FFmpeg
opusenc.c
Go to the documentation of this file.
1 /*
2  * Opus encoder
3  * Copyright (c) 2017 Rostislav Pehlivanov <atomnuker@gmail.com>
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 #include <float.h>
23 
24 #include "encode.h"
25 #include "opusenc.h"
26 #include "opus_pvq.h"
27 #include "opusenc_psy.h"
28 #include "opustab.h"
29 
31 #include "libavutil/float_dsp.h"
32 #include "libavutil/mem_internal.h"
33 #include "libavutil/opt.h"
34 #include "bytestream.h"
35 #include "audio_frame_queue.h"
36 #include "codec_internal.h"
37 
38 typedef struct OpusEncContext {
49 
50  uint8_t enc_id[64];
52 
54 
55  int channels;
56 
59 
60  /* Actual energy the decoder will have */
62 
63  DECLARE_ALIGNED(32, float, scratch)[2048];
65 
67 {
68  uint8_t *bs = avctx->extradata;
69 
70  bytestream_put_buffer(&bs, "OpusHead", 8);
71  bytestream_put_byte (&bs, 0x1);
72  bytestream_put_byte (&bs, avctx->ch_layout.nb_channels);
73  bytestream_put_le16 (&bs, avctx->initial_padding);
74  bytestream_put_le32 (&bs, avctx->sample_rate);
75  bytestream_put_le16 (&bs, 0x0);
76  bytestream_put_byte (&bs, 0x0); /* Default layout */
77 }
78 
79 static int opus_gen_toc(OpusEncContext *s, uint8_t *toc, int *size, int *fsize_needed)
80 {
81  int tmp = 0x0, extended_toc = 0;
82  static const int toc_cfg[][OPUS_MODE_NB][OPUS_BANDWITH_NB] = {
83  /* Silk Hybrid Celt Layer */
84  /* NB MB WB SWB FB NB MB WB SWB FB NB MB WB SWB FB Bandwidth */
85  { { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 }, { 17, 0, 21, 25, 29 } }, /* 2.5 ms */
86  { { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 }, { 18, 0, 22, 26, 30 } }, /* 5 ms */
87  { { 1, 5, 9, 0, 0 }, { 0, 0, 0, 13, 15 }, { 19, 0, 23, 27, 31 } }, /* 10 ms */
88  { { 2, 6, 10, 0, 0 }, { 0, 0, 0, 14, 16 }, { 20, 0, 24, 28, 32 } }, /* 20 ms */
89  { { 3, 7, 11, 0, 0 }, { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }, /* 40 ms */
90  { { 4, 8, 12, 0, 0 }, { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }, /* 60 ms */
91  };
92  int cfg = toc_cfg[s->packet.framesize][s->packet.mode][s->packet.bandwidth];
93  *fsize_needed = 0;
94  if (!cfg)
95  return 1;
96  if (s->packet.frames == 2) { /* 2 packets */
97  if (s->frame[0].framebits == s->frame[1].framebits) { /* same size */
98  tmp = 0x1;
99  } else { /* different size */
100  tmp = 0x2;
101  *fsize_needed = 1; /* put frame sizes in the packet */
102  }
103  } else if (s->packet.frames > 2) {
104  tmp = 0x3;
105  extended_toc = 1;
106  }
107  tmp |= (s->channels > 1) << 2; /* Stereo or mono */
108  tmp |= (cfg - 1) << 3; /* codec configuration */
109  *toc++ = tmp;
110  if (extended_toc) {
111  for (int i = 0; i < (s->packet.frames - 1); i++)
112  *fsize_needed |= (s->frame[i].framebits != s->frame[i + 1].framebits);
113  tmp = (*fsize_needed) << 7; /* vbr flag */
114  tmp |= (0) << 6; /* padding flag */
115  tmp |= s->packet.frames;
116  *toc++ = tmp;
117  }
118  *size = 1 + extended_toc;
119  return 0;
120 }
121 
123 {
124  AVFrame *cur = NULL;
125  const int subframesize = s->avctx->frame_size;
126  int subframes = OPUS_BLOCK_SIZE(s->packet.framesize) / subframesize;
127 
128  cur = ff_bufqueue_get(&s->bufqueue);
129 
130  for (int ch = 0; ch < f->channels; ch++) {
131  CeltBlock *b = &f->block[ch];
132  const void *input = cur->extended_data[ch];
133  size_t bps = av_get_bytes_per_sample(cur->format);
134  memcpy(b->overlap, input, bps*cur->nb_samples);
135  }
136 
137  av_frame_free(&cur);
138 
139  for (int sf = 0; sf < subframes; sf++) {
140  if (sf != (subframes - 1))
141  cur = ff_bufqueue_get(&s->bufqueue);
142  else
143  cur = ff_bufqueue_peek(&s->bufqueue, 0);
144 
145  for (int ch = 0; ch < f->channels; ch++) {
146  CeltBlock *b = &f->block[ch];
147  const void *input = cur->extended_data[ch];
148  const size_t bps = av_get_bytes_per_sample(cur->format);
149  const size_t left = (subframesize - cur->nb_samples)*bps;
150  const size_t len = FFMIN(subframesize, cur->nb_samples)*bps;
151  memcpy(&b->samples[sf*subframesize], input, len);
152  memset(&b->samples[cur->nb_samples], 0, left);
153  }
154 
155  /* Last frame isn't popped off and freed yet - we need it for overlap */
156  if (sf != (subframes - 1))
157  av_frame_free(&cur);
158  }
159 }
160 
161 /* Apply the pre emphasis filter */
163 {
164  const int subframesize = s->avctx->frame_size;
165  const int subframes = OPUS_BLOCK_SIZE(s->packet.framesize) / subframesize;
166 
167  /* Filter overlap */
168  for (int ch = 0; ch < f->channels; ch++) {
169  CeltBlock *b = &f->block[ch];
170  float m = b->emph_coeff;
171  for (int i = 0; i < CELT_OVERLAP; i++) {
172  float sample = b->overlap[i];
173  b->overlap[i] = sample - m;
174  m = sample * CELT_EMPH_COEFF;
175  }
176  b->emph_coeff = m;
177  }
178 
179  /* Filter the samples but do not update the last subframe's coeff - overlap ^^^ */
180  for (int sf = 0; sf < subframes; sf++) {
181  for (int ch = 0; ch < f->channels; ch++) {
182  CeltBlock *b = &f->block[ch];
183  float m = b->emph_coeff;
184  for (int i = 0; i < subframesize; i++) {
185  float sample = b->samples[sf*subframesize + i];
186  b->samples[sf*subframesize + i] = sample - m;
187  m = sample * CELT_EMPH_COEFF;
188  }
189  if (sf != (subframes - 1))
190  b->emph_coeff = m;
191  }
192  }
193 }
194 
195 /* Create the window and do the mdct */
197 {
198  float *win = s->scratch, *temp = s->scratch + 1920;
199 
200  if (f->transient) {
201  for (int ch = 0; ch < f->channels; ch++) {
202  CeltBlock *b = &f->block[ch];
203  float *src1 = b->overlap;
204  for (int t = 0; t < f->blocks; t++) {
205  float *src2 = &b->samples[CELT_OVERLAP*t];
206  s->dsp->vector_fmul(win, src1, ff_celt_window, 128);
207  s->dsp->vector_fmul_reverse(&win[CELT_OVERLAP], src2,
208  ff_celt_window_padded, 128);
209  src1 = src2;
210  s->tx_fn[0](s->tx[0], b->coeffs + t, win, sizeof(float)*f->blocks);
211  }
212  }
213  } else {
214  int blk_len = OPUS_BLOCK_SIZE(f->size), wlen = OPUS_BLOCK_SIZE(f->size + 1);
215  int rwin = blk_len - CELT_OVERLAP, lap_dst = (wlen - blk_len - CELT_OVERLAP) >> 1;
