1 /*
2 * Copyright (c) 2002 Mark Hills <mark@pogo.org.uk>
3 *
4 * This file is part of FFmpeg.
5 *
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.
10 *
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.
15 *
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
19 */
20
21 #include <vorbis/vorbisenc.h>
22
23 #include "libavutil/avassert.h"
24 #include "libavutil/channel_layout.h"
25 #include "libavutil/fifo.h"
26 #include "libavutil/opt.h"
27 #include "avcodec.h"
28 #include "audio_frame_queue.h"
29 #include "codec_internal.h"
30 #include "encode.h"
31 #include "internal.h"
32 #include "version.h"
33 #include "vorbis.h"
34 #include "vorbis_parser.h"
35
36
37 /* Number of samples the user should send in each call.
38 * This value is used because it is the LCD of all possible frame sizes, so
39 * an output packet will always start at the same point as one of the input
40 * packets.
41 */
42 #define LIBVORBIS_FRAME_SIZE 64
43
44 #define BUFFER_SIZE (1024 * 64)
45
46 typedef struct LibvorbisEncContext {
47 AVClass *av_class; /**< class for AVOptions */
48 vorbis_info vi; /**< vorbis_info used during init */
49 vorbis_dsp_state vd; /**< DSP state used for analysis */
50 vorbis_block vb; /**< vorbis_block used for analysis */
51 AVFifo *pkt_fifo; /**< output packet buffer */
52 int eof; /**< end-of-file flag */
53 int dsp_initialized; /**< vd has been initialized */
54 vorbis_comment vc; /**< VorbisComment info */
55 double iblock; /**< impulse block bias option */
56 AVVorbisParseContext *vp; /**< parse context to get durations */
57 AudioFrameQueue afq; /**< frame queue for timestamps */
58 } LibvorbisEncContext;
59
60 static const AVOption options[] = {
61 { "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 },
62 { NULL }
63 };
64
65 static const FFCodecDefault defaults[] = {
66 { "b", "0" },
67 { NULL },
68 };
69
70 static const AVClass vorbis_class = {
71 .class_name = "libvorbis",
72 .item_name = av_default_item_name,
73 .option = options,
74 .version = LIBAVUTIL_VERSION_INT,
75 };
76
77 static const uint8_t vorbis_encoding_channel_layout_offsets[8][8] = {
78 { 0 },
79 { 0, 1 },
80 { 0, 2, 1 },
81 { 0, 1, 2, 3 },
82 { 0, 2, 1, 3, 4 },
83 { 0, 2, 1, 4, 5, 3 },
84 { 0, 2, 1, 5, 6, 4, 3 },
85 { 0, 2, 1, 6, 7, 4, 5, 3 },
86 };
87
vorbis_error_to_averror(int ov_err)88 static int vorbis_error_to_averror(int ov_err)
89 {
90 switch (ov_err) {
91 case OV_EFAULT: return AVERROR_BUG;
92 case OV_EINVAL: return AVERROR(EINVAL);
93 case OV_EIMPL: return AVERROR(EINVAL);
94 default: return AVERROR_UNKNOWN;
95 }
96 }
97
libvorbis_setup(vorbis_info * vi,AVCodecContext * avctx)98 static av_cold int libvorbis_setup(vorbis_info *vi, AVCodecContext *avctx)
99 {
100 LibvorbisEncContext *s = avctx->priv_data;
101 int channels = avctx->ch_layout.nb_channels;
102 double cfreq;
103 int ret;
104
105 if (avctx->flags & AV_CODEC_FLAG_QSCALE || !avctx->bit_rate) {
106 /* variable bitrate
107 * NOTE: we use the oggenc range of -1 to 10 for global_quality for
108 * user convenience, but libvorbis uses -0.1 to 1.0.
