FFmpeg
enc.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 "enc.h"
26 #include "pvq.h"
27 #include "enc_psy.h"
28 #include "tab.h"
29 
31 #include "libavutil/float_dsp.h"
32 #include "libavutil/mem.h"
33 #include "libavutil/mem_internal.h"
34 #include "libavutil/opt.h"
35 #include "bytestream.h"
36 #include "audio_frame_queue.h"
37 #include "codec_internal.h"
38 
39 typedef struct OpusEncContext {
50 
51  uint8_t enc_id[64];
53 
55 
56  int channels;
57 
60 
61  /* Actual energy the decoder will have */
63 
64  DECLARE_ALIGNED(32, float, scratch)[2048];
66 
68 {
69  uint8_t *bs = avctx->extradata;
70 
71  bytestream_put_buffer(&bs, "OpusHead", 8);
72  bytestream_put_byte (&bs, 0x1);
73  bytestream_put_byte (&bs, avctx->ch_layout.nb_channels);
74  bytestream_put_le16 (&bs, avctx->initial_padding);
75  bytestream_put_le32 (&bs, avctx->sample_rate);
76  bytestream_put_le16 (&bs, 0x0);
77  bytestream_put_byte (&bs, 0x0); /* Default layout */
78 }
79 
80 static int opus_gen_toc(OpusEncContext *s, uint8_t *toc, int *size, int *fsize_needed)
81 {
82  int tmp = 0x0, extended_toc = 0;
83  static const int toc_cfg[][OPUS_MODE_NB][OPUS_BANDWITH_NB] = {
84  /* Silk Hybrid Celt Layer */
85  /* NB MB WB SWB FB NB MB WB SWB FB NB MB WB SWB FB Bandwidth */
86  { { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 }, { 17, 0, 21, 25, 29 } }, /* 2.5 ms */
87  { { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 }, { 18, 0, 22, 26, 30 } }, /* 5 ms */
88  { { 1, 5, 9, 0, 0 }, { 0, 0, 0, 13, 15 }, { 19, 0, 23, 27, 31 } }, /* 10 ms */
89  { { 2, 6, 10, 0, 0 }, { 0, 0, 0, 14, 16 }, { 20, 0, 24, 28, 32 } }, /* 20 ms */
90  { { 3, 7, 11, 0, 0 }, { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }, /* 40 ms */
91  { { 4, 8, 12, 0, 0 }, { 0, 0, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }, /* 60 ms */
92  };
93  int cfg = toc_cfg[s->packet.framesize][s->packet.mode][s->packet.bandwidth];
94  *fsize_needed = 0;
95  if (!cfg)
96  return 1;
97  if (s->packet.frames == 2) { /* 2 packets */
98  if (s->frame[0].framebits == s->frame[1].framebits) { /* same size */
99  tmp = 0x1;
100  } else { /* different size */
101  tmp = 0x2;
102  *fsize_needed = 1; /* put frame sizes in the packet */
103  }
104  } else if (s->packet.frames > 2) {
105  tmp = 0x3;
106  extended_toc = 1;
107  }
108  tmp |= (s->channels > 1) << 2; /* Stereo or mono */
109  tmp |= (cfg - 1) << 3; /* codec configuration */
110  *toc++ = tmp;
111  if (extended_toc) {
112  for (int i = 0; i < (s->packet.frames - 1); i++)
113  *fsize_needed |= (s->frame[i].framebits != s->frame[i + 1].framebits);
114  tmp = (*fsize_needed) << 7; /* vbr flag */
115  tmp |= (0) << 6; /* padding flag */
116  tmp |= s->packet.frames;
117  *toc++ = tmp;
118  }
119  *size = 1 + extended_toc;
120  return 0;
121 }
122 
124 {
125  AVFrame *cur = NULL;
126  const int subframesize = s->avctx->frame_size;
127  int subframes = OPUS_BLOCK_SIZE(s->packet.framesize) / subframesize;
128 
129  cur = ff_bufqueue_get(&s->bufqueue);
130 
131  for (int ch = 0; ch < f->channels; ch++) {
132  CeltBlock *b = &f->block[ch];
133  const void *input = cur->extended_data[ch];
134  size_t bps = av_get_bytes_per_sample(cur->format);
135  memcpy(b->overlap, input, bps*cur->nb_samples);
136  }
137 
138  av_frame_free(&cur);
139 
140  for (int sf = 0; sf < subframes; sf++) {
141  if (sf != (subframes - 1))
142  cur = ff_bufqueue_get(&s->bufqueue);
143  else
144  cur = ff_bufqueue_peek(&s->bufqueue, 0);
145 
146  for (int ch = 0; ch < f->channels; ch++) {
147  CeltBlock *b = &f->block[ch];
148  const void *input = cur->extended_data[ch];
149  const size_t bps = av_get_bytes_per_sample(cur->format);
150  const size_t left = (subframesize - cur->nb_samples)*bps;
151  const size_t len = FFMIN(subframesize, cur->nb_samples)*bps;
152  memcpy(&b->samples[sf*subframesize], input, len);
