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 "opusenc.h"
23 #include "opus_pvq.h"
24 #include "opusenc_psy.h"
25 #include "opustab.h"
26 
27 #include "libavutil/float_dsp.h"
28 #include "libavutil/mem_internal.h"
29 #include "libavutil/opt.h"
30 #include "internal.h"
31 #include "bytestream.h"
32 #include "audio_frame_queue.h"
33 
34 typedef struct OpusEncContext {
44 
45  uint8_t enc_id[64];
47 
49 
50  int channels;
51 
54 
55  /* Actual energy the decoder will have */
57 
58  DECLARE_ALIGNED(32, float, scratch)[2048];
60 
62 {
63  uint8_t *bs = avctx->extradata;
64 
65  bytestream_put_buffer(&bs, "OpusHead", 8);
66  bytestream_put_byte (&bs, 0x1);
67  bytestream_put_byte (&bs, avctx->channels);
68  bytestream_put_le16 (&bs, avctx->initial_padding);
69  bytestream_put_le32 (&bs, avctx->sample_rate);
70  bytestream_put_le16 (&bs, 0x0);
71  bytestream_put_byte (&bs, 0x0); /* Default layout */
72 }
73 
74 static int opus_gen_toc(OpusEncContext *s, uint8_t *toc, int *size, int *fsize_needed)
75 {
76  int tmp = 0x0, extended_toc = 0;
77  static const int toc_cfg[][OPUS_MODE_NB][OPUS_BANDWITH_NB] = {
78  /* Silk Hybrid Celt Layer */
79  /* NB MB WB SWB FB NB MB WB SWB FB NB MB WB SWB FB Bandwidth */
80  { { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 }, { 17, 0, 21, 25, 29 } }, /* 2.5 ms */
81  { { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 }, { 18, 0, 22, 26, 30 } }, /* 5 ms */
82  { { 1, 5, 9, 0, 0 }, { 0, 0, 0, 13, 15 }, { 19, 0, 23, 27, 31 } }, /* 10 ms */
83  { { 2, 6, 10, 0, 0 }, { 0, 0, 0, 14, 16 }, { 20, 0, 24, 28, 32 } }, /* 20 ms */
84  { { 3, 7, 11, 0, 0 }, { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }, /* 40 ms */
85  { { 4, 8, 12, 0, 0 }, { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }, /* 60 ms */
86  };
87  int cfg = toc_cfg[s->packet.framesize][s->packet.mode][s->packet.bandwidth];
88  *fsize_needed = 0;
89  if (!cfg)
90  return 1;
91  if (s->packet.frames == 2) { /* 2 packets */
92  if (s->frame[0].framebits == s->frame[1].framebits) { /* same size */
93  tmp = 0x1;
94  } else { /* different size */
95  tmp = 0x2;
96  *fsize_needed = 1; /* put frame sizes in the packet */
97  }
98  } else if (s->packet.frames > 2) {
99  tmp = 0x3;
100  extended_toc = 1;
101  }
102  tmp |= (s->channels > 1) << 2; /* Stereo or mono */
103  tmp |= (cfg - 1) << 3; /* codec configuration */
104  *toc++ = tmp;
105  if (extended_toc) {
106  for (int i = 0; i < (s->packet.frames - 1); i++)
107  *fsize_needed |= (s->frame[i].framebits != s->frame[i + 1].framebits);
108  tmp = (*fsize_needed) << 7; /* vbr flag */
109  tmp |= (0) << 6; /* padding flag */
110  tmp |= s->packet.frames;
111  *toc++ = tmp;
112  }
113  *size = 1 + extended_toc;
114  return 0;
115 }
116 
118 {
119  AVFrame *cur = NULL;
120  const int subframesize = s->avctx->frame_size;
121  int subframes = OPUS_BLOCK_SIZE(s->packet.framesize) / subframesize;
122 
123  cur = ff_bufqueue_get(&s->bufqueue);
124 
125  for (int ch = 0; ch < f->channels; ch++) {
126  CeltBlock *b = &f->block[ch];
127  const void *input = cur->extended_data[ch];
128  size_t bps = av_get_bytes_per_sample(cur->format);
129  memcpy(b->overlap, input, bps*cur->nb_samples);
130  }
131 
132  av_frame_free(&cur);
133 
134  for (int sf = 0; sf < subframes; sf++) {
135  if (sf != (subframes - 1))
136  cur = ff_bufqueue_get(&s->bufqueue);
137  else
138  cur = ff_bufqueue_peek(&s->bufqueue, 0);
139 
140  for (int ch = 0; ch < f->channels; ch++) {
141  CeltBlock *b = &f->block[ch];
142  const void *input = cur->extended_data[ch];
143  const size_t bps = av_get_bytes_per_sample(cur->format);
144  const size_t left = (subframesize - cur->nb_samples)*bps;
145  const size_t len = FFMIN(subframesize, cur->nb_samples)*bps;
146  memcpy(&b->samples[sf*subframesize], input, len);
147  memset(&b->samples[cur->nb_samples], 0, left);
148  }
149 
150  /* Last frame isn't popped off and freed yet - we need it for overlap */
151  if (sf != (subframes - 1))
152  av_frame_free(&cur);
153  }
154 }
155 
156 /* Apply the pre emphasis filter */
158 {
159  const int subframesize = s->avctx->frame_size;
160  const int subframes = OPUS_BLOCK_SIZE(s->packet.framesize) / subframesize;
161 
162  /* Filter overlap */
163  for (int ch = 0; ch < f->channels; ch++) {
164  CeltBlock *b = &f->block[ch];
165  float m = b->emph_coeff;
166  for (int i = 0; i < CELT_OVERLAP; i++) {
167  float sample = b->overlap[i];
168  b->overlap[i] = sample - m;
169  m = sample * CELT_EMPH_COEFF;
170  }
