1 /*
2 * Opus decoder
3 * Copyright (c) 2012 Andrew D'Addesio
4 * Copyright (c) 2013-2014 Mozilla Corporation
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 /**
24 * @file
25 * Opus decoder
26 * @author Andrew D'Addesio, Anton Khirnov
27 *
28 * Codec homepage: http://opus-codec.org/
29 * Specification: http://tools.ietf.org/html/rfc6716
30 * Ogg Opus specification: https://tools.ietf.org/html/draft-ietf-codec-oggopus-03
31 *
32 * Ogg-contained .opus files can be produced with opus-tools:
33 * http://git.xiph.org/?p=opus-tools.git
34 */
35
36 #include <stdint.h>
37
38 #include "libavutil/attributes.h"
39 #include "libavutil/audio_fifo.h"
40 #include "libavutil/channel_layout.h"
41 #include "libavutil/opt.h"
42
43 #include "libswresample/swresample.h"
44
45 #include "avcodec.h"
46 #include "get_bits.h"
47 #include "internal.h"
48 #include "mathops.h"
49 #include "opus.h"
50 #include "opustab.h"
51 #include "opus_celt.h"
52
53 static const uint16_t silk_frame_duration_ms[16] = {
54 10, 20, 40, 60,
55 10, 20, 40, 60,
56 10, 20, 40, 60,
57 10, 20,
58 10, 20,
59 };
60
61 /* number of samples of silence to feed to the resampler
62 * at the beginning */
63 static const int silk_resample_delay[] = {
64 4, 8, 11, 11, 11
65 };
66
get_silk_samplerate(int config)67 static int get_silk_samplerate(int config)
68 {
69 if (config < 4)
70 return 8000;
71 else if (config < 8)
72 return 12000;
73 return 16000;
74 }
75
opus_fade(float * out,const float * in1,const float * in2,const float * window,int len)76 static void opus_fade(float *out,
77 const float *in1, const float *in2,
78 const float *window, int len)
79 {
80 int i;
81 for (i = 0; i < len; i++)
82 out[i] = in2[i] * window[i] + in1[i] * (1.0 - window[i]);
83 }
84
opus_flush_resample(OpusStreamContext * s,int nb_samples)85 static int opus_flush_resample(OpusStreamContext *s, int nb_samples)
86 {
87 int celt_size = av_audio_fifo_size(s->celt_delay);
88 int ret, i;
89 ret = swr_convert(s->swr,
90 (uint8_t**)s->out, nb_samples,
91 NULL, 0);
92 if (ret < 0)
93 return ret;
94 else if (ret != nb_samples) {
95 av_log(s->avctx, AV_LOG_ERROR, "Wrong number of flushed samples: %d\n",
96 ret);
97 return AVERROR_BUG;
98 }
99
100 if (celt_size) {
101 if (celt_size != nb_samples) {
102 av_log(s->avctx, AV_LOG_ERROR, "Wrong number of CELT delay samples.\n");
103 return AVERROR_BUG;
104 }
105 av_audio_fifo_read(s->celt_delay, (void**)s->celt_output, nb_samples);
106 for (i = 0; i < s->output_channels; i++) {
107 s->fdsp->vector_fmac_scalar(s->out[i],
108 s->celt_output[i], 1.0,
109 nb_samples);
110 }
111 }
112
113 if (s->redundancy_idx) {
114 for (i = 0; i < s->output_channels; i++)
115 opus_fade(s->out[i], s->out[i],
116 s->redundancy_output[i] + 120 + s->redundancy_idx,
117 ff_celt_window2 + s->redundancy_idx, 120 - s->redundancy_idx);
118 s->redundancy_idx = 0;
119 }
120
121 s->out[0] += nb_samples;
122 s->out[1] += nb_samples;
123 s->out_size -= nb_samples * sizeof(float);
124
125 return 0;
126 }
127
opus_init_resample(OpusStreamContext * s)128 static int opus_init_resample(OpusStreamContext *s)
129 {
130 static const float delay[16] = { 0.0 };
131 const uint8_t *delayptr[2] = { (uint8_t*)delay, (uint8_t*)delay };
132 int ret;
133
134 av_opt_set_int(s->swr, "in_sample_rate", s->silk_samplerate, 0);
135 ret = swr_init(s->swr);
136 if (ret < 0) {
137 av_log(s->avctx, AV_LOG_ERROR, "Error opening the resampler.\n");