216  memset(win, 0, wlen*sizeof(float));
217  for (int ch = 0; ch < f->channels; ch++) {
218  CeltBlock *b = &f->block[ch];
219 
220  /* Overlap */
221  s->dsp->vector_fmul(temp, b->overlap, ff_celt_window, 128);
222  memcpy(win + lap_dst, temp, CELT_OVERLAP*sizeof(float));
223 
224  /* Samples, flat top window */
225  memcpy(&win[lap_dst + CELT_OVERLAP], b->samples, rwin*sizeof(float));
226 
227  /* Samples, windowed */
228  s->dsp->vector_fmul_reverse(temp, b->samples + rwin,
229  ff_celt_window_padded, 128);
230  memcpy(win + lap_dst + blk_len, temp, CELT_OVERLAP*sizeof(float));
231 
232  s->tx_fn[f->size](s->tx[f->size], b->coeffs, win, sizeof(float));
233  }
234  }
235 
236  for (int ch = 0; ch < f->channels; ch++) {
237  CeltBlock *block = &f->block[ch];
238  for (int i = 0; i < CELT_MAX_BANDS; i++) {
239  float ener = 0.0f;
240  int band_offset = ff_celt_freq_bands[i] << f->size;
241  int band_size = ff_celt_freq_range[i] << f->size;
242  float *coeffs = &block->coeffs[band_offset];
243 
244  for (int j = 0; j < band_size; j++)
245  ener += coeffs[j]*coeffs[j];
246 
247  block->lin_energy[i] = sqrtf(ener) + FLT_EPSILON;
248  ener = 1.0f/block->lin_energy[i];
249 
250  for (int j = 0; j < band_size; j++)
251  coeffs[j] *= ener;
252 
253  block->energy[i] = log2f(block->lin_energy[i]) - ff_celt_mean_energy[i];
254 
255  /* CELT_ENERGY_SILENCE is what the decoder uses and its not -infinity */
256  block->energy[i] = FFMAX(block->energy[i], CELT_ENERGY_SILENCE);
257  }
258  }
259 }
260 
262 {
263  int tf_select = 0, diff = 0, tf_changed = 0, tf_select_needed;
264  int bits = f->transient ? 2 : 4;
265 
266  tf_select_needed = ((f->size && (opus_rc_tell(rc) + bits + 1) <= f->framebits));
267 
268  for (int i = f->start_band; i < f->end_band; i++) {
269  if ((opus_rc_tell(rc) + bits + tf_select_needed) <= f->framebits) {
270  const int tbit = (diff ^ 1) == f->tf_change[i];
271  ff_opus_rc_enc_log(rc, tbit, bits);
272  diff ^= tbit;
273  tf_changed |= diff;
274  }
275  bits = f->transient ? 4 : 5;
276  }
277 
278  if (tf_select_needed && ff_celt_tf_select[f->size][f->transient][0][tf_changed] !=
279  ff_celt_tf_select[f->size][f->transient][1][tf_changed]) {
280  ff_opus_rc_enc_log(rc, f->tf_select, 1);
281  tf_select = f->tf_select;
282  }
283 
284  for (int i = f->start_band; i < f->end_band; i++)
285  f->tf_change[i] = ff_celt_tf_select[f->size][f->transient][tf_select][f->tf_change[i]];
286 }
287 
289 {
290  float gain = f->pf_gain;
291  int txval, octave = f->pf_octave, period = f->pf_period, tapset = f->pf_tapset;
292 
293  ff_opus_rc_enc_log(rc, f->pfilter, 1);
294  if (!f->pfilter)
295  return;
296 
297  /* Octave */
298  txval = FFMIN(octave, 6);
299  ff_opus_rc_enc_uint(rc, txval, 6);
300  octave = txval;
301  /* Period */
302  txval = av_clip(period - (16 << octave) + 1, 0, (1 << (4 + octave)) - 1);
303  ff_opus_rc_put_raw(rc, period, 4 + octave);
304  period = txval + (16 << octave) - 1;
305  /* Gain */
306  txval = FFMIN(((int)(gain / 0.09375f)) - 1, 7);
307  ff_opus_rc_put_raw(rc, txval, 3);
308  gain = 0.09375f * (txval + 1);
309  /* Tapset */
310  if ((opus_rc_tell(rc) + 2) <= f->framebits)
312  else
313  tapset = 0;
314  /* Finally create the coeffs */
315  for (int i = 0; i < 2; i++) {
316  CeltBlock *block = &f->block[i];
317 
318  block->pf_period_new = FFMAX(period, CELT_POSTFILTER_MINPERIOD);
319  block->pf_gains_new[0] = gain * ff_celt_postfilter_taps[tapset][0];
320  block->pf_gains_new[1] = gain * ff_celt_postfilter_taps[tapset][1];
321  block->pf_gains_new[2] = gain * ff_celt_postfilter_taps[tapset][2];
322  }
323 }
324 
326  float last_energy[][CELT_MAX_BANDS], int intra)
327 {
328  float alpha, beta, prev[2] = { 0, 0 };
329  const uint8_t *pmod = ff_celt_coarse_energy_dist[f->size][intra];
330 
331  /* Inter is really just differential coding */
332  if (opus_rc_tell(rc) + 3 <= f->framebits)
333  ff_opus_rc_enc_log(rc, intra, 3);
334  else
335  intra = 0;
336 
337  if (intra) {
338  alpha = 0.0f;
339  beta = 1.0f - (4915.0f/32768.0f);
340  } else {
341  alpha = ff_celt_alpha_coef[f->size];
342  beta = ff_celt_beta_coef[f->size];
343  }
344 
345  for (int i = f->start_band; i < f->end_band; i++) {
346  for (int ch = 0; ch < f->channels; ch++) {
347  CeltBlock *block = &f->block[ch];
348  const int left = f->framebits - opus_rc_tell(rc);
349  const float last = FFMAX(-9.0f, last_energy[ch][i]);
350  float diff = block->energy[i] - prev[ch] - last*alpha;
351  int q_en = lrintf(diff);
352  if (left >= 15) {
353  ff_opus_rc_enc_laplace(rc, &q_en, pmod[i << 1] << 7, pmod[(i << 1) + 1] << 6);
354  } else if (left >= 2) {
355  q_en = av_clip(q_en, -1, 1);
356  ff_opus_rc_enc_cdf(rc, 2*q_en + 3*(q_en < 0), ff_celt_model_energy_small);
357  } else if (left >= 1) {
358  q_en = av_clip(q_en, -1, 0);
359  ff_opus_rc_enc_log(rc, (q_en & 1), 1);
360  } else q_en = -1;
361 
362  block->error_energy[i] = q_en - diff;
363  prev[ch] += beta * q_en;
364  }
365  }
366 }
367 
369  float last_energy[][CELT_MAX_BANDS])
370 {
371  uint32_t inter, intra;
373 
374  exp_quant_coarse(rc, f, last_energy, 1);
375  intra = OPUS_RC_CHECKPOINT_BITS(rc);
376 
378 
379  exp_quant_coarse(rc, f, last_energy, 0);
380  inter = OPUS_RC_CHECKPOINT_BITS(rc);
381 
382  if (inter > intra) { /* Unlikely */
384  exp_quant_coarse(rc, f, last_energy, 1);
385  }
386 }
387 
389 {
390  for (int i = f->start_band; i < f->end_band; i++) {
391  if (!f->fine_bits[i])
392  continue;
393  for (int ch = 0; ch < f->channels; ch++) {
394  CeltBlock *block = &f->block[ch];
395  int quant, lim = (1 << f->fine_bits[i]);
396  float offset, diff = 0.5f - block->error_energy[i];
397  quant = av_clip(floor(diff*lim), 0, lim - 1);
398  ff_opus_rc_put_raw(rc, quant, f->fine_bits[i]);
399  offset = 0.5f - ((quant + 0.5f) * (1 << (14 - f->fine_bits[i])) / 16384.0f);
400  block->error_energy[i] -= offset;
401  }
402  }
403 }
404 
406 {
407  for (int priority = 0; priority < 2; priority++) {