153  memset(&b->samples[cur->nb_samples], 0, left);
154  }
155 
156  /* Last frame isn't popped off and freed yet - we need it for overlap */
157  if (sf != (subframes - 1))
158  av_frame_free(&cur);
159  }
160 }
161 
162 /* Apply the pre emphasis filter */
164 {
165  const int subframesize = s->avctx->frame_size;
166  const int subframes = OPUS_BLOCK_SIZE(s->packet.framesize) / subframesize;
167  const float c = ff_opus_deemph_weights[0];
168 
169  /* Filter overlap */
170  for (int ch = 0; ch < f->channels; ch++) {
171  CeltBlock *b = &f->block[ch];
172  float m = b->emph_coeff;
173  for (int i = 0; i < CELT_OVERLAP; i++) {
174  float sample = b->overlap[i];
175  b->overlap[i] = sample - m;
176  m = sample * c;
177  }
178  b->emph_coeff = m;
179  }
180 
181  /* Filter the samples but do not update the last subframe's coeff - overlap ^^^ */
182  for (int sf = 0; sf < subframes; sf++) {
183  for (int ch = 0; ch < f->channels; ch++) {
184  CeltBlock *b = &f->block[ch];
185  float m = b->emph_coeff;
186  for (int i = 0; i < subframesize; i++) {
187  float sample = b->samples[sf*subframesize + i];
188  b->samples[sf*subframesize + i] = sample - m;
189  m = sample * c;
190  }
191  if (sf != (subframes - 1))
192  b->emph_coeff = m;
193  }
194  }
195 }
196 
197 /* Create the window and do the mdct */
199 {
200  float *win = s->scratch, *temp = s->scratch + 1920;
201 
202  if (f->transient) {
203  for (int ch = 0; ch < f->channels; ch++) {
204  CeltBlock *b = &f->block[ch];
205  float *src1 = b->overlap;
206  for (int t = 0; t < f->blocks; t++) {
207  float *src2 = &b->samples[CELT_OVERLAP*t];
208  s->dsp->vector_fmul(win, src1, ff_celt_window, 128);
209  s->dsp->vector_fmul_reverse(&win[CELT_OVERLAP], src2,
210  ff_celt_window_padded, 128);
211  src1 = src2;
212  s->tx_fn[0](s->tx[0], b->coeffs + t, win, sizeof(float)*f->blocks);
213  }
214  }
215  } else {
216  int blk_len = OPUS_BLOCK_SIZE(f->size), wlen = OPUS_BLOCK_SIZE(f->size + 1);
217  int rwin = blk_len - CELT_OVERLAP, lap_dst = (wlen - blk_len - CELT_OVERLAP) >> 1;
218  memset(win, 0, wlen*sizeof(float));
219  for (int ch = 0; ch < f->channels; ch++) {
220  CeltBlock *b = &f->block[ch];
221 
222  /* Overlap */
223  s->dsp->vector_fmul(temp, b->overlap, ff_celt_window, 128);
224  memcpy(win + lap_dst, temp, CELT_OVERLAP*sizeof(float));
225 
226  /* Samples, flat top window */
227  memcpy(&win[lap_dst + CELT_OVERLAP], b->samples, rwin*sizeof(float));
228 
229  /* Samples, windowed */
230  s->dsp->vector_fmul_reverse(temp, b->samples + rwin,
231  ff_celt_window_padded, 128);
232  memcpy(win + lap_dst + blk_len, temp, CELT_OVERLAP*sizeof(float));
233 
234  s->tx_fn[f->size](s->tx[f->size], b->coeffs, win, sizeof(float));
235  }
236  }
237 
238  for (int ch = 0; ch < f->channels; ch++) {
239  CeltBlock *block = &f->block[ch];
240  for (int i = 0; i < CELT_MAX_BANDS; i++) {
241  float ener = 0.0f;
242  int band_offset = ff_celt_freq_bands[i] << f->size;
243  int band_size = ff_celt_freq_range[i] << f->size;
244  float *coeffs = &block->coeffs[band_offset];
245 
246  for (int j = 0; j < band_size; j++)
247  ener += coeffs[j]*coeffs[j];
248 
249  block->lin_energy[i] = sqrtf(ener) + FLT_EPSILON;
250  ener = 1.0f/block->lin_energy[i];
251 
252  for (int j = 0; j < band_size; j++)
253  coeffs[j] *= ener;
254 
255  block->energy[i] = log2f(block->lin_energy[i]) - ff_celt_mean_energy[i];
256 
257  /* CELT_ENERGY_SILENCE is what the decoder uses and its not -infinity */
258  block->energy[i] = FFMAX(block->energy[i], CELT_ENERGY_SILENCE);
259  }
260  }
261 }
262 
264 {
265  int tf_select = 0, diff = 0, tf_changed = 0, tf_select_needed;
266  int bits = f->transient ? 2 : 4;
267 
268  tf_select_needed = ((f->size && (opus_rc_tell(rc) + bits + 1) <= f->framebits));
269 
270  for (int i = f->start_band; i < f->end_band; i++) {
