av_class; /**< class for AVOptions */
53 vorbis_info vi; /**< vorbis_info used during init */
54 vorbis_dsp_state vd; /**< DSP state used for analysis */
55 vorbis_block vb; /**< vorbis_block used for analysis */
56 AVFifoBuffer *pkt_fifo; /**< output packet buffer */
57 int eof; /**< end-of-file flag */
58 int dsp_initialized; /**< vd has been initialized */
59 vorbis_comment vc; /**< VorbisComment info */
60 ogg_packet op; /**< ogg packet */
61 double iblock; /**< impulse block bias option */
62 VorbisParseContext vp; /**< parse context to get durations */
63 AudioFrameQueue afq; /**< frame queue for timestamps */
66 static const AVOption options[] = {
67 { "iblock", "Sets the impulse block bias", offsetof(OggVorbisContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
71 static const AVCodecDefault defaults[] = {
76 static const AVClass class = { "libvorbis", av_default_item_name, options, LIBAVUTIL_VERSION_INT };
79 static int vorbis_error_to_averror(int ov_err)
82 case OV_EFAULT: return AVERROR_BUG;
83 case OV_EINVAL: return AVERROR(EINVAL);
84 case OV_EIMPL: return AVERROR(EINVAL);
85 default: return AVERROR_UNKNOWN;
89 static av_cold int oggvorbis_init_encoder(vorbis_info *vi,
90 AVCodecContext *avctx)
92 OggVorbisContext *s = avctx->priv_data;
96 if (avctx->flags & CODEC_FLAG_QSCALE || !avctx->bit_rate) {
98 * NOTE: we use the oggenc range of -1 to 10 for global_quality for
99 * user convenience, but libvorbis uses -0.1 to 1.0.
101 float q = avctx->global_quality / (float)FF_QP2LAMBDA;
102 /* default to 3 if the user did not set quality or bitrate */
103 if (!(avctx->flags & CODEC_FLAG_QSCALE))
105 if ((ret = vorbis_encode_setup_vbr(vi, avctx->channels,
110 int minrate = avctx->rc_min_rate > 0 ? avctx->rc_min_rate : -1;
111 int maxrate = avctx->rc_max_rate > 0 ? avctx->rc_max_rate : -1;
113 /* average bitrate */
114 if ((ret = vorbis_encode_setup_managed(vi, avctx->channels,
115 avctx->sample_rate, maxrate,
116 avctx->bit_rate, minrate)))
119 /* variable bitrate by estimate, disable slow rate management */
120 if (minrate == -1 && maxrate == -1)
121 if ((ret = vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL)))
122 goto error; /* should not happen */
125 /* cutoff frequency */
126 if (avctx->cutoff > 0) {
127 cfreq = avctx->cutoff / 1000.0;
128 if ((ret = vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq)))
129 goto error; /* should not happen */
132 /* impulse block bias */
134 if ((ret = vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &s->iblock)))
138 if (avctx->channels == 3 &&
139 avctx->channel_layout != (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER) ||
140 avctx->channels == 4 &&
141 avctx->channel_layout != AV_CH_LAYOUT_2_2 &&
142 avctx->channel_layout != AV_CH_LAYOUT_QUAD ||
143 avctx->channels == 5 &&
144 avctx->channel_layout != AV_CH_LAYOUT_5POINT0 &&
145 avctx->channel_layout != AV_CH_LAYOUT_5POINT0_BACK ||
146 avctx->channels == 6 &&
147 avctx->channel_layout != AV_CH_LAYOUT_5POINT1 &&
148 avctx->channel_layout != AV_CH_LAYOUT_5POINT1_BACK ||
149 avctx->channels == 7 &&
150 avctx->channel_layout != (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER) ||
151 avctx->channels == 8 &&
152 avctx->channel_layout != AV_CH_LAYOUT_7POINT1) {
153 if (avctx->channel_layout) {