408  for (int i = f->start_band; i < f->end_band && (f->framebits - opus_rc_tell(rc)) >= f->channels; i++) {
409  if (f->fine_priority[i] != priority || f->fine_bits[i] >= CELT_MAX_FINE_BITS)
410  continue;
411  for (int ch = 0; ch < f->channels; ch++) {
412  CeltBlock *block = &f->block[ch];
413  const float err = block->error_energy[i];
414  const float offset = 0.5f * (1 << (14 - f->fine_bits[i] - 1)) / 16384.0f;
415  const int sign = FFABS(err + offset) < FFABS(err - offset);
416  ff_opus_rc_put_raw(rc, sign, 1);
417  block->error_energy[i] -= offset*(1 - 2*sign);
418  }
419  }
420  }
421 }
422 
424  CeltFrame *f, int index)
425 {
427 
428  ff_opus_psy_celt_frame_init(&s->psyctx, f, index);
429 
431 
432  if (f->silence) {
433  if (f->framebits >= 16)
434  ff_opus_rc_enc_log(rc, 1, 15); /* Silence (if using explicit singalling) */
435  for (int ch = 0; ch < s->channels; ch++)
436  memset(s->last_quantized_energy[ch], 0.0f, sizeof(float)*CELT_MAX_BANDS);
437  return;
438  }
439 
440  /* Filters */
442  if (f->pfilter) {
443  ff_opus_rc_enc_log(rc, 0, 15);
445  }
446 
447  /* Transform */
448  celt_frame_mdct(s, f);
449 
450  /* Need to handle transient/non-transient switches at any point during analysis */
451  while (ff_opus_psy_celt_frame_process(&s->psyctx, f, index))
452  celt_frame_mdct(s, f);
453 
455 
456  /* Silence */
457  ff_opus_rc_enc_log(rc, 0, 15);
458 
459  /* Pitch filter */
460  if (!f->start_band && opus_rc_tell(rc) + 16 <= f->framebits)
462 
463  /* Transient flag */
464  if (f->size && opus_rc_tell(rc) + 3 <= f->framebits)
465  ff_opus_rc_enc_log(rc, f->transient, 3);
466 
467  /* Main encoding */
468  celt_quant_coarse (f, rc, s->last_quantized_energy);
469  celt_enc_tf (f, rc);
470  ff_celt_bitalloc (f, rc, 1);
471  celt_quant_fine (f, rc);
472  ff_celt_quant_bands(f, rc);
473 
474  /* Anticollapse bit */
475  if (f->anticollapse_needed)
476  ff_opus_rc_put_raw(rc, f->anticollapse, 1);
477 
478  /* Final per-band energy adjustments from leftover bits */
479  celt_quant_final(s, rc, f);
480 
481  for (int ch = 0; ch < f->channels; ch++) {
482  CeltBlock *block = &f->block[ch];
483  for (int i = 0; i < CELT_MAX_BANDS; i++)
484  s->last_quantized_energy[ch][i] = block->energy[i] + block->error_energy[i];
485  }
486 }
487 
488 static inline int write_opuslacing(uint8_t *dst, int v)
489 {
490  dst[0] = FFMIN(v - FFALIGN(v - 255, 4), v);
491  dst[1] = v - dst[0] >> 2;
492  return 1 + (v >= 252);
493 }
494 
496 {
497  int offset, fsize_needed;
498 
499  /* Write toc */
500  opus_gen_toc(s, avpkt->data, &offset, &fsize_needed);
501 
502  /* Frame sizes if needed */
503  if (fsize_needed) {
504  for (int i = 0; i < s->packet.frames - 1; i++) {
505  offset += write_opuslacing(avpkt->data + offset,
506  s->frame[i].framebits >> 3);
507  }
508  }
509 
510  /* Packets */
511  for (int i = 0; i < s->packet.frames; i++) {
512  ff_opus_rc_enc_end(&s->rc[i], avpkt->data + offset,
513  s->frame[i].framebits >> 3);
514  offset += s->frame[i].framebits >> 3;
515  }
516 
517  avpkt->size = offset;
518 }
519 
520 /* Used as overlap for the first frame and padding for the last encoded packet */
522 {
523  AVFrame *f = av_frame_alloc();
524  int ret;
525  if (!f)
526  return NULL;
527  f->format = s->avctx->sample_fmt;
528  f->nb_samples = s->avctx->frame_size;
529  ret = av_channel_layout_copy(&f->ch_layout, &s->avctx->ch_layout);
530  if (ret < 0) {
531  av_frame_free(&f);
532  return NULL;
533  }
534  if (av_frame_get_buffer(f, 4)) {
535  av_frame_free(&f);
536  return NULL;
537  }
538  for (int i = 0; i < s->channels; i++) {
539  size_t bps = av_get_bytes_per_sample(f->format);
540  memset(f->extended_data[i], 0, bps*f->nb_samples);
541  }
542  return f;
543 }
544 
545 static int opus_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
546  const AVFrame *frame, int *got_packet_ptr)
547 {
548  OpusEncContext *s = avctx->priv_data;
549  int ret, frame_size, alloc_size = 0;
550 
551  if (frame) { /* Add new frame to queue */
552  if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
553  return ret;
554  ff_bufqueue_add(avctx, &s->bufqueue, av_frame_clone(frame));
555  } else {
556  ff_opus_psy_signal_eof(&s->psyctx);
557  if (!s->afq.remaining_samples || !avctx->frame_number)
558  return 0; /* We've been flushed and there's nothing left to encode */
559  }
560 
561  /* Run the psychoacoustic system */
562  if (ff_opus_psy_process(&s->psyctx, &s->packet))
563  return 0;
564 
565  frame_size = OPUS_BLOCK_SIZE(s->packet.framesize);
566 
567  if (!frame) {
568  /* This can go negative, that's not a problem, we only pad if positive */
569  int pad_empty = s->packet.frames*(frame_size/s->avctx->frame_size) - s->bufqueue.available + 1;
570  /* Pad with empty 2.5 ms frames to whatever framesize was decided,
571  * this should only happen at the very last flush frame. The frames
572  * allocated here will be freed (because they have no other references)
573  * after they get used by celt_frame_setup_input() */
574  for (int i = 0; i < pad_empty; i++) {
575  AVFrame *empty = spawn_empty_frame(s);
576  if (!empty)
577  return AVERROR(ENOMEM);
578  ff_bufqueue_add(avctx, &s->bufqueue, empty);
579  }
580  }
581 
582  for (int i = 0; i < s->packet.frames; i++) {
583  celt_encode_frame(s, &s->rc[i], &s->frame[i], i);
584  alloc_size += s->frame[i].framebits >> 3;
585  }
586 
587  /* Worst case toc + the frame lengths if needed */
588  alloc_size += 2 + s->packet.frames*2;
589 
590  if ((ret = ff_alloc_packet(avctx, avpkt, alloc_size)) < 0)
591  return ret;
592 
593  /* Assemble packet */
594  opus_packet_assembler(s, avpkt);
595 
596  /* Update the psychoacoustic system */
597  ff_opus_psy_postencode_update(&s->psyctx, s->frame);
598 
599  /* Remove samples from queue and skip if needed */
600  ff_af_queue_remove(&s->afq, s->packet.frames*frame_size, &avpkt->pts, &avpkt->duration);
601  if (s->packet.frames*frame_size > avpkt->duration) {
602  uint8_t *side = av_packet_new_side_data(avpkt, AV_PKT_DATA_SKIP_SAMPLES, 10);
603  if (!side)
604  return AVERROR(ENOMEM);