271  if ((opus_rc_tell(rc) + bits + tf_select_needed) <= f->framebits) {
272  const int tbit = (diff ^ 1) == f->tf_change[i];
273  ff_opus_rc_enc_log(rc, tbit, bits);
274  diff ^= tbit;
275  tf_changed |= diff;
276  }
277  bits = f->transient ? 4 : 5;
278  }
279 
280  if (tf_select_needed && ff_celt_tf_select[f->size][f->transient][0][tf_changed] !=
281  ff_celt_tf_select[f->size][f->transient][1][tf_changed]) {
282  ff_opus_rc_enc_log(rc, f->tf_select, 1);
283  tf_select = f->tf_select;
284  }
285 
286  for (int i = f->start_band; i < f->end_band; i++)
287  f->tf_change[i] = ff_celt_tf_select[f->size][f->transient][tf_select][f->tf_change[i]];
288 }
289 
291 {
292  float gain = f->pf_gain;
293  int txval, octave = f->pf_octave, period = f->pf_period, tapset = f->pf_tapset;
294 
295  ff_opus_rc_enc_log(rc, f->pfilter, 1);
296  if (!f->pfilter)
297  return;
298 
299  /* Octave */
300  txval = FFMIN(octave, 6);
301  ff_opus_rc_enc_uint(rc, txval, 6);
302  octave = txval;
303  /* Period */
304  txval = av_clip(period - (16 << octave) + 1, 0, (1 << (4 + octave)) - 1);
305  ff_opus_rc_put_raw(rc, period, 4 + octave);
306  period = txval + (16 << octave) - 1;
307  /* Gain */
308  txval = FFMIN(((int)(gain / 0.09375f)) - 1, 7);
309  ff_opus_rc_put_raw(rc, txval, 3);
310  gain = 0.09375f * (txval + 1);
311  /* Tapset */
312  if ((opus_rc_tell(rc) + 2) <= f->framebits)
314  else
315  tapset = 0;
316  /* Finally create the coeffs */
317  for (int i = 0; i < 2; i++) {
318  CeltBlock *block = &f->block[i];
319 
320  block->pf_period_new = FFMAX(period, CELT_POSTFILTER_MINPERIOD);
321  block->pf_gains_new[0] = gain * ff_celt_postfilter_taps[tapset][0];
322  block->pf_gains_new[1] = gain * ff_celt_postfilter_taps[tapset][1];
323  block->pf_gains_new[2] = gain * ff_celt_postfilter_taps[tapset][2];
324  }
325 }
326 
328  float last_energy[][CELT_MAX_BANDS], int intra)
329 {
330  float alpha, beta, prev[2] = { 0, 0 };
331  const uint8_t *pmod = ff_celt_coarse_energy_dist[f->size][intra];
332 
333  /* Inter is really just differential coding */
334  if (opus_rc_tell(rc) + 3 <= f->framebits)
335  ff_opus_rc_enc_log(rc, intra, 3);
336  else
337  intra = 0;
338 
339  if (intra) {
340  alpha = 0.0f;
341  beta = 1.0f - (4915.0f/32768.0f);
342  } else {
343  alpha = ff_celt_alpha_coef[f->size];
344  beta = ff_celt_beta_coef[f->size];
345  }
346 
347  for (int i = f->start_band; i < f->end_band; i++) {
348  for (int ch = 0; ch < f->channels; ch++) {
349  CeltBlock *block = &f->block[ch];
350  const int left = f->framebits - opus_rc_tell(rc);
351  const float last = FFMAX(-9.0f, last_energy[ch][i]);
352  float diff = block->energy[i] - prev[ch] - last*alpha;
353  int q_en = lrintf(diff);
354  if (left >= 15) {
355  ff_opus_rc_enc_laplace(rc, &q_en, pmod[i << 1] << 7, pmod[(i << 1) + 1] << 6);
356  } else if (left >= 2) {
357  q_en = av_clip(q_en, -1, 1);
358  ff_opus_rc_enc_cdf(rc, 2*q_en + 3*(q_en < 0), ff_celt_model_energy_small);
359  } else if (left >= 1) {
360  q_en = av_clip(q_en, -1, 0);
361  ff_opus_rc_enc_log(rc, (q_en & 1), 1);
362  } else q_en = -1;
363 
364  block->error_energy[i] = q_en - diff;
365  prev[ch] += beta * q_en;
366  }
367  }
368 }
369 
371  float last_energy[][CELT_MAX_BANDS])
372 {
373  uint32_t inter, intra;
375 
376  exp_quant_coarse(rc, f, last_energy, 1);
377  intra = OPUS_RC_CHECKPOINT_BITS(rc);
378 
380 
381  exp_quant_coarse(rc, f, last_energy, 0);
382  inter = OPUS_RC_CHECKPOINT_BITS(rc);
383 
384  if (inter > intra) { /* Unlikely */
386  exp_quant_coarse(rc, f, last_energy, 1);
387  }
388 }
389 
391 {
392  for (int i = f->start_band; i < f->end_band; i++) {
393  if (!f->fine_bits[i])
394  continue;
395  for (int ch = 0; ch < f->channels; ch++) {
396  CeltBlock *block = &f->block[ch];
397  int quant, lim = (1 << f->fine_bits[i]);
398  float offset, diff = 0.5f - block->error_energy[i];