138 return ret;
139 }
140
141 ret = swr_convert(s->swr,
142 NULL, 0,
143 delayptr, silk_resample_delay[s->packet.bandwidth]);
144 if (ret < 0) {
145 av_log(s->avctx, AV_LOG_ERROR,
146 "Error feeding initial silence to the resampler.\n");
147 return ret;
148 }
149
150 return 0;
151 }
152
opus_decode_redundancy(OpusStreamContext * s,const uint8_t * data,int size)153 static int opus_decode_redundancy(OpusStreamContext *s, const uint8_t *data, int size)
154 {
155 int ret = ff_opus_rc_dec_init(&s->redundancy_rc, data, size);
156 if (ret < 0)
157 goto fail;
158 ff_opus_rc_dec_raw_init(&s->redundancy_rc, data + size, size);
159
160 ret = ff_celt_decode_frame(s->celt, &s->redundancy_rc,
161 s->redundancy_output,
162 s->packet.stereo + 1, 240,
163 0, ff_celt_band_end[s->packet.bandwidth]);
164 if (ret < 0)
165 goto fail;
166
167 return 0;
168 fail:
169 av_log(s->avctx, AV_LOG_ERROR, "Error decoding the redundancy frame.\n");
170 return ret;
171 }
172
opus_decode_frame(OpusStreamContext * s,const uint8_t * data,int size)173 static int opus_decode_frame(OpusStreamContext *s, const uint8_t *data, int size)
174 {
175 int samples = s->packet.frame_duration;
176 int redundancy = 0;
177 int redundancy_size, redundancy_pos;
178 int ret, i, consumed;
179 int delayed_samples = s->delayed_samples;
180
181 ret = ff_opus_rc_dec_init(&s->rc, data, size);
182 if (ret < 0)
183 return ret;
184
185 /* decode the silk frame */
186 if (s->packet.mode == OPUS_MODE_SILK || s->packet.mode == OPUS_MODE_HYBRID) {
187 if (!swr_is_initialized(s->swr)) {
188 ret = opus_init_resample(s);
189 if (ret < 0)
190 return ret;
191 }
192
193 samples = ff_silk_decode_superframe(s->silk, &s->rc, s->silk_output,
194 FFMIN(s->packet.bandwidth, OPUS_BANDWIDTH_WIDEBAND),
195 s->packet.stereo + 1,
196 silk_frame_duration_ms[s->packet.config]);
197 if (samples < 0) {
198 av_log(s->avctx, AV_LOG_ERROR, "Error decoding a SILK frame.\n");
199 return samples;
200 }
201 samples = swr_convert(s->swr,
202 (uint8_t**)s->out, s->packet.frame_duration,
203 (const uint8_t**)s->silk_output, samples);
204 if (samples < 0) {
205 av_log(s->avctx, AV_LOG_ERROR, "Error resampling SILK data.\n");
206 return samples;
207 }
208 av_assert2((samples & 7) == 0);
209 s->delayed_samples += s->packet.frame_duration - samples;
210 } else
211 ff_silk_flush(s->silk);
212
213 // decode redundancy information
214 consumed = opus_rc_tell(&s->rc);
215 if (s->packet.mode == OPUS_MODE_HYBRID && consumed + 37 <= size * 8)
216 redundancy = ff_opus_rc_dec_log(&s->rc, 12);
217 else if (s->packet.mode == OPUS_MODE_SILK && consumed + 17 <= size * 8)
218 redundancy = 1;
219
220 if (redundancy) {
221 redundancy_pos = ff_opus_rc_dec_log(&s->rc, 1);
222
223 if (s->packet.mode == OPUS_MODE_HYBRID)
224 redundancy_size = ff_opus_rc_dec_uint(&s->rc, 256) + 2;
225 else
226 redundancy_size = size - (consumed + 7) / 8;
227 size -= redundancy_size;
228 if (size < 0) {
229 av_log(s->avctx, AV_LOG_ERROR, "Invalid redundancy frame size.\n");
230 return AVERROR_INVALIDDATA;
231 }
232
233 if (redundancy_pos) {
234 ret = opus_decode_redundancy(s, data + size, redundancy_size);
235 if (ret < 0)
236 return ret;
237 ff_celt_flush(s->celt);
238 }
239 }
240
241 /* decode the CELT frame */
242 if (s->packet.mode == OPUS_MODE_CELT || s->packet.mode == OPUS_MODE_HYBRID) {
243 float *out_tmp[2] = { s->out[0], s->out[1] };
244 float **dst = (s->packet.mode == OPUS_MODE_CELT) ?