605  AV_WL32(&side[4], s->packet.frames*frame_size - avpkt->duration + 120);
606  }
607 
608  *got_packet_ptr = 1;
609 
610  return 0;
611 }
612 
614 {
615  OpusEncContext *s = avctx->priv_data;
616 
617  for (int i = 0; i < CELT_BLOCK_NB; i++)
618  av_tx_uninit(&s->tx[i]);
619 
620  ff_celt_pvq_uninit(&s->pvq);
621  av_freep(&s->dsp);
622  av_freep(&s->frame);
623  av_freep(&s->rc);
624  ff_af_queue_close(&s->afq);
625  ff_opus_psy_end(&s->psyctx);
626  ff_bufqueue_discard_all(&s->bufqueue);
627 
628  return 0;
629 }
630 
632 {
633  int ret, max_frames;
634  OpusEncContext *s = avctx->priv_data;
635 
636  s->avctx = avctx;
637  s->channels = avctx->ch_layout.nb_channels;
638 
639  /* Opus allows us to change the framesize on each packet (and each packet may
640  * have multiple frames in it) but we can't change the codec's frame size on
641  * runtime, so fix it to the lowest possible number of samples and use a queue
642  * to accumulate AVFrames until we have enough to encode whatever the encoder
643  * decides is the best */
644  avctx->frame_size = 120;
645  /* Initial padding will change if SILK is ever supported */
646  avctx->initial_padding = 120;
647 
648  if (!avctx->bit_rate) {
649  int coupled = ff_opus_default_coupled_streams[s->channels - 1];
650  avctx->bit_rate = coupled*(96000) + (s->channels - coupled*2)*(48000);
651  } else if (avctx->bit_rate < 6000 || avctx->bit_rate > 255000 * s->channels) {
652  int64_t clipped_rate = av_clip(avctx->bit_rate, 6000, 255000 * s->channels);
653  av_log(avctx, AV_LOG_ERROR, "Unsupported bitrate %"PRId64" kbps, clipping to %"PRId64" kbps\n",
654  avctx->bit_rate/1000, clipped_rate/1000);
655  avctx->bit_rate = clipped_rate;
656  }
657 
658  /* Extradata */
659  avctx->extradata_size = 19;
661  if (!avctx->extradata)
662  return AVERROR(ENOMEM);
663  opus_write_extradata(avctx);
664 
665  ff_af_queue_init(avctx, &s->afq);
666 
667  if ((ret = ff_celt_pvq_init(&s->pvq, 1)) < 0)
668  return ret;
669 
670  if (!(s->dsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT)))
671  return AVERROR(ENOMEM);
672 
673  /* I have no idea why a base scaling factor of 68 works, could be the twiddles */
674  for (int i = 0; i < CELT_BLOCK_NB; i++) {
675  const float scale = 68 << (CELT_BLOCK_NB - 1 - i);
676  if ((ret = av_tx_init(&s->tx[i], &s->tx_fn[i], AV_TX_FLOAT_MDCT, 0, 15 << (i + 3), &scale, 0)))
677  return AVERROR(ENOMEM);
678  }
679 
680  /* Zero out previous energy (matters for inter first frame) */
681  for (int ch = 0; ch < s->channels; ch++)
682  memset(s->last_quantized_energy[ch], 0.0f, sizeof(float)*CELT_MAX_BANDS);
683 
684  /* Allocate an empty frame to use as overlap for the first frame of audio */
685  ff_bufqueue_add(avctx, &s->bufqueue, spawn_empty_frame(s));
686  if (!ff_bufqueue_peek(&s->bufqueue, 0))
687  return AVERROR(ENOMEM);
688 
689  if ((ret = ff_opus_psy_init(&s->psyctx, s->avctx, &s->bufqueue, &s->options)))
690  return ret;
691 
692  /* Frame structs and range coder buffers */
693  max_frames = ceilf(FFMIN(s->options.max_delay_ms, 120.0f)/2.5f);
694  s->frame = av_malloc(max_frames*sizeof(CeltFrame));
695  if (!s->frame)
696  return AVERROR(ENOMEM);
697  s->rc = av_malloc(max_frames*sizeof(OpusRangeCoder));
698  if (!s->rc)
699  return AVERROR(ENOMEM);
700 
701  for (int i = 0; i < max_frames; i++) {
702  s->frame[i].dsp = s->dsp;
703  s->frame[i].avctx = s->avctx;
704  s->frame[i].seed = 0;
705  s->frame[i].pvq = s->pvq;
706  s->frame[i].apply_phase_inv = s->options.apply_phase_inv;
707  s->frame[i].block[0].emph_coeff = s->frame[i].block[1].emph_coeff = 0.0f;
708  }
709 
710  return 0;
711 }
712 
713 #define OPUSENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
714 static const AVOption opusenc_options[] = {
715  { "opus_delay", "Maximum delay in milliseconds", offsetof(OpusEncContext, options.max_delay_ms), AV_OPT_TYPE_FLOAT, { .dbl = OPUS_MAX_LOOKAHEAD }, 2.5f, OPUS_MAX_LOOKAHEAD, OPUSENC_FLAGS, "max_delay_ms" },
716  { "apply_phase_inv", "Apply intensity stereo phase inversion", offsetof(OpusEncContext, options.apply_phase_inv), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, OPUSENC_FLAGS, "apply_phase_inv" },
717  { NULL },
718 };
719 
720 static const AVClass opusenc_class = {
721  .class_name = "Opus encoder",
722  .item_name = av_default_item_name,
723  .option = opusenc_options,
724  .version = LIBAVUTIL_VERSION_INT,
725 };
726 
728  { "b", "0" },
729  { "compression_level", "10" },
730  { NULL },
731 };
732 
734  .p.name = "opus",
735  CODEC_LONG_NAME("Opus"),
736  .p.type = AVMEDIA_TYPE_AUDIO,
737  .p.id = AV_CODEC_ID_OPUS,
738  .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY |
740  .defaults = opusenc_defaults,
741  .p.priv_class = &opusenc_class,
742  .priv_data_size = sizeof(OpusEncContext),
745  .close = opus_encode_end,
746  .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
747  .p.supported_samplerates = (const int []){ 48000, 0 },
749  .p.ch_layouts = (const AVChannelLayout []){ AV_CHANNEL_LAYOUT_MONO,
750  AV_CHANNEL_LAYOUT_STEREO, { 0 } },
751  .p.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
753 };
AVCodecContext::frame_size
int frame_size
Number of samples per channel in an audio frame.
Definition: avcodec.h:1035
AV_SAMPLE_FMT_FLTP
@ AV_SAMPLE_FMT_FLTP
float, planar
Definition: samplefmt.h:66
ff_celt_postfilter_taps
const float ff_celt_postfilter_taps[3][3]
Definition: opustab.c:1098
ff_opus_rc_enc_cdf
void ff_opus_rc_enc_cdf(OpusRangeCoder *rc, int val, const uint16_t *cdf)
Definition: opus_rc.c:109
OpusEncContext::av_class
AVClass * av_class
Definition: opusenc.c:39
ff_opus_psy_process
int ff_opus_psy_process(OpusPsyContext *s, OpusPacketInfo *p)
Definition: opusenc_psy.c:226
av_clip
#define av_clip
Definition: common.h:95
spawn_empty_frame
static AVFrame * spawn_empty_frame(OpusEncContext *s)
Definition: opusenc.c:521
FF_CODEC_CAP_INIT_CLEANUP
#define FF_CODEC_CAP_INIT_CLEANUP
The codec allows calling the close function for deallocation even if the init function returned a fai...