399  quant = av_clip(floor(diff*lim), 0, lim - 1);
400  ff_opus_rc_put_raw(rc, quant, f->fine_bits[i]);
401  offset = 0.5f - ((quant + 0.5f) * (1 << (14 - f->fine_bits[i])) / 16384.0f);
402  block->error_energy[i] -= offset;
403  }
404  }
405 }
406 
408 {
409  for (int priority = 0; priority < 2; priority++) {
410  for (int i = f->start_band; i < f->end_band && (f->framebits - opus_rc_tell(rc)) >= f->channels; i++) {
411  if (f->fine_priority[i] != priority || f->fine_bits[i] >= CELT_MAX_FINE_BITS)
412  continue;
413  for (int ch = 0; ch < f->channels; ch++) {
414  CeltBlock *block = &f->block[ch];
415  const float err = block->error_energy[i];
416  const float offset = 0.5f * (1 << (14 - f->fine_bits[i] - 1)) / 16384.0f;
417  const int sign = FFABS(err + offset) < FFABS(err - offset);
418  ff_opus_rc_put_raw(rc, sign, 1);
419  block->error_energy[i] -= offset*(1 - 2*sign);
420  }
421  }
422  }
423 }
424 
426  CeltFrame *f, int index)
427 {
429 
430  ff_opus_psy_celt_frame_init(&s->psyctx, f, index);
431 
433 
434  if (f->silence) {
435  if (f->framebits >= 16)
436  ff_opus_rc_enc_log(rc, 1, 15); /* Silence (if using explicit singalling) */
437  for (int ch = 0; ch < s->channels; ch++)
438  memset(s->last_quantized_energy[ch], 0.0f, sizeof(float)*CELT_MAX_BANDS);
439  return;
440  }
441 
442  /* Filters */
444  if (f->pfilter) {
445  ff_opus_rc_enc_log(rc, 0, 15);
447  }
448 
449  /* Transform */
450  celt_frame_mdct(s, f);
451 
452  /* Need to handle transient/non-transient switches at any point during analysis */
453  while (ff_opus_psy_celt_frame_process(&s->psyctx, f, index))
454  celt_frame_mdct(s, f);
455 
457 
458  /* Silence */
459  ff_opus_rc_enc_log(rc, 0, 15);
460 
461  /* Pitch filter */
462  if (!f->start_band && opus_rc_tell(rc) + 16 <= f->framebits)
464 
465  /* Transient flag */
466  if (f->size && opus_rc_tell(rc) + 3 <= f->framebits)
467  ff_opus_rc_enc_log(rc, f->transient, 3);
468 
469  /* Main encoding */
470  celt_quant_coarse (f, rc, s->last_quantized_energy);
471  celt_enc_tf (f, rc);
472  ff_celt_bitalloc (f, rc, 1);
473  celt_quant_fine (f, rc);
474  ff_celt_quant_bands(f, rc);
475 
476  /* Anticollapse bit */
477  if (f->anticollapse_needed)
478  ff_opus_rc_put_raw(rc, f->anticollapse, 1);
479 
480  /* Final per-band energy adjustments from leftover bits */
481  celt_quant_final(s, rc, f);
482 
483  for (int ch = 0; ch < f->channels; ch++) {
484  CeltBlock *block = &f->block[ch];
485  for (int i = 0; i < CELT_MAX_BANDS; i++)
486  s->last_quantized_energy[ch][i] = block->energy[i] + block->error_energy[i];
487  }
488 }
489 
490 static inline int write_opuslacing(uint8_t *dst, int v)
491 {
492  dst[0] = FFMIN(v - FFALIGN(v - 255, 4), v);
493  dst[1] = v - dst[0] >> 2;
494  return 1 + (v >= 252);
495 }
496 
498 {
499  int offset, fsize_needed;
500 
501  /* Write toc */
502  opus_gen_toc(s, avpkt->data, &offset, &fsize_needed);
503 
504  /* Frame sizes if needed */
505  if (fsize_needed) {
506  for (int i = 0; i < s->packet.frames - 1; i++) {
507  offset += write_opuslacing(avpkt->data + offset,
508  s->frame[i].framebits >> 3);
509  }
510  }
511 
512  /* Packets */
513  for (int i = 0; i < s->packet.frames; i++) {
514  ff_opus_rc_enc_end(&s->rc[i], avpkt->data + offset,
515  s->frame[i].framebits >> 3);
516  offset += s->frame[i].framebits >> 3;
517  }
518 
519  avpkt->size = offset;
520 }
521 
522 /* Used as overlap for the first frame and padding for the last encoded packet */
524 {
525  AVFrame *f = av_frame_alloc();
526  int ret;
527  if (!f)
528  return NULL;
529  f->format = s->avctx->sample_fmt;
530  f->nb_samples = s->avctx->frame_size;
531  ret = av_channel_layout_copy(&f->ch_layout, &s->avctx->ch_layout);
532  if (ret < 0) {
533  av_frame_free(&f);
534  return NULL;
535  }
536  if (av_frame_get_buffer(f, 4)) {
537  av_frame_free(&f);
538  return NULL;