245 out_tmp : s->celt_output;
246 int celt_output_samples = samples;
247 int delay_samples = av_audio_fifo_size(s->celt_delay);
248
249 if (delay_samples) {
250 if (s->packet.mode == OPUS_MODE_HYBRID) {
251 av_audio_fifo_read(s->celt_delay, (void**)s->celt_output, delay_samples);
252
253 for (i = 0; i < s->output_channels; i++) {
254 s->fdsp->vector_fmac_scalar(out_tmp[i], s->celt_output[i], 1.0,
255 delay_samples);
256 out_tmp[i] += delay_samples;
257 }
258 celt_output_samples -= delay_samples;
259 } else {
260 av_log(s->avctx, AV_LOG_WARNING,
261 "Spurious CELT delay samples present.\n");
262 av_audio_fifo_drain(s->celt_delay, delay_samples);
263 if (s->avctx->err_recognition & AV_EF_EXPLODE)
264 return AVERROR_BUG;
265 }
266 }
267
268 ff_opus_rc_dec_raw_init(&s->rc, data + size, size);
269
270 ret = ff_celt_decode_frame(s->celt, &s->rc, dst,
271 s->packet.stereo + 1,
272 s->packet.frame_duration,
273 (s->packet.mode == OPUS_MODE_HYBRID) ? 17 : 0,
274 ff_celt_band_end[s->packet.bandwidth]);
275 if (ret < 0)
276 return ret;
277
278 if (s->packet.mode == OPUS_MODE_HYBRID) {
279 int celt_delay = s->packet.frame_duration - celt_output_samples;
280 void *delaybuf[2] = { s->celt_output[0] + celt_output_samples,
281 s->celt_output[1] + celt_output_samples };
282
283 for (i = 0; i < s->output_channels; i++) {
284 s->fdsp->vector_fmac_scalar(out_tmp[i],
285 s->celt_output[i], 1.0,
286 celt_output_samples);
287 }
288
289 ret = av_audio_fifo_write(s->celt_delay, delaybuf, celt_delay);
290 if (ret < 0)
291 return ret;
292 }
293 } else
294 ff_celt_flush(s->celt);
295
296 if (s->redundancy_idx) {
297 for (i = 0; i < s->output_channels; i++)
298 opus_fade(s->out[i], s->out[i],
299 s->redundancy_output[i] + 120 + s->redundancy_idx,
300 ff_celt_window2 + s->redundancy_idx, 120 - s->redundancy_idx);
301 s->redundancy_idx = 0;
302 }
303 if (redundancy) {
304 if (!redundancy_pos) {
305 ff_celt_flush(s->celt);
306 ret = opus_decode_redundancy(s, data + size, redundancy_size);
307 if (ret < 0)
308 return ret;
309
310 for (i = 0; i < s->output_channels; i++) {
311 opus_fade(s->out[i] + samples - 120 + delayed_samples,
312 s->out[i] + samples - 120 + delayed_samples,
313 s->redundancy_output[i] + 120,
314 ff_celt_window2, 120 - delayed_samples);
315 if (delayed_samples)
316 s->redundancy_idx = 120 - delayed_samples;
317 }
318 } else {
319 for (i = 0; i < s->output_channels; i++) {
320 memcpy(s->out[i] + delayed_samples, s->redundancy_output[i], 120 * sizeof(float));
321 opus_fade(s->out[i] + 120 + delayed_samples,
322 s->redundancy_output[i] + 120,
323 s->out[i] + 120 + delayed_samples,
324 ff_celt_window2, 120);
325 }
326 }
327 }
328
329 return samples;
330 }
331
opus_decode_subpacket(OpusStreamContext * s,const uint8_t * buf,int buf_size,float ** out,int out_size,int nb_samples)332 static int opus_decode_subpacket(OpusStreamContext *s,
333 const uint8_t *buf, int buf_size,
334 float **out, int out_size,
335 int nb_samples)
336 {
337 int output_samples = 0;