Definition: codec_internal.h:42
AVERROR
Filter the word “frame” indicates either a video frame or a group of audio as stored in an AVFrame structure Format for each input and each output the list of supported formats For video that means pixel format For audio that means channel sample they are references to shared objects When the negotiation mechanism computes the intersection of the formats supported at each end of a all references to both lists are replaced with a reference to the intersection And when a single format is eventually chosen for a link amongst the remaining all references to the list are updated That means that if a filter requires that its input and output have the same format amongst a supported all it has to do is use a reference to the same list of formats query_formats can leave some formats unset and return AVERROR(EAGAIN) to cause the negotiation mechanism toagain later. That can be used by filters with complex requirements to use the format negotiated on one link to set the formats supported on another. Frame references ownership and permissions
opt.h
ff_celt_freq_bands
const uint8_t ff_celt_freq_bands[]
Definition: opustab.c:768
AV_WL32
#define AV_WL32(p, v)
Definition: intreadwrite.h:426
mem_internal.h
ff_af_queue_remove
void ff_af_queue_remove(AudioFrameQueue *afq, int nb_samples, int64_t *pts, int64_t *duration)
Remove frame(s) from the queue.
Definition: audio_frame_queue.c:75
av_frame_get_buffer
int av_frame_get_buffer(AVFrame *frame, int align)
Allocate new buffer(s) for audio or video data.
Definition: frame.c:259
AV_CHANNEL_LAYOUT_STEREO
#define AV_CHANNEL_LAYOUT_STEREO
Definition: channel_layout.h:369
AVCodecContext::sample_rate
int sample_rate
samples per second
Definition: avcodec.h:1007
ff_af_queue_close
void ff_af_queue_close(AudioFrameQueue *afq)
Close AudioFrameQueue.
Definition: audio_frame_queue.c:36
log2f
#define log2f(x)
Definition: libm.h:409
ff_opus_psy_celt_frame_init
void ff_opus_psy_celt_frame_init(OpusPsyContext *s, CeltFrame *f, int index)
Definition: opusenc_psy.c:257
src1
const pixel * src1
Definition: h264pred_template.c:421
AV_CH_LAYOUT_MONO
#define AV_CH_LAYOUT_MONO
Definition: channel_layout.h:210
AVTXContext
Definition: tx_priv.h:228
ff_opus_rc_enc_uint
void ff_opus_rc_enc_uint(OpusRangeCoder *rc, uint32_t val, uint32_t size)
CELT: write a uniformly distributed integer.
Definition: opus_rc.c:204
ff_af_queue_init
av_cold void ff_af_queue_init(AVCodecContext *avctx, AudioFrameQueue *afq)
Initialize AudioFrameQueue.
Definition: audio_frame_queue.c:28
av_frame_free
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:116
ff_celt_model_tapset
const uint16_t ff_celt_model_tapset[]
Definition: opustab.c:758
opusenc_options
static const AVOption opusenc_options[]
Definition: opusenc.c:714
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:325
tmp
static uint8_t tmp[11]
Definition: aes_ctr.c:28
AVPacket::data
uint8_t * data
Definition: packet.h:374
AVOption
AVOption.
Definition: opt.h:251
encode.h
b
#define b
Definition: input.c:41
OPUS_RC_CHECKPOINT_SPAWN
#define OPUS_RC_CHECKPOINT_SPAWN(rc)
Definition: opus_rc.h:115
FFCodec
Definition: codec_internal.h:119
opus_packet_assembler
static void opus_packet_assembler(OpusEncContext *s, AVPacket *avpkt)
Definition: opusenc.c:495
float.h
AVPacket::duration
int64_t duration
Duration of this packet in AVStream->time_base units, 0 if unknown.
Definition: packet.h:392
OpusEncContext::options
OpusEncOptions options
Definition: opusenc.c:40
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
opus_encode_end
static av_cold int opus_encode_end(AVCodecContext *avctx)
Definition: opusenc.c:613
AVChannelLayout::nb_channels
int nb_channels
Number of channels in this layout.
Definition: channel_layout.h:311
ff_celt_pvq_init
int av_cold ff_celt_pvq_init(CeltPVQ **pvq, int encode)
Definition: opus_pvq.c:906
exp_quant_coarse
static void exp_quant_coarse(OpusRangeCoder *rc, CeltFrame *f, float last_energy[][CELT_MAX_BANDS], int intra)
Definition: opusenc.c:325
ceilf
static __device__ float ceilf(float a)
Definition: cuda_runtime.h:175
av_tx_init
av_cold int av_tx_init(AVTXContext **ctx, av_tx_fn *tx, enum AVTXType type, int inv, int len, const void *scale, uint64_t flags)
Initialize a transform context with the given configuration (i)MDCTs with an odd length are currently...
Definition: tx.c:883
ff_bufqueue_get
static AVFrame * ff_bufqueue_get(struct FFBufQueue *queue)
Get the first buffer from the queue and remove it.
Definition: bufferqueue.h:98
opus_rc_tell
static av_always_inline uint32_t opus_rc_tell(const OpusRangeCoder *rc)
CELT: estimate bits of entropy that have thus far been consumed for the current CELT frame,...
Definition: opus_rc.h:60
ff_celt_coarse_energy_dist
const uint8_t ff_celt_coarse_energy_dist[4][2][42]
Definition: opustab.c:808
av_malloc
#define av_malloc(s)
Definition: tableprint_vlc.h:30
CeltBlock
Definition: opus_celt.h:70
win
static float win(SuperEqualizerContext *s, float n, int N)
Definition: af_superequalizer.c:119
CeltPVQ
Definition: opus_pvq.h:37
FFCodecDefault
Definition: codec_internal.h:89
FFCodec::p
AVCodec p
The public AVCodec.
Definition: codec_internal.h:123
OPUSENC_FLAGS
#define OPUSENC_FLAGS
Definition: opusenc.c:713
AVCodecContext::ch_layout
AVChannelLayout ch_layout
Audio channel layout.
Definition: avcodec.h:2059
audio_frame_queue.h
AVCodecContext::initial_padding
int initial_padding
Audio only.
Definition: avcodec.h:1726
opus_encode_frame
static int opus_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, const AVFrame *frame, int *got_packet_ptr)
Definition: opusenc.c:545
AVCodecContext::flags
int flags
AV_CODEC_FLAG_*.
Definition: avcodec.h:478
OpusEncContext::frame
CeltFrame * frame
Definition: opusenc.c:57
scale
static av_always_inline float scale(float x, float s)
Definition: vf_v360.c:1389
ff_opus_psy_end
av_cold int ff_opus_psy_end(OpusPsyContext *s)
Definition: opusenc_psy.c:594
FF_CODEC_ENCODE_CB
#define FF_CODEC_ENCODE_CB(func)
Definition: codec_internal.h:307
ff_af_queue_add
int ff_af_queue_add(AudioFrameQueue *afq, const AVFrame *f)
Add a frame to the queue.
Definition: audio_frame_queue.c:44
AV_CH_LAYOUT_STEREO
#define AV_CH_LAYOUT_STEREO
Definition: channel_layout.h:211
quant
static int quant(float coef, const float Q, const float rounding)
Quantize one coefficient.
Definition: aacenc_utils.h:59
av_frame_alloc
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
Definition: frame.c:104
AV_LOG_ERROR
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:180
OPUS_MAX_LOOKAHEAD
#define OPUS_MAX_LOOKAHEAD
Definition: opusenc.h:32
OpusEncContext::afq
AudioFrameQueue afq
Definition: opusenc.c:43
AV_CODEC_CAP_EXPERIMENTAL
#define AV_CODEC_CAP_EXPERIMENTAL
Codec is experimental and is thus avoided in favor of non experimental encoders.