171  b->emph_coeff = m;
172  }
173 
174  /* Filter the samples but do not update the last subframe's coeff - overlap ^^^ */
175  for (int sf = 0; sf < subframes; sf++) {
176  for (int ch = 0; ch < f->channels; ch++) {
177  CeltBlock *b = &f->block[ch];
178  float m = b->emph_coeff;
179  for (int i = 0; i < subframesize; i++) {
180  float sample = b->samples[sf*subframesize + i];
181  b->samples[sf*subframesize + i] = sample - m;
182  m = sample * CELT_EMPH_COEFF;
183  }
184  if (sf != (subframes - 1))
185  b->emph_coeff = m;
186  }
187  }
188 }
189 
190 /* Create the window and do the mdct */
192 {
193  float *win = s->scratch, *temp = s->scratch + 1920;
194 
195  if (f->transient) {
196  for (int ch = 0; ch < f->channels; ch++) {
197  CeltBlock *b = &f->block[ch];
198  float *src1 = b->overlap;
199  for (int t = 0; t < f->blocks; t++) {
200  float *src2 = &b->samples[CELT_OVERLAP*t];
201  s->dsp->vector_fmul(win, src1, ff_celt_window, 128);
202  s->dsp->vector_fmul_reverse(&win[CELT_OVERLAP], src2,
203  ff_celt_window - 8, 128);
204  src1 = src2;
205  s->mdct[0]->mdct(s->mdct[0], b->coeffs + t, win, f->blocks);
206  }
207  }
208  } else {
209  int blk_len = OPUS_BLOCK_SIZE(f->size), wlen = OPUS_BLOCK_SIZE(f->size + 1);
210  int rwin = blk_len - CELT_OVERLAP, lap_dst = (wlen - blk_len - CELT_OVERLAP) >> 1;
211  memset(win, 0, wlen*sizeof(float));
212  for (int ch = 0; ch < f->channels; ch++) {
213  CeltBlock *b = &f->block[ch];
214 
215  /* Overlap */
216  s->dsp->vector_fmul(temp, b->overlap, ff_celt_window, 128);
217  memcpy(win + lap_dst, temp, CELT_OVERLAP*sizeof(float));
218 
219  /* Samples, flat top window */
220  memcpy(&win[lap_dst + CELT_OVERLAP], b->samples, rwin*sizeof(float));
221 
222  /* Samples, windowed */
223  s->dsp->vector_fmul_reverse(temp, b->samples + rwin,
224  ff_celt_window - 8, 128);
225  memcpy(win + lap_dst + blk_len, temp, CELT_OVERLAP*sizeof(float));
226 
227  s->mdct[f->size]->mdct(s->mdct[f->size], b->coeffs, win, 1);
228  }
229  }
230 
231  for (int ch = 0; ch < f->channels; ch++) {
232  CeltBlock *block = &f->block[ch];
233  for (int i = 0; i < CELT_MAX_BANDS; i++) {
234  float ener = 0.0f;
235  int band_offset = ff_celt_freq_bands[i] << f->size;
236  int band_size = ff_celt_freq_range[i] << f->size;
237  float *coeffs = &block->coeffs[band_offset];
238 
239  for (int j = 0; j < band_size; j++)
240  ener += coeffs[j]*coeffs[j];
241 
242  block->lin_energy[i] = sqrtf(ener) + FLT_EPSILON;
243  ener = 1.0f/block->lin_energy[i];
244 
245  for (int j = 0; j < band_size; j++)
246  coeffs[j] *= ener;
247 
248  block->energy[i] = log2f(block->lin_energy[i]) - ff_celt_mean_energy[i];
249 
250  /* CELT_ENERGY_SILENCE is what the decoder uses and its not -infinity */
251  block->energy[i] = FFMAX(block->energy[i], CELT_ENERGY_SILENCE);
252  }
253  }
254 }
255 
257 {
258  int tf_select = 0, diff = 0, tf_changed = 0, tf_select_needed;
259  int bits = f->transient ? 2 : 4;
260 
261  tf_select_needed = ((f->size && (opus_rc_tell(rc) + bits + 1) <= f->framebits));
262 
263  for (int i = f->start_band; i < f->end_band; i++) {
264  if ((opus_rc_tell(rc) + bits + tf_select_needed) <= f->framebits) {
265  const int tbit = (diff ^ 1) == f->tf_change[i];
266  ff_opus_rc_enc_log(rc, tbit, bits);
267  diff ^= tbit;
268  tf_changed |= diff;
269  }
270  bits = f->transient ? 4 : 5;
271  }
272 
273  if (tf_select_needed && ff_celt_tf_select[f->size][f->transient][0][tf_changed] !=
274  ff_celt_tf_select[f->size][f->transient][1][tf_changed]) {
275  ff_opus_rc_enc_log(rc, f->tf_select, 1);
276  tf_select = f->tf_select;
277  }
278 
279  for (int i = f->start_band; i < f->end_band; i++)
280  f->tf_change[i] = ff_celt_tf_select[f->size][f->transient][tf_select][f->tf_change[i]];
281 }
282 
284 {
285  float gain = f->pf_gain;
286  int txval, octave = f->pf_octave, period = f->pf_period, tapset = f->pf_tapset;
287 
288  ff_opus_rc_enc_log(rc, f->pfilter, 1);
289  if (!f->pfilter)
290  return;
291 
292  /* Octave */
293  txval = FFMIN(octave, 6);
294  ff_opus_rc_enc_uint(rc, txval, 6);
295  octave = txval;
296  /* Period */
297  txval = av_clip(period - (16 << octave) + 1, 0, (1 << (4 + octave)) - 1);
298  ff_opus_rc_put_raw(rc, period, 4 + octave);
299  period = txval + (16 << octave) - 1;
300  /* Gain */
301  txval = FFMIN(((int)(gain / 0.09375f)) - 1, 7);
302  ff_opus_rc_put_raw(rc, txval, 3);
303  gain = 0.09375f * (txval + 1);
304  /* Tapset */
305  if ((opus_rc_tell(rc) + 2) <= f->framebits)
307  else
308  tapset = 0;