539  }
540  for (int i = 0; i < s->channels; i++) {
541  size_t bps = av_get_bytes_per_sample(f->format);
542  memset(f->extended_data[i], 0, bps*f->nb_samples);
543  }
544  return f;
545 }
546 
547 static int opus_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
548  const AVFrame *frame, int *got_packet_ptr)
549 {
550  OpusEncContext *s = avctx->priv_data;
551  int ret, frame_size, alloc_size = 0;
552 
553  if (frame) { /* Add new frame to queue */
554  if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
555  return ret;
556  ff_bufqueue_add(avctx, &s->bufqueue, av_frame_clone(frame));
557  } else {
558  ff_opus_psy_signal_eof(&s->psyctx);
559  if (!s->afq.remaining_samples || !avctx->frame_num)
560  return 0; /* We've been flushed and there's nothing left to encode */
561  }
562 
563  /* Run the psychoacoustic system */
564  if (ff_opus_psy_process(&s->psyctx, &s->packet))
565  return 0;
566 
567  frame_size = OPUS_BLOCK_SIZE(s->packet.framesize);
568 
569  if (!frame) {
570  /* This can go negative, that's not a problem, we only pad if positive */
571  int pad_empty = s->packet.frames*(frame_size/s->avctx->frame_size) - s->bufqueue.available + 1;
572  /* Pad with empty 2.5 ms frames to whatever framesize was decided,
573  * this should only happen at the very last flush frame. The frames
574  * allocated here will be freed (because they have no other references)
575  * after they get used by celt_frame_setup_input() */
576  for (int i = 0; i < pad_empty; i++) {
577  AVFrame *empty = spawn_empty_frame(s);
578  if (!empty)
579  return AVERROR(ENOMEM);
580  ff_bufqueue_add(avctx, &s->bufqueue, empty);
581  }
582  }
583 
584  for (int i = 0; i < s->packet.frames; i++) {
585  celt_encode_frame(s, &s->rc[i], &s->frame[i], i);
586  alloc_size += s->frame[i].framebits >> 3;
587  }
588 
589  /* Worst case toc + the frame lengths if needed */
590  alloc_size += 2 + s->packet.frames*2;
591 
592  if ((ret = ff_alloc_packet(avctx, avpkt, alloc_size)) < 0)
593  return ret;
594 
595  /* Assemble packet */
596  opus_packet_assembler(s, avpkt);
597 
598  /* Update the psychoacoustic system */
599  ff_opus_psy_postencode_update(&s->psyctx, s->frame);
600 
601  /* Remove samples from queue and skip if needed */
602  ff_af_queue_remove(&s->afq, s->packet.frames*frame_size, &avpkt->pts, &avpkt->duration);
603  if (s->packet.frames*frame_size > avpkt->duration) {
604  uint8_t *side = av_packet_new_side_data(avpkt, AV_PKT_DATA_SKIP_SAMPLES, 10);
605  if (!side)
606  return AVERROR(ENOMEM);
607  AV_WL32(&side[4], s->packet.frames*frame_size - avpkt->duration + 120);
608  }
609 
610  *got_packet_ptr = 1;
611 
612  return 0;
613 }
614 
616 {
617  OpusEncContext *s = avctx->priv_data;
618 
619  for (int i = 0; i < CELT_BLOCK_NB; i++)
620  av_tx_uninit(&s->tx[i]);
621 
622  ff_celt_pvq_uninit(&s->pvq);
623  av_freep(&s->dsp);
624  av_freep(&s->frame);
625  av_freep(&s->rc);
626  ff_af_queue_close(&s->afq);
627  ff_opus_psy_end(&s->psyctx);
628  ff_bufqueue_discard_all(&s->bufqueue);
629 
630  return 0;
631 }
632 
634 {
635  int ret, max_frames;
636  OpusEncContext *s = avctx->priv_data;
637 
638  s->avctx = avctx;
639  s->channels = avctx->ch_layout.nb_channels;
640 
641  /* Opus allows us to change the framesize on each packet (and each packet may
642  * have multiple frames in it) but we can't change the codec's frame size on
643  * runtime, so fix it to the lowest possible number of samples and use a queue
644  * to accumulate AVFrames until we have enough to encode whatever the encoder
645  * decides is the best */
646  avctx->frame_size = 120;
647  /* Initial padding will change if SILK is ever supported */
648  avctx->initial_padding = 120;
649 
650  if (!avctx->bit_rate) {
651  int coupled = ff_opus_default_coupled_streams[s->channels - 1];
652  avctx->bit_rate = coupled*(96000) + (s->channels - coupled*2)*(48000);