Definition: codec.h:105
av_cold
#define av_cold
Definition: attributes.h:90
OPUS_BLOCK_SIZE
#define OPUS_BLOCK_SIZE(x)
Definition: opusenc.h:39
av_tx_fn
void(* av_tx_fn)(AVTXContext *s, void *out, void *in, ptrdiff_t stride)
Function pointer to a function to perform the transform.
Definition: tx.h:127
float
float
Definition: af_crystalizer.c:122
AVCodecContext::extradata_size
int extradata_size
Definition: avcodec.h:500
AV_TX_FLOAT_MDCT
@ AV_TX_FLOAT_MDCT
Standard MDCT with a sample data type of float, double or int32_t, respecively.
Definition: tx.h:68
OpusEncContext::rc
OpusRangeCoder * rc
Definition: opusenc.c:58
OpusEncContext::scratch
float scratch[2048]
Definition: opusenc.c:63
s
#define s(width, name)
Definition: cbs_vp9.c:256
floor
static __device__ float floor(float a)
Definition: cuda_runtime.h:173
frame_size
int frame_size
Definition: mxfenc.c:2202
AVMEDIA_TYPE_AUDIO
@ AVMEDIA_TYPE_AUDIO
Definition: avutil.h:202
init
int(* init)(AVBSFContext *ctx)
Definition: dts2pts_bsf.c:363
ff_opus_psy_celt_frame_process
int ff_opus_psy_celt_frame_process(OpusPsyContext *s, CeltFrame *f, int index)
Definition: opusenc_psy.c:458
CELT_MAX_FINE_BITS
#define CELT_MAX_FINE_BITS
Definition: opus_celt.h:47
bits
uint8_t bits
Definition: vp3data.h:128
CODEC_OLD_CHANNEL_LAYOUTS
#define CODEC_OLD_CHANNEL_LAYOUTS(...)
Definition: codec_internal.h:294
AudioFrameQueue
Definition: audio_frame_queue.h:32
av_frame_clone
AVFrame * av_frame_clone(const AVFrame *src)
Create a new frame that references the same data as src.
Definition: frame.c:474
OpusPsyContext
Definition: opusenc_psy.h:52
CODEC_LONG_NAME
#define CODEC_LONG_NAME(str)
Definition: codec_internal.h:264
FFABS
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition: common.h:64
OPUS_MODE_NB
@ OPUS_MODE_NB
Definition: opus.h:46
OpusPacketInfo
Definition: opusenc.h:48
LIBAVUTIL_VERSION_INT
#define LIBAVUTIL_VERSION_INT
Definition: version.h:85
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:66
NULL
#define NULL
Definition: coverity.c:32
OPUS_BANDWITH_NB
@ OPUS_BANDWITH_NB
Definition: opus.h:56
opusenc.h
OpusEncContext::packet
OpusPacketInfo packet
Definition: opusenc.c:53
ff_celt_pvq_uninit
void av_cold ff_celt_pvq_uninit(CeltPVQ **pvq)
Definition: opus_pvq.c:926
period
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf default minimum maximum flags name is the option keep it simple and lowercase description are in without period
Definition: writing_filters.txt:89
AVCodecContext::bit_rate
int64_t bit_rate
the average bitrate
Definition: avcodec.h:448
av_default_item_name
const char * av_default_item_name(void *ptr)
Return the context name.
Definition: log.c:237
ff_bufqueue_discard_all
static void ff_bufqueue_discard_all(struct FFBufQueue *queue)
Unref and remove all buffers from the queue.
Definition: bufferqueue.h:111
celt_frame_mdct
static void celt_frame_mdct(OpusEncContext *s, CeltFrame *f)
Definition: opusenc.c:196
celt_quant_fine
static void celt_quant_fine(CeltFrame *f, OpusRangeCoder *rc)
Definition: opusenc.c:388
ff_celt_window
static const float *const ff_celt_window
Definition: opustab.h:162
celt_frame_setup_input
static void celt_frame_setup_input(OpusEncContext *s, CeltFrame *f)
Definition: opusenc.c:122
sqrtf
static __device__ float sqrtf(float a)
Definition: cuda_runtime.h:184
opus_gen_toc
static int opus_gen_toc(OpusEncContext *s, uint8_t *toc, int *size, int *fsize_needed)
Definition: opusenc.c:79
ff_celt_freq_range
const uint8_t ff_celt_freq_range[]
Definition: opustab.c:772
index
int index
Definition: gxfenc.c:89
float_dsp.h
ff_opus_psy_postencode_update
void ff_opus_psy_postencode_update(OpusPsyContext *s, CeltFrame *f)
Definition: opusenc_psy.c:479
CELT_ENERGY_SILENCE
#define CELT_ENERGY_SILENCE
Definition: opus_celt.h:52
OpusEncContext::channels
int channels
Definition: opusenc.c:55
options
const OptionDef options[]
CELT_OVERLAP
#define CELT_OVERLAP
Definition: opus_celt.h:39
opustab.h
CELT_MAX_BANDS
#define CELT_MAX_BANDS
Definition: opus_celt.h:42
f
f
Definition: af_crystalizer.c:122
ff_opus_rc_enc_init
void ff_opus_rc_enc_init(OpusRangeCoder *rc)
Definition: opus_rc.c:402
AV_CODEC_CAP_DR1
#define AV_CODEC_CAP_DR1
Codec uses get_buffer() or get_encode_buffer() for allocating buffers and supports custom allocators.
Definition: codec.h:52
AVPacket::size
int size
Definition: packet.h:375
AVChannelLayout
An AVChannelLayout holds information about the channel layout of audio data.
Definition: channel_layout.h:301
codec_internal.h
opus_pvq.h
opusenc_psy.h
for
for(k=2;k<=8;++k)
Definition: h264pred_template.c:425
bps
unsigned bps
Definition: movenc.c:1648
AV_SAMPLE_FMT_NONE
@ AV_SAMPLE_FMT_NONE
Definition: samplefmt.h:56
sample
#define sample
Definition: flacdsp_template.c:44
size
int size
Definition: twinvq_data.h:10344
OpusEncContext::enc_id
uint8_t enc_id[64]
Definition: opusenc.c:50
OpusRangeCoder
Definition: opus_rc.h:39
OpusEncContext::psyctx
OpusPsyContext psyctx
Definition: opusenc.c:41
AV_CODEC_ID_OPUS
@ AV_CODEC_ID_OPUS
Definition: codec_id.h:495
AVFloatDSPContext
Definition: float_dsp.h:24
AVFrame::format
int format
format of the frame, -1 if unknown or unset Values correspond to enum AVPixelFormat for video frames,...
Definition: frame.h:412
ff_celt_tf_select
const int8_t ff_celt_tf_select[4][2][2][2]
Definition: opustab.c:782
offset
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf offset
Definition: writing_filters.txt:86
ff_bufqueue_add
static void ff_bufqueue_add(void *log, struct FFBufQueue *queue, AVFrame *buf)
Add a buffer to the queue.
Definition: bufferqueue.h:71
write_opuslacing
static int write_opuslacing(uint8_t *dst, int v)
Definition: opusenc.c:488
OpusEncContext
Definition: opusenc.c:38
input
and forward the test the status of outputs and forward it to the corresponding return FFERROR_NOT_READY If the filters stores internally one or a few frame for some input
Definition: filter_design.txt:172
opus_write_extradata
static void opus_write_extradata(AVCodecContext *avctx)
Definition: opusenc.c:66
ff_celt_beta_coef
const float ff_celt_beta_coef[]
Definition: opustab.c:804
av_tx_uninit
av_cold void av_tx_uninit(AVTXContext **ctx)
Frees a context and sets *ctx to NULL, does nothing when *ctx == NULL.