109 */
110 float q = avctx->global_quality / (float)FF_QP2LAMBDA;
111 /* default to 3 if the user did not set quality or bitrate */
112 if (!(avctx->flags & AV_CODEC_FLAG_QSCALE))
113 q = 3.0;
114 if ((ret = vorbis_encode_setup_vbr(vi, channels,
115 avctx->sample_rate,
116 q / 10.0)))
117 goto error;
118 } else {
119 int minrate = avctx->rc_min_rate > 0 ? avctx->rc_min_rate : -1;
120 int maxrate = avctx->rc_max_rate > 0 ? avctx->rc_max_rate : -1;
121
122 /* average bitrate */
123 if ((ret = vorbis_encode_setup_managed(vi, channels,
124 avctx->sample_rate, maxrate,
125 avctx->bit_rate, minrate)))
126 goto error;
127
128 /* variable bitrate by estimate, disable slow rate management */
129 if (minrate == -1 && maxrate == -1)
130 if ((ret = vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL)))
131 goto error; /* should not happen */
132 }
133
134 /* cutoff frequency */
135 if (avctx->cutoff > 0) {
136 cfreq = avctx->cutoff / 1000.0;
137 if ((ret = vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq)))
138 goto error; /* should not happen */
139 }
140
141 /* impulse block bias */
142 if (s->iblock) {
143 if ((ret = vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &s->iblock)))
144 goto error;
145 }
146
147 if ((channels == 3 &&
148 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_SURROUND)) ||
149 (channels == 4 &&
150 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_2_2) &&
151 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_QUAD)) ||
152 (channels == 5 &&
153 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_5POINT0) &&
154 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_5POINT0_BACK)) ||
155 (channels == 6 &&
156 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_5POINT1) &&
157 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_5POINT1_BACK)) ||
158 (channels == 7 &&
159 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_6POINT1)) ||
160 (channels == 8 &&
161 av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_7POINT1))) {
162 if (avctx->ch_layout.order != AV_CHANNEL_ORDER_UNSPEC) {
163 char name[32];
164 av_channel_layout_describe(&avctx->ch_layout, name, sizeof(name));
165 av_log(avctx, AV_LOG_ERROR, "%s not supported by Vorbis: "
166 "output stream will have incorrect "
167 "channel layout.\n", name);
168 } else {
169 av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The encoder "
170 "will use Vorbis channel layout for "
171 "%d channels.\n", channels);
172 }
173 }
174
175 if ((ret = vorbis_encode_setup_init(vi)))
176 goto error;
177
178 return 0;
179 error:
180 return vorbis_error_to_averror(ret);
181 }
182
183 /* How many bytes are needed for a buffer of length 'l' */
xiph_len(int l)184 static int xiph_len(int l)
185 {
186 return 1 + l / 255 + l;
187 }
188
libvorbis_encode_close(AVCodecContext * avctx)189 static av_cold int libvorbis_encode_close(AVCodecContext *avctx)
190 {
191 LibvorbisEncContext *s = avctx->priv_data;
192
193 /* notify vorbisenc this is EOF */
194 if (s->dsp_initialized)
195 vorbis_analysis_wrote(&s->vd, 0);
196
197 vorbis_block_clear(&s->vb);
198 vorbis_dsp_clear(&s->vd);
199 vorbis_info_clear(&s->vi);
200
201 av_fifo_freep2(&s->pkt_fifo);
202 ff_af_queue_close(&s->afq);
203
204 av_vorbis_parse_free(&s->vp);
205
206 return 0;
207 }
208
libvorbis_encode_init(AVCodecContext * avctx)209 static av_cold int libvorbis_encode_init(AVCodecContext *avctx)
210 {
211 LibvorbisEncContext *s = avctx->priv_data;
212 ogg_packet header, header_comm, header_code;
213 uint8_t *p;
214 unsigned int offset;
215 int ret;
216
217 vorbis_info_init(&s->vi);
218 if ((ret = libvorbis_setup(&s->vi, avctx))) {
219 av_log(avctx, AV_LOG_ERROR, "encoder setup failed\n");
220 goto error;
221 }
222 if ((ret = vorbis_analysis_init(&s->vd, &s->vi))) {
223 av_log(avctx, AV_LOG_ERROR, "analysis init failed\n");
224 ret = vorbis_error_to_averror(ret);
225 goto error;
226 }
227 s->dsp_initialized = 1;
228 if ((ret = vorbis_block_init(&s->vd, &s->vb))) {
229 av_log(avctx, AV_LOG_ERROR, "dsp init failed\n");
230 ret = vorbis_error_to_averror(ret);
231 goto error;
232 }
233
234 vorbis_comment_init(&s->vc);
235 if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT))
236 vorbis_comment_add_tag(&s->vc, "encoder", LIBAVCODEC_IDENT);
237
238 if ((ret = vorbis_analysis_headerout(&s->vd, &s->vc, &header, &header_comm,
239 &header_code))) {
240 ret = vorbis_error_to_averror(ret);
241 goto error;
242 }
243
244 avctx->extradata_size = 1 + xiph_len(header.bytes) +
245 xiph_len(header_comm.bytes) +
246 header_code.bytes;
247 p = avctx->extradata = av_malloc(avctx->extradata_size +
248 AV_INPUT_BUFFER_PADDING_SIZE);
249 if (!p) {
250 ret = AVERROR(ENOMEM);
251 goto error;
252 }
253 p[0] = 2;
254 offset = 1;