309  /* Finally create the coeffs */
310  for (int i = 0; i < 2; i++) {
311  CeltBlock *block = &f->block[i];
312 
313  block->pf_period_new = FFMAX(period, CELT_POSTFILTER_MINPERIOD);
314  block->pf_gains_new[0] = gain * ff_celt_postfilter_taps[tapset][0];
315  block->pf_gains_new[1] = gain * ff_celt_postfilter_taps[tapset][1];
316  block->pf_gains_new[2] = gain * ff_celt_postfilter_taps[tapset][2];
317  }
318 }
319 
321  float last_energy[][CELT_MAX_BANDS], int intra)
322 {
323  float alpha, beta, prev[2] = { 0, 0 };
324  const uint8_t *pmod = ff_celt_coarse_energy_dist[f->size][intra];
325 
326  /* Inter is really just differential coding */
327  if (opus_rc_tell(rc) + 3 <= f->framebits)
328  ff_opus_rc_enc_log(rc, intra, 3);
329  else
330  intra = 0;
331 
332  if (intra) {
333  alpha = 0.0f;
334  beta = 1.0f - (4915.0f/32768.0f);
335  } else {
336  alpha = ff_celt_alpha_coef[f->size];
337  beta = ff_celt_beta_coef[f->size];
338  }
339 
340  for (int i = f->start_band; i < f->end_band; i++) {
341  for (int ch = 0; ch < f->channels; ch++) {
342  CeltBlock *block = &f->block[ch];
343  const int left = f->framebits - opus_rc_tell(rc);
344  const float last = FFMAX(-9.0f, last_energy[ch][i]);
345  float diff = block->energy[i] - prev[ch] - last*alpha;
346  int q_en = lrintf(diff);
347  if (left >= 15) {
348  ff_opus_rc_enc_laplace(rc, &q_en, pmod[i << 1] << 7, pmod[(i << 1) + 1] << 6);
349  } else if (left >= 2) {
350  q_en = av_clip(q_en, -1, 1);
351  ff_opus_rc_enc_cdf(rc, 2*q_en + 3*(q_en < 0), ff_celt_model_energy_small);
352  } else if (left >= 1) {
353  q_en = av_clip(q_en, -1, 0);
354  ff_opus_rc_enc_log(rc, (q_en & 1), 1);
355  } else q_en = -1;
356 
357  block->error_energy[i] = q_en - diff;
358  prev[ch] += beta * q_en;
359  }
360  }
361 }
362 
364  float last_energy[][CELT_MAX_BANDS])
365 {
366  uint32_t inter, intra;
368 
369  exp_quant_coarse(rc, f, last_energy, 1);
370  intra = OPUS_RC_CHECKPOINT_BITS(rc);
371 
373 
374  exp_quant_coarse(rc, f, last_energy, 0);
375  inter = OPUS_RC_CHECKPOINT_BITS(rc);
376 
377  if (inter > intra) { /* Unlikely */
379  exp_quant_coarse(rc, f, last_energy, 1);
380  }
381 }
382 
384 {
385  for (int i = f->start_band; i < f->end_band; i++) {
386  if (!f->fine_bits[i])
387  continue;
388  for (int ch = 0; ch < f->channels; ch++) {
389  CeltBlock *block = &f->block[ch];
390  int quant, lim = (1 << f->fine_bits[i]);
391  float offset, diff = 0.5f - block->error_energy[i];
392  quant = av_clip(floor(diff*lim), 0, lim - 1);
393  ff_opus_rc_put_raw(rc, quant, f->fine_bits[i]);
394  offset = 0.5f - ((quant + 0.5f) * (1 << (14 - f->fine_bits[i])) / 16384.0f);
395  block->error_energy[i] -= offset;
396  }
397  }
398 }
399 
401 {
402  for (int priority = 0; priority < 2; priority++) {
403  for (int i = f->start_band; i < f->end_band && (f->framebits - opus_rc_tell(rc)) >= f->channels; i++) {
404  if (f->fine_priority[i] != priority || f->fine_bits[i] >= CELT_MAX_FINE_BITS)
405  continue;
406  for (int ch = 0; ch < f->channels; ch++) {
407  CeltBlock *block = &f->block[ch];
408  const float err = block->error_energy[i];
409  const float offset = 0.5f * (1 << (14 - f->fine_bits[i] - 1)) / 16384.0f;
410  const int sign = FFABS(err + offset) < FFABS(err - offset);
411  ff_opus_rc_put_raw(rc, sign, 1);
412  block->error_energy[i] -= offset*(1 - 2*sign);
413  }
414  }
415  }
416 }
417 
419  CeltFrame *f, int index)
420 {
422 
423  ff_opus_psy_celt_frame_init(&s->psyctx, f, index);
424 
426 
427  if (f->silence) {
428  if (f->framebits >= 16)
429  ff_opus_rc_enc_log(rc, 1, 15); /* Silence (if using explicit singalling) */
430  for (int ch = 0; ch < s->channels; ch++)
431  memset(s->last_quantized_energy[ch], 0.0f, sizeof(float)*CELT_MAX_BANDS);
432  return;
433  }
434 
435  /* Filters */
437  if (f->pfilter) {
438  ff_opus_rc_enc_log(rc, 0, 15);
440  }
441 
442  /* Transform */
443  celt_frame_mdct(s, f);
444 
445  /* Need to handle transient/non-transient switches at any point during analysis */
446  while (ff_opus_psy_celt_frame_process(&s->psyctx, f, index))
447  celt_frame_mdct(s, f);
448 
450 
451  /* Silence */
452  ff_opus_rc_enc_log(rc, 0, 15);
453 
454  /* Pitch filter */
455  if (!f->start_band && opus_rc_tell(rc) + 16 <= f->framebits)
457 
458  /* Transient flag */
459  if (f->size && opus_rc_tell(rc) + 3 <= f->framebits)
460  ff_opus_rc_enc_log(rc, f->transient, 3);
461 
462  /* Main encoding */
463  celt_quant_coarse (f, rc, s->last_quantized_energy);
464  celt_enc_tf (f, rc);