Definition: tx.c:294
AV_OPT_TYPE_FLOAT
@ AV_OPT_TYPE_FLOAT
Definition: opt.h:228
DECLARE_ALIGNED
#define DECLARE_ALIGNED(n, t, v)
Definition: mem.h:116
OpusEncContext::last_quantized_energy
float last_quantized_energy[OPUS_MAX_CHANNELS][CELT_MAX_BANDS]
Definition: opusenc.c:61
bytestream_put_buffer
static av_always_inline void bytestream_put_buffer(uint8_t **b, const uint8_t *src, unsigned int size)
Definition: bytestream.h:372
AVFrame::nb_samples
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:405
lrintf
#define lrintf(x)
Definition: libm_mips.h:72
ff_bufqueue_peek
static AVFrame * ff_bufqueue_peek(struct FFBufQueue *queue, unsigned index)
Get a buffer from the queue without altering it.
Definition: bufferqueue.h:87
ff_celt_window_padded
const float ff_celt_window_padded[136]
Definition: opustab.c:1104
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:269
AVPacket::pts
int64_t pts
Presentation timestamp in AVStream->time_base units; the time at which the decompressed packet will b...
Definition: packet.h:367
ff_opus_rc_enc_laplace
void ff_opus_rc_enc_laplace(OpusRangeCoder *rc, int *value, uint32_t symbol, int decay)
Definition: opus_rc.c:314
av_get_bytes_per_sample
int av_get_bytes_per_sample(enum AVSampleFormat sample_fmt)
Return number of bytes per sample.
Definition: samplefmt.c:108
AVCodecContext::extradata
uint8_t * extradata
some codecs need / can use extradata like Huffman tables.
Definition: avcodec.h:499
FFBufQueue
Structure holding the queue.
Definition: bufferqueue.h:49
AVFrame::extended_data
uint8_t ** extended_data
pointers to the data planes/channels.
Definition: frame.h:386
src2
const pixel * src2
Definition: h264pred_template.c:422
ff_opus_rc_put_raw
void ff_opus_rc_put_raw(OpusRangeCoder *rc, uint32_t val, uint32_t count)
CELT: write 0 - 31 bits to the rawbits buffer.
Definition: opus_rc.c:161
celt_encode_frame
static void celt_encode_frame(OpusEncContext *s, OpusRangeCoder *rc, CeltFrame *f, int index)
Definition: opusenc.c:423
AVSampleFormat
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:55
OPUS_MAX_CHANNELS
#define OPUS_MAX_CHANNELS
Definition: opusenc.h:34
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
ff_opus_psy_signal_eof
void ff_opus_psy_signal_eof(OpusPsyContext *s)
Definition: opusenc_psy.c:589
AVCodec::name
const char * name
Name of the codec implementation.
Definition: codec.h:211
AV_PKT_DATA_SKIP_SAMPLES
@ AV_PKT_DATA_SKIP_SAMPLES
Recommmends skipping the specified number of samples.
Definition: packet.h:157
len
int len
Definition: vorbis_enc_data.h:426
CELT_POSTFILTER_MINPERIOD
#define CELT_POSTFILTER_MINPERIOD
Definition: opus_celt.h:51
ff_celt_quant_bands
void ff_celt_quant_bands(CeltFrame *f, OpusRangeCoder *rc)
Definition: opus_celt.c:28
ff_opus_default_coupled_streams
const uint8_t ff_opus_default_coupled_streams[]
Definition: opustab.c:27
opusenc_class
static const AVClass opusenc_class
Definition: opusenc.c:720
ff_celt_bitalloc
void ff_celt_bitalloc(CeltFrame *f, OpusRangeCoder *rc, int encode)
Definition: opus_celt.c:137
OpusEncContext::tx
AVTXContext * tx[CELT_BLOCK_NB]
Definition: opusenc.c:45
ret
ret
Definition: filter_design.txt:187
OpusEncContext::avctx
AVCodecContext * avctx
Definition: opusenc.c:42
AVClass::class_name
const char * class_name
The name of the class; usually it is the same name as the context structure type to which the AVClass...
Definition: log.h:71
frame
these buffered frames must be flushed immediately if a new input produces new the filter must not call request_frame to get more It must just process the frame or queue it The task of requesting more frames is left to the filter s request_frame method or the application If a filter has several the filter must be ready for frames arriving randomly on any input any filter with several inputs will most likely require some kind of queuing mechanism It is perfectly acceptable to have a limited queue and to drop frames when the inputs are too unbalanced request_frame For filters that do not use the this method is called when a frame is wanted on an output For a it should directly call filter_frame on the corresponding output For a if there are queued frames already one of these frames should be pushed If the filter should request a frame on one of its repeatedly until at least one frame has been pushed Return or at least make progress towards producing a frame
Definition: filter_design.txt:264
opusenc_defaults
static const FFCodecDefault opusenc_defaults[]
Definition: opusenc.c:727
AV_INPUT_BUFFER_PADDING_SIZE
#define AV_INPUT_BUFFER_PADDING_SIZE
Definition: defs.h:40
OpusEncContext::enc_id_bits
int enc_id_bits
Definition: opusenc.c:51
left
Tag MUST be and< 10hcoeff half pel interpolation filter coefficients, hcoeff[0] are the 2 middle coefficients[1] are the next outer ones and so on, resulting in a filter like:...eff[2], hcoeff[1], hcoeff[0], hcoeff[0], hcoeff[1], hcoeff[2] ... the sign of the coefficients is not explicitly stored but alternates after each coeff and coeff[0] is positive, so ...,+,-,+,-,+,+,-,+,-,+,... hcoeff[0] is not explicitly stored but found by subtracting the sum of all stored coefficients with signs from 32 hcoeff[0]=32 - hcoeff[1] - hcoeff[2] - ... a good choice for hcoeff and htaps is htaps=6 hcoeff={40,-10, 2} an alternative which requires more computations at both encoder and decoder side and may or may not be better is htaps=8 hcoeff={42,-14, 6,-2}ref_frames minimum of the number of available reference frames and max_ref_frames for example the first frame after a key frame always has ref_frames=1spatial_decomposition_type wavelet type 0 is a 9/7 symmetric compact integer wavelet 1 is a 5/3 symmetric compact integer wavelet others are reserved stored as delta from last, last is reset to 0 if always_reset||keyframeqlog quality(logarithmic quantizer scale) stored as delta from last, last is reset to 0 if always_reset||keyframemv_scale stored as delta from last, last is reset to 0 if always_reset||keyframe FIXME check that everything works fine if this changes between framesqbias dequantization bias stored as delta from last, last is reset to 0 if always_reset||keyframeblock_max_depth maximum depth of the block tree stored as delta from last, last is reset to 0 if always_reset||keyframequant_table quantization tableHighlevel bitstream structure:==============================--------------------------------------------|Header|--------------------------------------------|------------------------------------|||Block0||||split?||||yes no||||......... intra?||||:Block01 :yes no||||:Block02 :....... ..........||||:Block03 ::y DC ::ref index:||||:Block04 ::cb DC ::motion x :||||......... :cr DC ::motion y :||||....... ..........|||------------------------------------||------------------------------------|||Block1|||...