653  } else if (avctx->bit_rate < 6000 || avctx->bit_rate > 255000 * s->channels) {
654  int64_t clipped_rate = av_clip(avctx->bit_rate, 6000, 255000 * s->channels);
655  av_log(avctx, AV_LOG_ERROR, "Unsupported bitrate %"PRId64" kbps, clipping to %"PRId64" kbps\n",
656  avctx->bit_rate/1000, clipped_rate/1000);
657  avctx->bit_rate = clipped_rate;
658  }
659 
660  /* Extradata */
661  avctx->extradata_size = 19;
663  if (!avctx->extradata)
664  return AVERROR(ENOMEM);
665  opus_write_extradata(avctx);
666 
667  ff_af_queue_init(avctx, &s->afq);
668 
669  if ((ret = ff_celt_pvq_init(&s->pvq, 1)) < 0)
670  return ret;
671 
672  if (!(s->dsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT)))
673  return AVERROR(ENOMEM);
674 
675  /* I have no idea why a base scaling factor of 68 works, could be the twiddles */
676  for (int i = 0; i < CELT_BLOCK_NB; i++) {
677  const float scale = 68 << (CELT_BLOCK_NB - 1 - i);
678  if ((ret = av_tx_init(&s->tx[i], &s->tx_fn[i], AV_TX_FLOAT_MDCT, 0, 15 << (i + 3), &scale, 0)))
679  return AVERROR(ENOMEM);
680  }
681 
682  /* Zero out previous energy (matters for inter first frame) */
683  for (int ch = 0; ch < s->channels; ch++)
684  memset(s->last_quantized_energy[ch], 0.0f, sizeof(float)*CELT_MAX_BANDS);
685 
686  /* Allocate an empty frame to use as overlap for the first frame of audio */
687  ff_bufqueue_add(avctx, &s->bufqueue, spawn_empty_frame(s));
688  if (!ff_bufqueue_peek(&s->bufqueue, 0))
689  return AVERROR(ENOMEM);
690 
691  if ((ret = ff_opus_psy_init(&s->psyctx, s->avctx, &s->bufqueue, &s->options)))
692  return ret;
693 
694  /* Frame structs and range coder buffers */
695  max_frames = ceilf(FFMIN(s->options.max_delay_ms, 120.0f)/2.5f);
696  s->frame = av_malloc(max_frames*sizeof(CeltFrame));
697  if (!s->frame)
698  return AVERROR(ENOMEM);
699  s->rc = av_malloc(max_frames*sizeof(OpusRangeCoder));
700  if (!s->rc)
701  return AVERROR(ENOMEM);
702 
703  for (int i = 0; i < max_frames; i++) {
704  s->frame[i].dsp = s->dsp;
705  s->frame[i].avctx = s->avctx;
706  s->frame[i].seed = 0;
707  s->frame[i].pvq = s->pvq;
708  s->frame[i].apply_phase_inv = s->options.apply_phase_inv;
709  s->frame[i].block[0].emph_coeff = s->frame[i].block[1].emph_coeff = 0.0f;
710  }
711 
712  return 0;
713 }
714 
715 #define OPUSENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
716 static const AVOption opusenc_options[] = {
717  { "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, .unit = "max_delay_ms" },
718  { "apply_phase_inv", "Apply intensity stereo phase inversion", offsetof(OpusEncContext, options.apply_phase_inv), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, OPUSENC_FLAGS, .unit = "apply_phase_inv" },
719  { NULL },
720 };
721 
722 static const AVClass opusenc_class = {
723  .class_name = "Opus encoder",
724  .item_name = av_default_item_name,
725  .option = opusenc_options,
726  .version = LIBAVUTIL_VERSION_INT,
727 };
728 
730  { "b", "0" },
731  { "compression_level", "10" },
732  { NULL },
733 };
734 
736  .p.name = "opus",
737  CODEC_LONG_NAME("Opus"),
738  .p.type = AVMEDIA_TYPE_AUDIO,
739  .p.id = AV_CODEC_ID_OPUS,
740  .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY |
742  .defaults = opusenc_defaults,
743  .p.priv_class = &opusenc_class,
744  .priv_data_size = sizeof(OpusEncContext),
747  .close = opus_encode_end,
748  .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
749  .p.supported_samplerates = (const int []){ 48000, 0 },
750  .p.ch_layouts = (const AVChannelLayout []){ AV_CHANNEL_LAYOUT_MONO,
751  AV_CHANNEL_LAYOUT_STEREO, { 0 } },
752  .p.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
754 };
ff_celt_window_padded
const float ff_celt_window_padded[136]
Definition: tab.c:1104
AVCodecContext::frame_size
int frame_size
Number of samples per channel in an audio frame.