255 offset += av_xiphlacing(&p[offset], header.bytes);
256 offset += av_xiphlacing(&p[offset], header_comm.bytes);
257 memcpy(&p[offset], header.packet, header.bytes);
258 offset += header.bytes;
259 memcpy(&p[offset], header_comm.packet, header_comm.bytes);
260 offset += header_comm.bytes;
261 memcpy(&p[offset], header_code.packet, header_code.bytes);
262 offset += header_code.bytes;
263 av_assert0(offset == avctx->extradata_size);
264
265 s->vp = av_vorbis_parse_init(avctx->extradata, avctx->extradata_size);
266 if (!s->vp) {
267 av_log(avctx, AV_LOG_ERROR, "invalid extradata\n");
268 return ret;
269 }
270
271 vorbis_comment_clear(&s->vc);
272
273 avctx->frame_size = LIBVORBIS_FRAME_SIZE;
274 ff_af_queue_init(avctx, &s->afq);
275
276 s->pkt_fifo = av_fifo_alloc2(BUFFER_SIZE, 1, 0);
277 if (!s->pkt_fifo) {
278 ret = AVERROR(ENOMEM);
279 goto error;
280 }
281
282 return 0;
283 error:
284 libvorbis_encode_close(avctx);
285 return ret;
286 }
287
libvorbis_encode_frame(AVCodecContext * avctx,AVPacket * avpkt,const AVFrame * frame,int * got_packet_ptr)288 static int libvorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
289 const AVFrame *frame, int *got_packet_ptr)
290 {
291 LibvorbisEncContext *s = avctx->priv_data;
292 ogg_packet op;
293 int ret, duration;
294
295 /* send samples to libvorbis */
296 if (frame) {
297 const int samples = frame->nb_samples;
298 float **buffer;
299 int c, channels = s->vi.channels;
300
301 buffer = vorbis_analysis_buffer(&s->vd, samples);
302 for (c = 0; c < channels; c++) {
303 int co = (channels > 8) ? c :
304 vorbis_encoding_channel_layout_offsets[channels - 1][c];
305 memcpy(buffer[c], frame->extended_data[co],
306 samples * sizeof(*buffer[c]));
307 }
308 if ((ret = vorbis_analysis_wrote(&s->vd, samples)) < 0) {
309 av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
310 return vorbis_error_to_averror(ret);
311 }
312 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
313 return ret;
314 } else {
315 if (!s->eof && s->afq.frame_alloc)
316 if ((ret = vorbis_analysis_wrote(&s->vd, 0)) < 0) {
317 av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
318 return vorbis_error_to_averror(ret);
319 }
320 s->eof = 1;
321 }
322
323 /* retrieve available packets from libvorbis */
324 while ((ret = vorbis_analysis_blockout(&s->vd, &s->vb)) == 1) {
325 if ((ret = vorbis_analysis(&s->vb, NULL)) < 0)
326 break;
327 if ((ret = vorbis_bitrate_addblock(&s->vb)) < 0)
328 break;
329
330 /* add any available packets to the output packet buffer */
331 while ((ret = vorbis_bitrate_flushpacket(&s->vd, &op)) == 1) {
332 if (av_fifo_can_write(s->pkt_fifo) < sizeof(ogg_packet) + op.bytes) {
333 av_log(avctx, AV_LOG_ERROR, "packet buffer is too small\n");
334 return AVERROR_BUG;
335 }
336 av_fifo_write(s->pkt_fifo, &op, sizeof(ogg_packet));
337 av_fifo_write(s->pkt_fifo, op.packet, op.bytes);
338 }
339 if (ret < 0) {
340 av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
341 break;
342 }
343 }
344 if (ret < 0) {
345 av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
346 return vorbis_error_to_averror(ret);
347 }
348
349 /* Read an available packet if possible */
350 if (av_fifo_read(s->pkt_fifo, &op, sizeof(ogg_packet)) < 0)
351 return 0;
352
353 if ((ret = ff_get_encode_buffer(avctx, avpkt, op.bytes, 0)) < 0)
354 return ret;
355 av_fifo_read(s->pkt_fifo, avpkt->data, op.bytes);
356
357 avpkt->pts = ff_samples_to_time_base(avctx, op.granulepos);
358
359 duration = av_vorbis_parse_frame(s->vp, avpkt->data, avpkt->size);
360 if (duration > 0) {
361 /* we do not know encoder delay until we get the first packet from
362 * libvorbis, so we have to update the AudioFrameQueue counts */
363 if (!avctx->initial_padding && s->afq.frames) {
364 avctx->initial_padding = duration;
365 av_assert0(!s->afq.remaining_delay);
366 s->afq.frames->duration += duration;
367 if (s->afq.frames->pts != AV_NOPTS_VALUE)
368 s->afq.frames->pts -= duration;
369 s->afq.remaining_samples += duration;
370 }
371 ff_af_queue_remove(&s->afq, duration, &avpkt->pts, &avpkt->duration);
372 }
373
374 *got_packet_ptr = 1;
375 return 0;
376 }
377
378 const FFCodec ff_libvorbis_encoder = {
379 .p.name = "libvorbis",
380 .p.long_name = NULL_IF_CONFIG_SMALL("libvorbis"),
381 .p.type = AVMEDIA_TYPE_AUDIO,
382 .p.id = AV_CODEC_ID_VORBIS,
383 .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY |
384 AV_CODEC_CAP_SMALL_LAST_FRAME,
385 .priv_data_size = sizeof(LibvorbisEncContext),
386 .init = libvorbis_encode_init,
387 FF_CODEC_ENCODE_CB(libvorbis_encode_frame),
388 .close = libvorbis_encode_close,
389 .p.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
390 AV_SAMPLE_FMT_NONE },
391 .p.priv_class = &vorbis_class,
392 .defaults = defaults,
393 .p.wrapper_name = "libvorbis",
394 };
395