155 av_get_channel_layout_string(name, sizeof(name), avctx->channels,
156 avctx->channel_layout);
157 av_log(avctx, AV_LOG_ERROR, "%s not supported by Vorbis: "
158 "output stream will have incorrect "
159 "channel layout.\n", name);
161 av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The encoder "
162 "will use Vorbis channel layout for "
163 "%d channels.\n", avctx->channels);
167 if ((ret = vorbis_encode_setup_init(vi)))
172 return vorbis_error_to_averror(ret);
175 /* How many bytes are needed for a buffer of length 'l' */
176 static int xiph_len(int l)
178 return 1 + l / 255 + l;
181 static av_cold int oggvorbis_encode_close(AVCodecContext *avctx)
183 OggVorbisContext *s = avctx->priv_data;
185 /* notify vorbisenc this is EOF */
186 if (s->dsp_initialized)
187 vorbis_analysis_wrote(&s->vd, 0);
189 vorbis_block_clear(&s->vb);
190 vorbis_dsp_clear(&s->vd);
191 vorbis_info_clear(&s->vi);
193 av_fifo_free(s->pkt_fifo);
194 ff_af_queue_close(&s->afq);
195 #if FF_API_OLD_ENCODE_AUDIO
196 av_freep(&avctx->coded_frame);
198 av_freep(&avctx->extradata);
203 static av_cold int oggvorbis_encode_init(AVCodecContext *avctx)
205 OggVorbisContext *s = avctx->priv_data;
206 ogg_packet header, header_comm, header_code;
211 vorbis_info_init(&s->vi);
212 if ((ret = oggvorbis_init_encoder(&s->vi, avctx))) {
213 av_log(avctx, AV_LOG_ERROR, "encoder setup failed\n");
216 if ((ret = vorbis_analysis_init(&s->vd, &s->vi))) {
217 av_log(avctx, AV_LOG_ERROR, "analysis init failed\n");
218 ret = vorbis_error_to_averror(ret);
221 s->dsp_initialized = 1;
222 if ((ret = vorbis_block_init(&s->vd, &s->vb))) {
223 av_log(avctx, AV_LOG_ERROR, "dsp init failed\n");
224 ret = vorbis_error_to_averror(ret);
228 vorbis_comment_init(&s->vc);
229 vorbis_comment_add_tag(&s->vc, "encoder", LIBAVCODEC_IDENT);
231 if ((ret = vorbis_analysis_headerout(&s->vd, &s->vc, &header, &header_comm,
233 ret = vorbis_error_to_averror(ret);
237 avctx->extradata_size = 1 + xiph_len(header.bytes) +
238 xiph_len(header_comm.bytes) +
240 p = avctx->extradata = av_malloc(avctx->extradata_size +
241 FF_INPUT_BUFFER_PADDING_SIZE);
243 ret = AVERROR(ENOMEM);
248 offset += av_xiphlacing(&p[offset], header.bytes);
249 offset += av_xiphlacing(&p[offset], header_comm.bytes);
250 memcpy(&p[offset], header.packet, header.bytes);
251 offset += header.bytes;
252 memcpy(&p[offset], header_comm.packet, header_comm.bytes);
253 offset += header_comm.bytes;
254 memcpy(&p[offset], header_code.packet, header_code.bytes);
255 offset += header_code.bytes;
256 assert(offset == avctx->extradata_size);
258 if ((ret = avpriv_vorbis_parse_extradata(avctx, &s->vp)) < 0) {
259 av_log(avctx, AV_LOG_ERROR, "invalid extradata\n");
263 vorbis_comment_clear(&s->vc);
265 avctx->frame_size = OGGVORBIS_FRAME_SIZE;
266 ff_af_queue_init(avctx, &s->afq);
268 s->pkt_fifo = av_fifo_alloc(BUFFER_SIZE);
270 ret = AVERROR(ENOMEM);
274 #if FF_API_OLD_ENCODE_AUDIO
275 avctx->coded_frame = avcodec_alloc_frame();
276 if (!avctx->coded_frame) {
277 ret = AVERROR(ENOMEM);
284 oggvorbis_encode_close(avctx);
288 static int oggvorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
289 const AVFrame *frame, int *got_packet_ptr)
291 OggVorbisContext *s = avctx->priv_data;
295 /* send samples to libvorbis */
297 const float *audio = (const float *)frame->data[0];