465  ff_celt_bitalloc (f, rc, 1);
466  celt_quant_fine (f, rc);
467  ff_celt_quant_bands(f, rc);
468 
469  /* Anticollapse bit */
470  if (f->anticollapse_needed)
471  ff_opus_rc_put_raw(rc, f->anticollapse, 1);
472 
473  /* Final per-band energy adjustments from leftover bits */
474  celt_quant_final(s, rc, f);
475 
476  for (int ch = 0; ch < f->channels; ch++) {
477  CeltBlock *block = &f->block[ch];
478  for (int i = 0; i < CELT_MAX_BANDS; i++)
479  s->last_quantized_energy[ch][i] = block->energy[i] + block->error_energy[i];
480  }
481 }
482 
483 static inline int write_opuslacing(uint8_t *dst, int v)
484 {
485  dst[0] = FFMIN(v - FFALIGN(v - 255, 4), v);
486  dst[1] = v - dst[0] >> 2;
487  return 1 + (v >= 252);
488 }
489 
491 {
492  int offset, fsize_needed;
493 
494  /* Write toc */
495  opus_gen_toc(s, avpkt->data, &offset, &fsize_needed);
496 
497  /* Frame sizes if needed */
498  if (fsize_needed) {
499  for (int i = 0; i < s->packet.frames - 1; i++) {
500  offset += write_opuslacing(avpkt->data + offset,
501  s->frame[i].framebits >> 3);
502  }
503  }
504 
505  /* Packets */
506  for (int i = 0; i < s->packet.frames; i++) {
507  ff_opus_rc_enc_end(&s->rc[i], avpkt->data + offset,
508  s->frame[i].framebits >> 3);
509  offset += s->frame[i].framebits >> 3;
510  }
511 
512  avpkt->size = offset;
513 }
514 
515 /* Used as overlap for the first frame and padding for the last encoded packet */
517 {
518  AVFrame *f = av_frame_alloc();
519  if (!f)
520  return NULL;
521  f->format = s->avctx->sample_fmt;
522  f->nb_samples = s->avctx->frame_size;
523  f->channel_layout = s->avctx->channel_layout;
524  if (av_frame_get_buffer(f, 4)) {
525  av_frame_free(&f);
526  return NULL;
527  }
528  for (int i = 0; i < s->channels; i++) {
529  size_t bps = av_get_bytes_per_sample(f->format);
530  memset(f->extended_data[i], 0, bps*f->nb_samples);
531  }
532  return f;
533 }
534 
535 static int opus_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
536  const AVFrame *frame, int *got_packet_ptr)
537 {
538  OpusEncContext *s = avctx->priv_data;
539  int ret, frame_size, alloc_size = 0;
540 
541  if (frame) { /* Add new frame to queue */
542  if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
543  return ret;
544  ff_bufqueue_add(avctx, &s->bufqueue, av_frame_clone(frame));
545  } else {
546  ff_opus_psy_signal_eof(&s->psyctx);
547  if (!s->afq.remaining_samples || !avctx->frame_number)
548  return 0; /* We've been flushed and there's nothing left to encode */
549  }
550 
551  /* Run the psychoacoustic system */
552  if (ff_opus_psy_process(&s->psyctx, &s->packet))
553  return 0;
554 
555  frame_size = OPUS_BLOCK_SIZE(s->packet.framesize);
556 
557  if (!frame) {
558  /* This can go negative, that's not a problem, we only pad if positive */
559  int pad_empty = s->packet.frames*(frame_size/s->avctx->frame_size) - s->bufqueue.available + 1;
560  /* Pad with empty 2.5 ms frames to whatever framesize was decided,
561  * this should only happen at the very last flush frame. The frames
562  * allocated here will be freed (because they have no other references)
563  * after they get used by celt_frame_setup_input() */
564  for (int i = 0; i < pad_empty; i++) {
565  AVFrame *empty = spawn_empty_frame(s);
566  if (!empty)
567  return AVERROR(ENOMEM);
568  ff_bufqueue_add(avctx, &s->bufqueue, empty);
569  }
570  }
571 
572  for (int i = 0; i < s->packet.frames; i++) {
573  celt_encode_frame(s, &s->rc[i], &s->frame[i], i);
574  alloc_size += s->frame[i].framebits >> 3;
575  }
576 
577  /* Worst case toc + the frame lengths if needed */
578  alloc_size += 2 + s->packet.frames*2;
579 
580  if ((ret = ff_alloc_packet2(avctx, avpkt, alloc_size, 0)) < 0)
581  return ret;
582 
583  /* Assemble packet */
584  opus_packet_assembler(s, avpkt);
585 
586  /* Update the psychoacoustic system */
587  ff_opus_psy_postencode_update(&s->psyctx, s->frame, s->rc);
588 
589  /* Remove samples from queue and skip if needed */
590  ff_af_queue_remove(&s->afq, s->packet.frames*frame_size, &avpkt->pts, &avpkt->duration);
591  if (s->packet.frames*frame_size > avpkt->duration) {
592  uint8_t *side = av_packet_new_side_data(avpkt, AV_PKT_DATA_SKIP_SAMPLES, 10);
593  if (!side)
594  return AVERROR(ENOMEM);
595  AV_WL32(&side[4], s->packet.frames*frame_size - avpkt->duration + 120);
596  }
597 
598  *got_packet_ptr = 1;
599 
600  return 0;
601 }
602 
604 {
605  OpusEncContext *s = avctx->priv_data;
606 
607  for (int i = 0; i < CELT_BLOCK_NB; i++)