Definition: avcodec.h:1083
AV_SAMPLE_FMT_FLTP
@ AV_SAMPLE_FMT_FLTP
float, planar
Definition: samplefmt.h:66
OpusEncContext::av_class
AVClass * av_class
Definition: enc.c:40
celt_quant_fine
static void celt_quant_fine(CeltFrame *f, OpusRangeCoder *rc)
Definition: enc.c:390
av_clip
#define av_clip
Definition: common.h:100
opus_encode_init
static av_cold int opus_encode_init(AVCodecContext *avctx)
Definition: enc.c:633
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:43
CELT_MAX_BANDS
#define CELT_MAX_BANDS
Definition: celt.h:43
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
AV_WL32
#define AV_WL32(p, v)
Definition: intreadwrite.h:422
opus_packet_assembler
static void opus_packet_assembler(OpusEncContext *s, AVPacket *avpkt)
Definition: enc.c:497
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:304
AV_CHANNEL_LAYOUT_STEREO
#define AV_CHANNEL_LAYOUT_STEREO
Definition: channel_layout.h:393
AVCodecContext::sample_rate
int sample_rate
samples per second
Definition: avcodec.h:1056
ff_af_queue_close
void ff_af_queue_close(AudioFrameQueue *afq)
Close AudioFrameQueue.
Definition: audio_frame_queue.c:36
CELT_ENERGY_SILENCE
#define CELT_ENERGY_SILENCE
Definition: celt.h:53
log2f
#define log2f(x)
Definition: libm.h:409
opusenc_options
static const AVOption opusenc_options[]
Definition: enc.c:716
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: rc.h:62
src1
const pixel * src1
Definition: h264pred_template.c:421
AVTXContext
Definition: tx_priv.h:235
int64_t
long long int64_t
Definition: coverity.c:34
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:162
opusenc_class
static const AVClass opusenc_class
Definition: enc.c:722
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:389
tmp
static uint8_t tmp[11]
Definition: aes_ctr.c:28
CELT_BLOCK_NB
@ CELT_BLOCK_NB
Definition: celt.h:68
AVPacket::data
uint8_t * data
Definition: packet.h:539
AVOption
AVOption.
Definition: opt.h:429
encode.h
b
#define b
Definition: input.c:41
CELT_MAX_FINE_BITS
#define CELT_MAX_FINE_BITS
Definition: celt.h:48
FFCodec
Definition: codec_internal.h:127
celt_enc_tf
static void celt_enc_tf(CeltFrame *f, OpusRangeCoder *rc)
Definition: enc.c:263
float.h
opus_gen_toc
static int opus_gen_toc(OpusEncContext *s, uint8_t *toc, int *size, int *fsize_needed)
Definition: enc.c:80
ff_opus_rc_enc_laplace
void ff_opus_rc_enc_laplace(OpusRangeCoder *rc, int *value, uint32_t symbol, int decay)
Definition: rc.c:314
AVPacket::duration
int64_t duration
Duration of this packet in AVStream->time_base units, 0 if unknown.
Definition: packet.h:557
OpusEncContext::options
OpusEncOptions options
Definition: enc.c:41
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
ff_opus_deemph_weights
const float ff_opus_deemph_weights[]
Definition: tab.c:1169
AVChannelLayout::nb_channels
int nb_channels
Number of channels in this layout.
Definition: channel_layout.h:327
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:903
OPUS_RC_CHECKPOINT_SPAWN
#define OPUS_RC_CHECKPOINT_SPAWN(rc)
Definition: rc.h:117
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
av_malloc
#define av_malloc(s)
Definition: tableprint_vlc.h:30
CeltBlock
Definition: celt.h:71
win
static float win(SuperEqualizerContext *s, float n, int N)
Definition: af_superequalizer.c:119
CeltPVQ
Definition: pvq.h:37
ff_opus_rc_enc_init
void ff_opus_rc_enc_init(OpusRangeCoder *rc)
Definition: rc.c:402
ff_opus_psy_postencode_update
void ff_opus_psy_postencode_update(OpusPsyContext *s, CeltFrame *f)
Definition: enc_psy.c:480
FFCodecDefault
Definition: codec_internal.h:97
FFCodec::p
AVCodec p
The public AVCodec.
Definition: codec_internal.h:131
ff_opus_rc_enc_end
void ff_opus_rc_enc_end(OpusRangeCoder *rc, uint8_t *dst, int size)
Definition: rc.c:360
AVCodecContext::ch_layout
AVChannelLayout ch_layout
Audio channel layout.