338 int flush_needed = 0;
339 int i, j, ret;
340
341 s->out[0] = out[0];
342 s->out[1] = out[1];
343 s->out_size = out_size;
344
345 /* check if we need to flush the resampler */
346 if (swr_is_initialized(s->swr)) {
347 if (buf) {
348 int64_t cur_samplerate;
349 av_opt_get_int(s->swr, "in_sample_rate", 0, &cur_samplerate);
350 flush_needed = (s->packet.mode == OPUS_MODE_CELT) || (cur_samplerate != s->silk_samplerate);
351 } else {
352 flush_needed = !!s->delayed_samples;
353 }
354 }
355
356 if (!buf && !flush_needed)
357 return 0;
358
359 /* use dummy output buffers if the channel is not mapped to anything */
360 if (!s->out[0] ||
361 (s->output_channels == 2 && !s->out[1])) {
362 av_fast_malloc(&s->out_dummy, &s->out_dummy_allocated_size, s->out_size);
363 if (!s->out_dummy)
364 return AVERROR(ENOMEM);
365 if (!s->out[0])
366 s->out[0] = s->out_dummy;
367 if (!s->out[1])
368 s->out[1] = s->out_dummy;
369 }
370
371 /* flush the resampler if necessary */
372 if (flush_needed) {
373 ret = opus_flush_resample(s, s->delayed_samples);
374 if (ret < 0) {
375 av_log(s->avctx, AV_LOG_ERROR, "Error flushing the resampler.\n");
376 return ret;
377 }
378 swr_close(s->swr);
379 output_samples += s->delayed_samples;
380 s->delayed_samples = 0;
381
382 if (!buf)
383 goto finish;
384 }
385
386 /* decode all the frames in the packet */
387 for (i = 0; i < s->packet.frame_count; i++) {
388 int size = s->packet.frame_size[i];
389 int samples = opus_decode_frame(s, buf + s->packet.frame_offset[i], size);
390
391 if (samples < 0) {
392 av_log(s->avctx, AV_LOG_ERROR, "Error decoding an Opus frame.\n");
393 if (s->avctx->err_recognition & AV_EF_EXPLODE)
394 return samples;
395
396 for (j = 0; j < s->output_channels; j++)
397 memset(s->out[j], 0, s->packet.frame_duration * sizeof(float));
398 samples = s->packet.frame_duration;
399 }
400 output_samples += samples;
401
402 for (j = 0; j < s->output_channels; j++)
403 s->out[j] += samples;
404 s->out_size -= samples * sizeof(float);
405 }
406
407 finish:
408 s->out[0] = s->out[1] = NULL;
409 s->out_size = 0;
410
411 return output_samples;
412 }
413
opus_decode_packet(AVCodecContext * avctx,void * data,int * got_frame_ptr,AVPacket * avpkt)414 static int opus_decode_packet(AVCodecContext *avctx, void *data,
415 int *got_frame_ptr, AVPacket *avpkt)
416 {
417 OpusContext *c = avctx->priv_data;
418 AVFrame *frame = data;
419 const uint8_t *buf = avpkt->data;
420 int buf_size = avpkt->size;
421 int coded_samples = 0;
422 int decoded_samples = INT_MAX;
423 int delayed_samples = 0;
424 int i, ret;
425
426 /* calculate the number of delayed samples */
427 for (i = 0; i < c->nb_streams; i++) {
428 OpusStreamContext *s = &c->streams[i];
429 s->out[0] =
430 s->out[1] = NULL;
431 delayed_samples = FFMAX(delayed_samples,
432 s->delayed_samples + av_audio_fifo_size(c->sync_buffers[i]));
433 }
434
435 /* decode the header of the first sub-packet to find out the sample count */
436 if (buf) {
437 OpusPacket *pkt = &c->streams[0].packet;