608  ff_mdct15_uninit(&s->mdct[i]);
609 
610  ff_celt_pvq_uninit(&s->pvq);
611  av_freep(&s->dsp);
612  av_freep(&s->frame);
613  av_freep(&s->rc);
614  ff_af_queue_close(&s->afq);
615  ff_opus_psy_end(&s->psyctx);
616  ff_bufqueue_discard_all(&s->bufqueue);
617 
618  return 0;
619 }
620 
622 {
623  int ret, max_frames;
624  OpusEncContext *s = avctx->priv_data;
625 
626  s->avctx = avctx;
627  s->channels = avctx->channels;
628 
629  /* Opus allows us to change the framesize on each packet (and each packet may
630  * have multiple frames in it) but we can't change the codec's frame size on
631  * runtime, so fix it to the lowest possible number of samples and use a queue
632  * to accumulate AVFrames until we have enough to encode whatever the encoder
633  * decides is the best */
634  avctx->frame_size = 120;
635  /* Initial padding will change if SILK is ever supported */
636  avctx->initial_padding = 120;
637 
638  if (!avctx->bit_rate) {
639  int coupled = ff_opus_default_coupled_streams[s->channels - 1];
640  avctx->bit_rate = coupled*(96000) + (s->channels - coupled*2)*(48000);
641  } else if (avctx->bit_rate < 6000 || avctx->bit_rate > 255000 * s->channels) {
642  int64_t clipped_rate = av_clip(avctx->bit_rate, 6000, 255000 * s->channels);
643  av_log(avctx, AV_LOG_ERROR, "Unsupported bitrate %"PRId64" kbps, clipping to %"PRId64" kbps\n",
644  avctx->bit_rate/1000, clipped_rate/1000);
645  avctx->bit_rate = clipped_rate;
646  }
647 
648  /* Extradata */
649  avctx->extradata_size = 19;
651  if (!avctx->extradata)
652  return AVERROR(ENOMEM);
653  opus_write_extradata(avctx);
654 
655  ff_af_queue_init(avctx, &s->afq);
656 
657  if ((ret = ff_celt_pvq_init(&s->pvq, 1)) < 0)
658  return ret;
659 
660  if (!(s->dsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT)))
661  return AVERROR(ENOMEM);
662 
663  /* I have no idea why a base scaling factor of 68 works, could be the twiddles */
664  for (int i = 0; i < CELT_BLOCK_NB; i++)
665  if ((ret = ff_mdct15_init(&s->mdct[i], 0, i + 3, 68 << (CELT_BLOCK_NB - 1 - i))))
666  return AVERROR(ENOMEM);
667 
668  /* Zero out previous energy (matters for inter first frame) */
669  for (int ch = 0; ch < s->channels; ch++)
670  memset(s->last_quantized_energy[ch], 0.0f, sizeof(float)*CELT_MAX_BANDS);
671 
672  /* Allocate an empty frame to use as overlap for the first frame of audio */
673  ff_bufqueue_add(avctx, &s->bufqueue, spawn_empty_frame(s));
674  if (!ff_bufqueue_peek(&s->bufqueue, 0))
675  return AVERROR(ENOMEM);
676 
677  if ((ret = ff_opus_psy_init(&s->psyctx, s->avctx, &s->bufqueue, &s->options)))
678  return ret;
679 
680  /* Frame structs and range coder buffers */
681  max_frames = ceilf(FFMIN(s->options.max_delay_ms, 120.0f)/2.5f);
682  s->frame = av_malloc(max_frames*sizeof(CeltFrame));
683  if (!s->frame)
684  return AVERROR(ENOMEM);
685  s->rc = av_malloc(max_frames*sizeof(OpusRangeCoder));
686  if (!s->rc)
687  return AVERROR(ENOMEM);
688 
689  for (int i = 0; i < max_frames; i++) {
690  s->frame[i].dsp = s->dsp;
691  s->frame[i].avctx = s->avctx;
692  s->frame[i].seed = 0;
693  s->frame[i].pvq = s->pvq;
694  s->frame[i].apply_phase_inv = s->options.apply_phase_inv;
695  s->frame[i].block[0].emph_coeff = s->frame[i].block[1].emph_coeff = 0.0f;
696  }
697 
698  return 0;
699 }
700 
701 #define OPUSENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
702 static const AVOption opusenc_options[] = {
703  { "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" },
704  { "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" },
705  { NULL },
706 };
707 
708 static const AVClass opusenc_class = {
709  .class_name = "Opus encoder",
710  .item_name = av_default_item_name,
711  .option = opusenc_options,
712  .version = LIBAVUTIL_VERSION_INT,
713 };
714 
716  { "b", "0" },
717  { "compression_level", "10" },
718  { NULL },
719 };
720 
722  .name = "opus",
723  .long_name = NULL_IF_CONFIG_SMALL("Opus"),
724  .type = AVMEDIA_TYPE_AUDIO,
725  .id = AV_CODEC_ID_OPUS,
726  .defaults = opusenc_defaults,
727  .priv_class = &opusenc_class,
728  .priv_data_size = sizeof(OpusEncContext),
730  .encode2 = opus_encode_frame,
731  .close = opus_encode_end,
734  .supported_samplerates = (const int []){ 48000, 0 },
735  .channel_layouts = (const uint64_t []){ AV_CH_LAYOUT_MONO,
736  AV_CH_LAYOUT_STEREO, 0 },
737  .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
739 };
AVCodecContext::frame_size
int frame_size
Number of samples per channel in an audio frame.