298 const int samples = frame->nb_samples;
300 int c, channels = s->vi.channels;
302 buffer = vorbis_analysis_buffer(&s->vd, samples);
303 for (c = 0; c < channels; c++) {
305 int co = (channels > 8) ? c :
306 ff_vorbis_encoding_channel_layout_offsets[channels - 1][c];
307 for (i = 0; i < samples; i++)
308 buffer[c][i] = audio[i * channels + co];
310 if ((ret = vorbis_analysis_wrote(&s->vd, samples)) < 0) {
311 av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
312 return vorbis_error_to_averror(ret);
314 if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
318 if ((ret = vorbis_analysis_wrote(&s->vd, 0)) < 0) {
319 av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
320 return vorbis_error_to_averror(ret);
325 /* retrieve available packets from libvorbis */
326 while ((ret = vorbis_analysis_blockout(&s->vd, &s->vb)) == 1) {
327 if ((ret = vorbis_analysis(&s->vb, NULL)) < 0)
329 if ((ret = vorbis_bitrate_addblock(&s->vb)) < 0)
332 /* add any available packets to the output packet buffer */
333 while ((ret = vorbis_bitrate_flushpacket(&s->vd, &op)) == 1) {
334 if (av_fifo_space(s->pkt_fifo) < sizeof(ogg_packet) + op.bytes) {
335 av_log(avctx, AV_LOG_ERROR, "packet buffer is too small\n");
338 av_fifo_generic_write(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
339 av_fifo_generic_write(s->pkt_fifo, op.packet, op.bytes, NULL);
342 av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
347 av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
348 return vorbis_error_to_averror(ret);
351 /* check for available packets */
352 if (av_fifo_size(s->pkt_fifo) < sizeof(ogg_packet))
355 av_fifo_generic_read(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
357 if ((ret = ff_alloc_packet2(avctx, avpkt, op.bytes)))
359 av_fifo_generic_read(s->pkt_fifo, avpkt->data, op.bytes, NULL);
361 avpkt->pts = ff_samples_to_time_base(avctx, op.granulepos);
363 duration = avpriv_vorbis_parse_frame(&s->vp, avpkt->data, avpkt->size);
365 /* we do not know encoder delay until we get the first packet from
366 * libvorbis, so we have to update the AudioFrameQueue counts */
368 avctx->delay = duration;
369 s->afq.remaining_delay += duration;
370 s->afq.remaining_samples += duration;
372 ff_af_queue_remove(&s->afq, duration, &avpkt->pts, &avpkt->duration);
379 AVCodec ff_libvorbis_encoder = {
381 .type = AVMEDIA_TYPE_AUDIO,
382 .id = CODEC_ID_VORBIS,
383 .priv_data_size = sizeof(OggVorbisContext),
384 .init = oggvorbis_encode_init,
385 .encode2 = oggvorbis_encode_frame,
386 .close = oggvorbis_encode_close,
387 .capabilities = CODEC_CAP_DELAY,
388 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
389 AV_SAMPLE_FMT_NONE },
390 .long_name = NULL_IF_CONFIG_SMALL("libvorbis Vorbis"),
391 .priv_class = &class,
392 .defaults = defaults,
395 static int oggvorbis_decode_init(AVCodecContext *avccontext) {
396 OggVorbisContext *context = avccontext->priv_data ;
397 uint8_t *p= avccontext->extradata;
399 unsigned char *headers[3], *extradata = avccontext->extradata;
401 vorbis_info_init(&context->vi) ;
402 vorbis_comment_init(&context->vc) ;
404 if(! avccontext->extradata_size || ! p) {
405 av_log(avccontext, AV_LOG_ERROR, "vorbis extradata absent\n");
409 if(p[0] == 0 && p[1] == 30) {
410 for(i = 0; i < 3; i++){
411 hsizes[i] = bytestream_get_be16(&p);