438 ret = ff_opus_parse_packet(pkt, buf, buf_size, c->nb_streams > 1);
439 if (ret < 0) {
440 av_log(avctx, AV_LOG_ERROR, "Error parsing the packet header.\n");
441 return ret;
442 }
443 coded_samples += pkt->frame_count * pkt->frame_duration;
444 c->streams[0].silk_samplerate = get_silk_samplerate(pkt->config);
445 }
446
447 frame->nb_samples = coded_samples + delayed_samples;
448
449 /* no input or buffered data => nothing to do */
450 if (!frame->nb_samples) {
451 *got_frame_ptr = 0;
452 return 0;
453 }
454
455 /* setup the data buffers */
456 ret = ff_get_buffer(avctx, frame, 0);
457 if (ret < 0)
458 return ret;
459 frame->nb_samples = 0;
460
461 memset(c->out, 0, c->nb_streams * 2 * sizeof(*c->out));
462 for (i = 0; i < avctx->channels; i++) {
463 ChannelMap *map = &c->channel_maps[i];
464 if (!map->copy)
465 c->out[2 * map->stream_idx + map->channel_idx] = (float*)frame->extended_data[i];
466 }
467
468 /* read the data from the sync buffers */
469 for (i = 0; i < c->nb_streams; i++) {
470 float **out = c->out + 2 * i;
471 int sync_size = av_audio_fifo_size(c->sync_buffers[i]);
472
473 float sync_dummy[32];
474 int out_dummy = (!out[0]) | ((!out[1]) << 1);
475
476 if (!out[0])
477 out[0] = sync_dummy;
478 if (!out[1])
479 out[1] = sync_dummy;
480 if (out_dummy && sync_size > FF_ARRAY_ELEMS(sync_dummy))
481 return AVERROR_BUG;
482
483 ret = av_audio_fifo_read(c->sync_buffers[i], (void**)out, sync_size);
484 if (ret < 0)
485 return ret;
486
487 if (out_dummy & 1)
488 out[0] = NULL;
489 else
490 out[0] += ret;
491 if (out_dummy & 2)
492 out[1] = NULL;
493 else
494 out[1] += ret;
495
496 c->out_size[i] = frame->linesize[0] - ret * sizeof(float);
497 }
498
499 /* decode each sub-packet */
500 for (i = 0; i < c->nb_streams; i++) {
501 OpusStreamContext *s = &c->streams[i];
502
503 if (i && buf) {
504 ret = ff_opus_parse_packet(&s->packet, buf, buf_size, i != c->nb_streams - 1);
505 if (ret < 0) {
506 av_log(avctx, AV_LOG_ERROR, "Error parsing the packet header.\n");
507 return ret;
508 }
509 if (coded_samples != s->packet.frame_count * s->packet.frame_duration) {
510 av_log(avctx, AV_LOG_ERROR,
511 "Mismatching coded sample count in substream %d.\n", i);
512 return AVERROR_INVALIDDATA;
513 }
514
515 s->silk_samplerate = get_silk_samplerate(s->packet.config);
516 }
517
518 ret = opus_decode_subpacket(&c->streams[i], buf, s->packet.data_size,
519 c->out + 2 * i, c->out_size[i], coded_samples);
520 if (ret < 0)
521 return ret;
522 c->decoded_samples[i] = ret;
523 decoded_samples = FFMIN(decoded_samples, ret);
524
525 buf += s->packet.packet_size;
526 buf_size -= s->packet.packet_size;
527 }
528
529 /* buffer the extra samples */
530 for (i = 0; i < c->nb_streams; i++) {
531 int buffer_samples = c->decoded_samples[i] - decoded_samples;
532 if (buffer_samples) {
533 float *buf[2] = { c->out[2 * i + 0] ? c->out[2 * i + 0] : (float*)frame->extended_data[0],
534 c->out[2 * i + 1] ? c->out[2 * i + 1] : (float*)frame->extended_data[0] };