Definition: avcodec.h:1118
AVCodec
AVCodec.
Definition: codec.h:197
AV_SAMPLE_FMT_FLTP
@ AV_SAMPLE_FMT_FLTP
float, planar
Definition: samplefmt.h:69
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
FF_CODEC_CAP_INIT_THREADSAFE
#define FF_CODEC_CAP_INIT_THREADSAFE
The codec does not modify any global variables in the init function, allowing to call the init functi...
Definition: internal.h:41
OpusEncContext::av_class
AVClass * av_class
Definition: opusenc.c:35
ff_opus_psy_process
int ff_opus_psy_process(OpusPsyContext *s, OpusPacketInfo *p)
Definition: opusenc_psy.c:223
av_clip
#define av_clip
Definition: common.h:122
spawn_empty_frame
static AVFrame * spawn_empty_frame(OpusEncContext *s)
Definition: opusenc.c:516
init
static av_cold int init(AVCodecContext *avctx)
Definition: avrndec.c:31
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:245
AVCodecContext::sample_rate
int sample_rate
samples per second
Definition: avcodec.h:1098
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:254
AV_CH_LAYOUT_MONO
#define AV_CH_LAYOUT_MONO
Definition: channel_layout.h:90
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:111
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:702
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:303
tmp
static uint8_t tmp[11]
Definition: aes_ctr.c:27
index
fg index
Definition: ffmpeg_filter.c:168
internal.h
AVPacket::data
uint8_t * data
Definition: packet.h:365
opusenc_defaults
static const AVCodecDefault opusenc_defaults[]
Definition: opusenc.c:715
AVOption
AVOption.
Definition: opt.h:248
b
#define b
Definition: input.c:41
OPUS_RC_CHECKPOINT_SPAWN
#define OPUS_RC_CHECKPOINT_SPAWN(rc)
Definition: opus_rc.h:116
opus_packet_assembler
static void opus_packet_assembler(OpusEncContext *s, AVPacket *avpkt)
Definition: opusenc.c:490
ff_mdct15_init
av_cold int ff_mdct15_init(MDCT15Context **ps, int inverse, int N, double scale)
Definition: mdct15.c:247
AVPacket::duration
int64_t duration
Duration of this packet in AVStream->time_base units, 0 if unknown.
Definition: packet.h:383
OpusEncContext::options
OpusEncOptions options
Definition: opusenc.c:36
opus_encode_end
static av_cold int opus_encode_end(AVCodecContext *avctx)
Definition: opusenc.c:603
ff_celt_pvq_init
int av_cold ff_celt_pvq_init(CeltPVQ **pvq, int encode)
Definition: opus_pvq.c:897
exp_quant_coarse
static void exp_quant_coarse(OpusRangeCoder *rc, CeltFrame *f, float last_energy[][CELT_MAX_BANDS], int intra)
Definition: opusenc.c:320
ceilf
static __device__ float ceilf(float a)
Definition: cuda_runtime.h:175
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:61
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:31
CeltBlock
Definition: opus_celt.h:66
win
static float win(SuperEqualizerContext *s, float n, int N)
Definition: af_superequalizer.c:119
CeltPVQ
Definition: opus_pvq.h:37
OPUSENC_FLAGS
#define OPUSENC_FLAGS
Definition: opusenc.c:701
audio_frame_queue.h
AVCodecContext::initial_padding
int initial_padding
Audio only.
Definition: avcodec.h:1807
opus_encode_frame
static int opus_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, const AVFrame *frame, int *got_packet_ptr)
Definition: opusenc.c:535
AVCodecContext::flags
int flags
AV_CODEC_FLAG_*.
Definition: avcodec.h:581
OpusEncContext::frame
CeltFrame * frame
Definition: opusenc.c:52
ff_opus_psy_end
av_cold int ff_opus_psy_end(OpusPsyContext *s)
Definition: opusenc_psy.c:593
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:91
quant
static int quant(float coef, const float Q, const float rounding)
Quantize one coefficient.