|--------------------------------------------|------------ ------------ ------------|||Y subbands||Cb subbands||Cr subbands||||--- ---||--- ---||--- ---|||||LL0||HL0||||LL0||HL0||||LL0||HL0|||||--- ---||--- ---||--- ---||||--- ---||--- ---||--- ---|||||LH0||HH0||||LH0||HH0||||LH0||HH0|||||--- ---||--- ---||--- ---||||--- ---||--- ---||--- ---|||||HL1||LH1||||HL1||LH1||||HL1||LH1|||||--- ---||--- ---||--- ---||||--- ---||--- ---||--- ---|||||HH1||HL2||||HH1||HL2||||HH1||HL2|||||...||...||...|||------------ ------------ ------------|--------------------------------------------Decoding process:=================------------|||Subbands|------------||||------------|Intra DC||||LL0 subband prediction ------------|\ Dequantization ------------------- \||Reference frames|\ IDWT|------- -------|Motion \|||Frame 0||Frame 1||Compensation . OBMC v -------|------- -------|--------------. \------> Frame n output Frame Frame<----------------------------------/|...|------------------- Range Coder:============Binary Range Coder:------------------- The implemented range coder is an adapted version based upon "Range encoding: an algorithm for removing redundancy from a digitised message." by G. N. N. Martin. The symbols encoded by the Snow range coder are bits(0|1). The associated probabilities are not fix but change depending on the symbol mix seen so far. bit seen|new state ---------+----------------------------------------------- 0|256 - state_transition_table[256 - old_state];1|state_transition_table[old_state];state_transition_table={ 0, 0, 0, 0, 0, 0, 0, 0, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 190, 191, 192, 194, 194, 195, 196, 197, 198, 199, 200, 201, 202, 202, 204, 205, 206, 207, 208, 209, 209, 210, 211, 212, 213, 215, 215, 216, 217, 218, 219, 220, 220, 222, 223, 224, 225, 226, 227, 227, 229, 229, 230, 231, 232, 234, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 248, 0, 0, 0, 0, 0, 0, 0};FIXME Range Coding of integers:------------------------- FIXME Neighboring Blocks:===================left and top are set to the respective blocks unless they are outside of the image in which case they are set to the Null block top-left is set to the top left block unless it is outside of the image in which case it is set to the left block if this block has no larger parent block or it is at the left side of its parent block and the top right block is not outside of the image then the top right block is used for top-right else the top-left block is used Null block y, cb, cr are 128 level, ref, mx and my are 0 Motion Vector Prediction:=========================1. the motion vectors of all the neighboring blocks are scaled to compensate for the difference of reference frames scaled_mv=(mv *(256 *(current_reference+1)/(mv.reference+1))+128)> the median of the scaled left
Definition: snow.txt:386
CELT_EMPH_COEFF
#define CELT_EMPH_COEFF
Definition: opusdsp.h:22
AVCodecContext
main external API structure.
Definition: avcodec.h:398
ff_celt_model_energy_small
const uint16_t ff_celt_model_energy_small[]
Definition: opustab.c:766
channel_layout.h
av_packet_new_side_data
uint8_t * av_packet_new_side_data(AVPacket *pkt, enum AVPacketSideDataType type, size_t size)
Allocate new information of a packet.
Definition: avpacket.c:230
celt_quant_coarse
static void celt_quant_coarse(CeltFrame *f, OpusRangeCoder *rc, float last_energy[][CELT_MAX_BANDS])
Definition: opusenc.c:368
ff_opus_rc_enc_log
void ff_opus_rc_enc_log(OpusRangeCoder *rc, int val, uint32_t bits)
Definition: opus_rc.c:131
OpusEncContext::tx_fn
av_tx_fn tx_fn[CELT_BLOCK_NB]
Definition: opusenc.c:46
temp
else temp
Definition: vf_mcdeint.c:248
AV_CODEC_CAP_DELAY
#define AV_CODEC_CAP_DELAY
Encoder or decoder requires flushing with NULL input at the end in order to give the complete and cor...
Definition: codec.h:82
ff_celt_alpha_coef
const float ff_celt_alpha_coef[]
Definition: opustab.c:800
celt_enc_quant_pfilter
static void celt_enc_quant_pfilter(OpusRangeCoder *rc, CeltFrame *f)
Definition: opusenc.c:288
av_channel_layout_copy
int av_channel_layout_copy(AVChannelLayout *dst, const AVChannelLayout *src)
Make a copy of a channel layout.
Definition: channel_layout.c:639
celt_quant_final
static void celt_quant_final(OpusEncContext *s, OpusRangeCoder *rc, CeltFrame *f)
Definition: opusenc.c:405
ff_opus_encoder
const FFCodec ff_opus_encoder
Definition: opusenc.c:733
AV_CODEC_FLAG_BITEXACT
#define AV_CODEC_FLAG_BITEXACT
Use only bitexact stuff (except (I)DCT).
Definition: avcodec.h:288
AVCodecContext::frame_number
int frame_number
Frame counter, set by libavcodec.
Definition: avcodec.h:1046
ff_opus_psy_init
av_cold int ff_opus_psy_init(OpusPsyContext *s, AVCodecContext *avctx, struct FFBufQueue *bufqueue, OpusEncOptions *options)
Definition: opusenc_psy.c:515
AV_CHANNEL_LAYOUT_MONO
#define AV_CHANNEL_LAYOUT_MONO
Definition: channel_layout.h:368
diff
static av_always_inline int diff(const uint32_t a, const uint32_t b)
Definition: vf_palettegen.c:139
OpusEncOptions
Definition: opusenc.h:43
FFALIGN
#define FFALIGN(x, a)
Definition: macros.h:78
alpha
static const int16_t alpha[]
Definition: ilbcdata.h:55
AVCodecContext::priv_data
void * priv_data
Definition: avcodec.h:425
AVPacket
This structure stores compressed data.
Definition: packet.h:351
AV_OPT_TYPE_BOOL
@ AV_OPT_TYPE_BOOL
Definition: opt.h:244
OpusEncContext::pvq
CeltPVQ * pvq
Definition: opusenc.c:47
av_freep
#define av_freep(p)
Definition: tableprint_vlc.h:34
avpriv_float_dsp_alloc
av_cold AVFloatDSPContext * avpriv_float_dsp_alloc(int bit_exact)
Allocate a float DSP context.
Definition: float_dsp.c:135
CELT_BLOCK_NB
@ CELT_BLOCK_NB
Definition: opus_celt.h:67
bytestream.h
block
The exact code depends on how similar the blocks are and how related they are to the block
Definition: filter_design.txt:207
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:27
celt_apply_preemph_filter
static void celt_apply_preemph_filter(OpusEncContext *s, CeltFrame *f)
Definition: opusenc.c:162
ff_celt_mean_energy
const float ff_celt_mean_energy[]
Definition: opustab.c:792
OpusEncContext::bufqueue
struct FFBufQueue bufqueue
Definition: opusenc.c:48
OPUS_RC_CHECKPOINT_BITS
#define OPUS_RC_CHECKPOINT_BITS(rc)
Definition: opus_rc.h:119
AV_CODEC_CAP_SMALL_LAST_FRAME
#define AV_CODEC_CAP_SMALL_LAST_FRAME
Codec can be fed a final frame with a smaller size.
Definition: codec.h:87
OPUS_RC_CHECKPOINT_ROLLBACK
#define OPUS_RC_CHECKPOINT_ROLLBACK(rc)
Definition: opus_rc.h:122
ff_alloc_packet
int ff_alloc_packet(AVCodecContext *avctx, AVPacket *avpkt, int64_t size)
Check AVPacket size and allocate data.
Definition: encode.c:35
celt_enc_tf
static void celt_enc_tf(CeltFrame *f, OpusRangeCoder *rc)
Definition: opusenc.c:261
opus_encode_init
static av_cold int opus_encode_init(AVCodecContext *avctx)
Definition: opusenc.c:631
ff_opus_rc_enc_end
void ff_opus_rc_enc_end(OpusRangeCoder *rc, uint8_t *dst, int size)
Definition: opus_rc.c:360
CeltFrame
Definition: opus_celt.h:97
OpusEncContext::dsp
AVFloatDSPContext * dsp
Definition: opusenc.c:44