535 buf[0] += decoded_samples;
536 buf[1] += decoded_samples;
537 ret = av_audio_fifo_write(c->sync_buffers[i], (void**)buf, buffer_samples);
538 if (ret < 0)
539 return ret;
540 }
541 }
542
543 for (i = 0; i < avctx->channels; i++) {
544 ChannelMap *map = &c->channel_maps[i];
545
546 /* handle copied channels */
547 if (map->copy) {
548 memcpy(frame->extended_data[i],
549 frame->extended_data[map->copy_idx],
550 frame->linesize[0]);
551 } else if (map->silence) {
552 memset(frame->extended_data[i], 0, frame->linesize[0]);
553 }
554
555 if (c->gain_i && decoded_samples > 0) {
556 c->fdsp->vector_fmul_scalar((float*)frame->extended_data[i],
557 (float*)frame->extended_data[i],
558 c->gain, FFALIGN(decoded_samples, 8));
559 }
560 }
561
562 frame->nb_samples = decoded_samples;
563 *got_frame_ptr = !!decoded_samples;
564
565 return avpkt->size;
566 }
567
opus_decode_flush(AVCodecContext * ctx)568 static av_cold void opus_decode_flush(AVCodecContext *ctx)
569 {
570 OpusContext *c = ctx->priv_data;
571 int i;
572
573 for (i = 0; i < c->nb_streams; i++) {
574 OpusStreamContext *s = &c->streams[i];
575
576 memset(&s->packet, 0, sizeof(s->packet));
577 s->delayed_samples = 0;
578
579 if (s->celt_delay)
580 av_audio_fifo_drain(s->celt_delay, av_audio_fifo_size(s->celt_delay));
581 swr_close(s->swr);
582
583 av_audio_fifo_drain(c->sync_buffers[i], av_audio_fifo_size(c->sync_buffers[i]));
584
585 ff_silk_flush(s->silk);
586 ff_celt_flush(s->celt);
587 }
588 }
589
opus_decode_close(AVCodecContext * avctx)590 static av_cold int opus_decode_close(AVCodecContext *avctx)
591 {
592 OpusContext *c = avctx->priv_data;
593 int i;
594
595 for (i = 0; i < c->nb_streams; i++) {
596 OpusStreamContext *s = &c->streams[i];
597
598 ff_silk_free(&s->silk);
599 ff_celt_free(&s->celt);
600
601 av_freep(&s->out_dummy);
602 s->out_dummy_allocated_size = 0;
603
604 av_audio_fifo_free(s->celt_delay);
605 swr_free(&s->swr);
606 }
607
608 av_freep(&c->streams);
609
610 if (c->sync_buffers) {
611 for (i = 0; i < c->nb_streams; i++)
612 av_audio_fifo_free(c->sync_buffers[i]);
613 }
614 av_freep(&c->sync_buffers);
615 av_freep(&c->decoded_samples);
616 av_freep(&c->out);
617 av_freep(&c->out_size);
618
619 c->nb_streams = 0;
620
621 av_freep(&c->channel_maps);
622 av_freep(&c->fdsp);
623
624 return 0;
625 }
626
opus_decode_init(AVCodecContext * avctx)627 static av_cold int opus_decode_init(AVCodecContext *avctx)
628 {
629 OpusContext *c = avctx->priv_data;
630 int ret, i, j;
631
632 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
633 avctx->sample_rate = 48000;
634
635 c->fdsp = avpriv_float_dsp_alloc(0);
636 if (!c->fdsp)
637 return AVERROR(ENOMEM);
638
639 /* find out the channel configuration */
640 ret = ff_opus_parse_extradata(avctx, c);
641 if (ret < 0) {
642 av_freep(&c->fdsp);
643 return ret;
644 }
645
646 /* allocate and init each independent decoder */
647 c->streams = av_mallocz_array(c->nb_streams, sizeof(*c->streams));
648 c->out = av_mallocz_array(c->nb_streams, 2 * sizeof(*c->out));