Definition: aacenc_utils.h:59
ff_opus_psy_postencode_update
void ff_opus_psy_postencode_update(OpusPsyContext *s, CeltFrame *f, OpusRangeCoder *rc)
Definition: opusenc_psy.c:479
av_frame_alloc
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
Definition: frame.c:98
AV_LOG_ERROR
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:181
OPUS_MAX_LOOKAHEAD
#define OPUS_MAX_LOOKAHEAD
Definition: opusenc.h:32
OpusEncContext::afq
AudioFrameQueue afq
Definition: opusenc.c:39
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:100
av_cold
#define av_cold
Definition: attributes.h:90
OPUS_BLOCK_SIZE
#define OPUS_BLOCK_SIZE(x)
Definition: opusenc.h:39
AVCodecContext::extradata_size
int extradata_size
Definition: avcodec.h:603
OpusEncContext::rc
OpusRangeCoder * rc
Definition: opusenc.c:53
OpusEncContext::scratch
float scratch[2048]
Definition: opusenc.c:58
s
#define s(width, name)
Definition: cbs_vp9.c:257
floor
static __device__ float floor(float a)
Definition: cuda_runtime.h:173
frame_size
int frame_size
Definition: mxfenc.c:2206
AVMEDIA_TYPE_AUDIO
@ AVMEDIA_TYPE_AUDIO
Definition: avutil.h:202
ff_opus_psy_celt_frame_process
int ff_opus_psy_celt_frame_process(OpusPsyContext *s, CeltFrame *f, int index)
Definition: opusenc_psy.c:455
CELT_MAX_FINE_BITS
#define CELT_MAX_FINE_BITS
Definition: opus_celt.h:41
bits
uint8_t bits
Definition: vp3data.h:141
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:423
MDCT15Context
Definition: mdct15.h:30
f
#define f(width, name)
Definition: cbs_vp9.c:255
OpusPsyContext
Definition: opusenc_psy.h:56
OpusEncContext::mdct
MDCT15Context * mdct[CELT_BLOCK_NB]
Definition: opusenc.c:41
FFABS
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition: common.h:72
AVCodecDefault
Definition: internal.h:207
OPUS_MODE_NB
@ OPUS_MODE_NB
Definition: opus.h:68
ff_celt_bitalloc
void ff_celt_bitalloc(CeltFrame *f, OpusRangeCoder *rc, int encode)
Definition: opus.c:555
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:67
NULL
#define NULL
Definition: coverity.c:32
OPUS_BANDWITH_NB
@ OPUS_BANDWITH_NB
Definition: opus.h:78
opusenc.h
OpusEncContext::packet
OpusPacketInfo packet
Definition: opusenc.c:48
ff_celt_pvq_uninit
void av_cold ff_celt_pvq_uninit(CeltPVQ **pvq)
Definition: opus_pvq.c:914
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:551
av_default_item_name
const char * av_default_item_name(void *ptr)
Return the context name.
Definition: log.c:235
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:191
CELT_MAX_BANDS
#define CELT_MAX_BANDS
Definition: opus.h:46
celt_quant_fine
static void celt_quant_fine(CeltFrame *f, OpusRangeCoder *rc)
Definition: opusenc.c:383
celt_frame_setup_input
static void celt_frame_setup_input(OpusEncContext *s, CeltFrame *f)
Definition: opusenc.c:117
opus_gen_toc
static int opus_gen_toc(OpusEncContext *s, uint8_t *toc, int *size, int *fsize_needed)
Definition: opusenc.c:74
ff_celt_freq_range
const uint8_t ff_celt_freq_range[]
Definition: opustab.c:772
float_dsp.h
for
for(j=16;j >0;--j)
Definition: h264pred_template.c:469
CELT_ENERGY_SILENCE
#define CELT_ENERGY_SILENCE
Definition: opus_celt.h:46
OpusEncContext::channels
int channels
Definition: opusenc.c:50
options
const OptionDef options[]
opustab.h
ff_opus_rc_enc_init
void ff_opus_rc_enc_init(OpusRangeCoder *rc)
Definition: opus_rc.c:402
AVPacket::size
int size
Definition: packet.h:366
NULL_IF_CONFIG_SMALL
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
Definition: internal.h:117
opus_pvq.h
opusenc_psy.h
FFMAX
#define FFMAX(a, b)
Definition: common.h:103
bps
unsigned bps
Definition: movenc.c:1595
AV_SAMPLE_FMT_NONE
@ AV_SAMPLE_FMT_NONE
Definition: samplefmt.h:59
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:45
OpusRangeCoder
Definition: opus_rc.h:40
OpusEncContext::psyctx
OpusPsyContext psyctx
Definition: opusenc.c:37
AV_CODEC_ID_OPUS
@ AV_CODEC_ID_OPUS
Definition: codec_id.h:481
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:376
FFMIN
#define FFMIN(a, b)
Definition: common.h:105
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:483
OpusEncContext
Definition: opusenc.c:34
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:61
ff_celt_beta_coef
const float ff_celt_beta_coef[]
Definition: opustab.c:804
AVCodecContext::channels
int channels
number of audio channels
Definition: avcodec.h:1099
AV_OPT_TYPE_FLOAT
@ AV_OPT_TYPE_FLOAT
Definition: opt.h:228
DECLARE_ALIGNED
#define DECLARE_ALIGNED(n, t, v)
Definition: mem.h:117
src1
#define src1
Definition: h264pred.c:140
OpusEncContext::last_quantized_energy
float last_quantized_energy[OPUS_MAX_CHANNELS][CELT_MAX_BANDS]
Definition: opusenc.c:56
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:369
i
int i
Definition: input.c:407
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
AVPacket::pts
int64_t pts
Presentation timestamp in AVStream->time_base units; the time at which the decompressed packet will b...