416 unsigned int offset = 1;
420 while((*p == 0xFF) && (offset < avccontext->extradata_size)) {
425 if(offset >= avccontext->extradata_size - 1) {
426 av_log(avccontext, AV_LOG_ERROR,
427 "vorbis header sizes damaged\n");
434 hsizes[2] = avccontext->extradata_size - hsizes[0]-hsizes[1]-offset;
436 av_log(avccontext, AV_LOG_DEBUG,
437 "vorbis header sizes: %d, %d, %d, / extradata_len is %d \n",
438 hsizes[0], hsizes[1], hsizes[2], avccontext->extradata_size);
440 headers[0] = extradata + offset;
441 headers[1] = extradata + offset + hsizes[0];
442 headers[2] = extradata + offset + hsizes[0] + hsizes[1];
444 av_log(avccontext, AV_LOG_ERROR,
445 "vorbis initial header len is wrong: %d\n", *p);
450 context->op.b_o_s= i==0;
451 context->op.bytes = hsizes[i];
452 context->op.packet = headers[i];
453 if(vorbis_synthesis_headerin(&context->vi, &context->vc, &context->op)<0){
454 av_log(avccontext, AV_LOG_ERROR, "%d. vorbis header damaged\n", i+1);
459 avccontext->channels = context->vi.channels;
460 avccontext->sample_rate = context->vi.rate;
461 avccontext->time_base= (AVRational){1, avccontext->sample_rate};
463 vorbis_synthesis_init(&context->vd, &context->vi);
464 vorbis_block_init(&context->vd, &context->vb);
470 static inline int conv(int samples, float **pcm, char *buf, int channels) {
472 ogg_int16_t *ptr, *data = (ogg_int16_t*)buf ;
475 for(i = 0 ; i < channels ; i++){
479 for(j = 0 ; j < samples ; j++) {
480 *ptr = av_clip_int16(mono[j] * 32767.f);
488 static int oggvorbis_decode_frame(AVCodecContext *avccontext, void *data,
489 int *got_frame_ptr, AVPacket *avpkt)
491 OggVorbisContext *context = avccontext->priv_data ;
493 ogg_packet *op= &context->op;
494 int samples, total_samples, total_bytes;
503 context->frame.nb_samples = 8192*4;
504 if ((ret = avccontext->get_buffer(avccontext, &context->frame)) < 0) {
505 av_log(avccontext, AV_LOG_ERROR, "get_buffer() failed\n");
508 output = (int16_t *)context->frame.data[0];
511 op->packet = avpkt->data;
512 op->bytes = avpkt->size;
514 // av_log(avccontext, AV_LOG_DEBUG, "%d %d %d %"PRId64" %"PRId64" %d %d\n", op->bytes, op->b_o_s, op->e_o_s, op->granulepos, op->packetno, buf_size, context->vi.rate);
516 /* for(i=0; i<op->bytes; i++)
517 av_log(avccontext, AV_LOG_DEBUG, "%02X ", op->packet[i]);
518 av_log(avccontext, AV_LOG_DEBUG, "\n");*/
520 if(vorbis_synthesis(&context->vb, op) == 0)
521 vorbis_synthesis_blockin(&context->vd, &context->vb) ;
526 while((samples = vorbis_synthesis_pcmout(&context->vd, &pcm)) > 0) {
527 conv(samples, pcm, (char*)output + total_bytes, context->vi.channels) ;
528 total_bytes += samples * 2 * context->vi.channels ;
529 total_samples += samples ;
530 vorbis_synthesis_read(&context->vd, samples) ;
533 context->frame.nb_samples = total_samples;
535 *(AVFrame *)data = context->frame;
540 static int oggvorbis_decode_close(AVCodecContext *avccontext) {
541 OggVorbisContext *context = avccontext->priv_data ;
543 vorbis_info_clear(&context->vi) ;
544 vorbis_comment_clear(&context->vc) ;
550 AVCodec ff_libvorbis_decoder = {
552 .type = AVMEDIA_TYPE_AUDIO,
553 .id = CODEC_ID_VORBIS,
554 .priv_data_size = sizeof(OggVorbisContext),
555 .init = oggvorbis_decode_init,
556 .decode = oggvorbis_decode_frame,
557 .close = oggvorbis_decode_close,
558 .capabilities = CODEC_CAP_DELAY,