649 c->out_size = av_mallocz_array(c->nb_streams, sizeof(*c->out_size));
650 c->sync_buffers = av_mallocz_array(c->nb_streams, sizeof(*c->sync_buffers));
651 c->decoded_samples = av_mallocz_array(c->nb_streams, sizeof(*c->decoded_samples));
652 if (!c->streams || !c->sync_buffers || !c->decoded_samples || !c->out || !c->out_size) {
653 c->nb_streams = 0;
654 ret = AVERROR(ENOMEM);
655 goto fail;
656 }
657
658 for (i = 0; i < c->nb_streams; i++) {
659 OpusStreamContext *s = &c->streams[i];
660 uint64_t layout;
661
662 s->output_channels = (i < c->nb_stereo_streams) ? 2 : 1;
663
664 s->avctx = avctx;
665
666 for (j = 0; j < s->output_channels; j++) {
667 s->silk_output[j] = s->silk_buf[j];
668 s->celt_output[j] = s->celt_buf[j];
669 s->redundancy_output[j] = s->redundancy_buf[j];
670 }
671
672 s->fdsp = c->fdsp;
673
674 s->swr =swr_alloc();
675 if (!s->swr)
676 goto fail;
677
678 layout = (s->output_channels == 1) ? AV_CH_LAYOUT_MONO : AV_CH_LAYOUT_STEREO;
679 av_opt_set_int(s->swr, "in_sample_fmt", avctx->sample_fmt, 0);
680 av_opt_set_int(s->swr, "out_sample_fmt", avctx->sample_fmt, 0);
681 av_opt_set_int(s->swr, "in_channel_layout", layout, 0);
682 av_opt_set_int(s->swr, "out_channel_layout", layout, 0);
683 av_opt_set_int(s->swr, "out_sample_rate", avctx->sample_rate, 0);
684 av_opt_set_int(s->swr, "filter_size", 16, 0);
685
686 ret = ff_silk_init(avctx, &s->silk, s->output_channels);
687 if (ret < 0)
688 goto fail;
689
690 ret = ff_celt_init(avctx, &s->celt, s->output_channels, c->apply_phase_inv);
691 if (ret < 0)
692 goto fail;
693
694 s->celt_delay = av_audio_fifo_alloc(avctx->sample_fmt,
695 s->output_channels, 1024);
696 if (!s->celt_delay) {
697 ret = AVERROR(ENOMEM);
698 goto fail;
699 }
700
701 c->sync_buffers[i] = av_audio_fifo_alloc(avctx->sample_fmt,
702 s->output_channels, 32);
703 if (!c->sync_buffers[i]) {
704 ret = AVERROR(ENOMEM);
705 goto fail;
706 }
707 }
708
709 return 0;
710 fail:
711 opus_decode_close(avctx);
712 return ret;
713 }
714
715 #define OFFSET(x) offsetof(OpusContext, x)
716 #define AD AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_DECODING_PARAM
717 static const AVOption opus_options[] = {
718 { "apply_phase_inv", "Apply intensity stereo phase inversion", OFFSET(apply_phase_inv), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, AD },
719 { NULL },
720 };
721
722 static const AVClass opus_class = {
723 .class_name = "Opus Decoder",
724 .item_name = av_default_item_name,
725 .option = opus_options,
726 .version = LIBAVUTIL_VERSION_INT,
727 };
728
729 AVCodec ff_opus_decoder = {
730 .name = "opus",
731 .long_name = NULL_IF_CONFIG_SMALL("Opus"),
732 .priv_class = &opus_class,
733 .type = AVMEDIA_TYPE_AUDIO,
734 .id = AV_CODEC_ID_OPUS,
735 .priv_data_size = sizeof(OpusContext),
736 .init = opus_decode_init,
737 .close = opus_decode_close,
738 .decode = opus_decode_packet,
739 .flush = opus_decode_flush,
740 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY,
741 };
742