Definition: packet.h:358
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:106
AVCodecContext::extradata
uint8_t * extradata
some codecs need / can use extradata like Huffman tables.
Definition: avcodec.h:602
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: internal.h:49
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:350
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:418
AVSampleFormat
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:58
OPUS_MAX_CHANNELS
#define OPUS_MAX_CHANNELS
Definition: opusenc.h:34
ff_opus_psy_signal_eof
void ff_opus_psy_signal_eof(OpusPsyContext *s)
Definition: opusenc_psy.c:588
ff_mdct15_uninit
av_cold void ff_mdct15_uninit(MDCT15Context **ps)
Definition: mdct15.c:43
AVCodec::name
const char * name
Name of the codec implementation.
Definition: codec.h:204
AV_PKT_DATA_SKIP_SAMPLES
@ AV_PKT_DATA_SKIP_SAMPLES
Recommmends skipping the specified number of samples.
Definition: packet.h:156
len
int len
Definition: vorbis_enc_data.h:426
CELT_POSTFILTER_MINPERIOD
#define CELT_POSTFILTER_MINPERIOD
Definition: opus_celt.h:45
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:708
ret
ret
Definition: filter_design.txt:187
OpusEncContext::avctx
AVCodecContext * avctx
Definition: opusenc.c:38
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:72
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
AV_INPUT_BUFFER_PADDING_SIZE
#define AV_INPUT_BUFFER_PADDING_SIZE
Definition: avcodec.h:192
OpusEncContext::enc_id_bits
int enc_id_bits
Definition: opusenc.c:46
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:24
AVCodecContext
main external API structure.
Definition: avcodec.h:501
ff_celt_model_energy_small
const uint16_t ff_celt_model_energy_small[]
Definition: opustab.c:766
ff_celt_window
const float *const ff_celt_window
Definition: opustab.c:1135
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:220
celt_quant_coarse
static void celt_quant_coarse(CeltFrame *f, OpusRangeCoder *rc, float last_energy[][CELT_MAX_BANDS])
Definition: opusenc.c:363
ff_opus_encoder
const AVCodec ff_opus_encoder
Definition: opusenc.c:721
ff_celt_quant_bands
void ff_celt_quant_bands(CeltFrame *f, OpusRangeCoder *rc)
Definition: opus.c:446
ff_opus_rc_enc_log
void ff_opus_rc_enc_log(OpusRangeCoder *rc, int val, uint32_t bits)
Definition: opus_rc.c:131
temp
else temp
Definition: vf_mcdeint.c:259
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:77
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:283
celt_quant_final
static void celt_quant_final(OpusEncContext *s, OpusRangeCoder *rc, CeltFrame *f)
Definition: opusenc.c:400
AV_CODEC_FLAG_BITEXACT
#define AV_CODEC_FLAG_BITEXACT
Use only bitexact stuff (except (I)DCT).
Definition: avcodec.h:310
AVCodecContext::frame_number
int frame_number
Frame counter, set by libavcodec.
Definition: avcodec.h:1129
ff_opus_psy_init
av_cold int ff_opus_psy_init(OpusPsyContext *s, AVCodecContext *avctx, struct FFBufQueue *bufqueue, OpusEncOptions *options)
Definition: opusenc_psy.c:516
diff
static av_always_inline int diff(const uint32_t a, const uint32_t b)
Definition: vf_palettegen.c:136
OpusEncOptions
Definition: opusenc.h:43
FFALIGN
#define FFALIGN(x, a)
Definition: macros.h:48
alpha
static const int16_t alpha[]
Definition: ilbcdata.h:55
AVCodecContext::priv_data
void * priv_data
Definition: avcodec.h:528
AVPacket
This structure stores compressed data.
Definition: packet.h:342
AV_OPT_TYPE_BOOL
@ AV_OPT_TYPE_BOOL
Definition: opt.h:242
OpusEncContext::pvq
CeltPVQ * pvq
Definition: opusenc.c:42
av_freep
#define av_freep(p)
Definition: tableprint_vlc.h:35
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:63
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:28
celt_apply_preemph_filter
static void celt_apply_preemph_filter(OpusEncContext *s, CeltFrame *f)
Definition: opusenc.c:157
ff_celt_mean_energy
const float ff_celt_mean_energy[]
Definition: opustab.c:792
OpusEncContext::bufqueue
struct FFBufQueue bufqueue
Definition: opusenc.c:43
CELT_OVERLAP
#define CELT_OVERLAP
Definition: opus.h:43
OPUS_RC_CHECKPOINT_BITS
#define OPUS_RC_CHECKPOINT_BITS(rc)
Definition: opus_rc.h:120
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:82
OPUS_RC_CHECKPOINT_ROLLBACK
#define OPUS_RC_CHECKPOINT_ROLLBACK(rc)
Definition: opus_rc.h:123
ff_alloc_packet2
int ff_alloc_packet2(AVCodecContext *avctx, AVPacket *avpkt, int64_t size, int64_t min_size)
Check AVPacket size and/or allocate data.
Definition: encode.c:33
celt_enc_tf
static void celt_enc_tf(CeltFrame *f, OpusRangeCoder *rc)
Definition: opusenc.c:256
opus_encode_init
static av_cold int opus_encode_init(AVCodecContext *avctx)
Definition: opusenc.c:621
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:93
OpusEncContext::dsp
AVFloatDSPContext * dsp
Definition: opusenc.c:40