Definition: avcodec.h:1071
audio_frame_queue.h
AVCodecContext::initial_padding
int initial_padding
Audio only.
Definition: avcodec.h:1128
AVCodecContext::flags
int flags
AV_CODEC_FLAG_*.
Definition: avcodec.h:508
OpusEncContext::frame
CeltFrame * frame
Definition: enc.c:58
OPUS_MAX_LOOKAHEAD
#define OPUS_MAX_LOOKAHEAD
Definition: enc.h:32
FF_CODEC_ENCODE_CB
#define FF_CODEC_ENCODE_CB(func)
Definition: codec_internal.h:320
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
ff_celt_model_tapset
const uint16_t ff_celt_model_tapset[]
Definition: tab.c:758
ff_opus_psy_signal_eof
void ff_opus_psy_signal_eof(OpusPsyContext *s)
Definition: enc_psy.c:590
av_frame_alloc
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
Definition: frame.c:150
quant
static const uint8_t quant[64]
Definition: vmixdec.c:71
AV_LOG_ERROR
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:209
OpusEncContext::afq
AudioFrameQueue afq
Definition: enc.c:44
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:102
av_cold
#define av_cold
Definition: attributes.h:90
ff_celt_postfilter_taps
const float ff_celt_postfilter_taps[3][3]
Definition: tab.c:1098
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:151
celt_frame_mdct
static void celt_frame_mdct(OpusEncContext *s, CeltFrame *f)
Definition: enc.c:198
float
float
Definition: af_crystalizer.c:122
AVCodecContext::extradata_size
int extradata_size
Definition: avcodec.h:530
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: enc.c:59
OpusEncContext::scratch
float scratch[2048]
Definition: enc.c:64
s
#define s(width, name)
Definition: cbs_vp9.c:198
floor
static __device__ float floor(float a)
Definition: cuda_runtime.h:173
frame_size
int frame_size
Definition: mxfenc.c:2429
AVMEDIA_TYPE_AUDIO
@ AVMEDIA_TYPE_AUDIO
Definition: avutil.h:202
pvq.h
bits
uint8_t bits
Definition: vp3data.h:128
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:609
ff_celt_pvq_uninit
void av_cold ff_celt_pvq_uninit(CeltPVQ **pvq)
Definition: pvq.c:927
OPUS_BLOCK_SIZE
#define OPUS_BLOCK_SIZE(x)
Definition: enc.h:39
opus_encode_end
static av_cold int opus_encode_end(AVCodecContext *avctx)
Definition: enc.c:615
OpusPsyContext
Definition: enc_psy.h:52
CODEC_LONG_NAME
#define CODEC_LONG_NAME(str)
Definition: codec_internal.h:296
FFABS
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition: common.h:74
OPUS_MODE_NB
@ OPUS_MODE_NB
Definition: opus.h:46
OpusPacketInfo
Definition: enc.h:48
ff_celt_tf_select
const int8_t ff_celt_tf_select[4][2][2][2]
Definition: tab.c:782
LIBAVUTIL_VERSION_INT
#define LIBAVUTIL_VERSION_INT
Definition: version.h:85
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:75
NULL
#define NULL
Definition: coverity.c:32
OPUS_BANDWITH_NB
@ OPUS_BANDWITH_NB
Definition: opus.h:56
celt_quant_coarse
static void celt_quant_coarse(CeltFrame *f, OpusRangeCoder *rc, float last_energy[][CELT_MAX_BANDS])
Definition: enc.c:370
ff_celt_freq_range
const uint8_t ff_celt_freq_range[]
Definition: tab.c:772
OpusEncContext::packet
OpusPacketInfo packet
Definition: enc.c:54
ff_opus_psy_end
av_cold int ff_opus_psy_end(OpusPsyContext *s)
Definition: enc_psy.c:595
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:501
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
options
Definition: swscale.c:42
sqrtf
static __device__ float sqrtf(float a)
Definition: cuda_runtime.h:184
CELT_POSTFILTER_MINPERIOD
#define CELT_POSTFILTER_MINPERIOD
Definition: celt.h:52
enc_psy.h
ff_opus_psy_celt_frame_init
void ff_opus_psy_celt_frame_init(OpusPsyContext *s, CeltFrame *f, int index)
Definition: enc_psy.c:258
ff_celt_coarse_energy_dist
const uint8_t ff_celt_coarse_energy_dist[4][2][42]
Definition: tab.c:808
celt_enc_quant_pfilter
static void celt_enc_quant_pfilter(OpusRangeCoder *rc, CeltFrame *f)
Definition: enc.c:290
index
int index
Definition: gxfenc.c:90
c
Undefined Behavior In the C some operations are like signed integer dereferencing freed accessing outside allocated Undefined Behavior must not occur in a C it is not safe even if the output of undefined operations is unused The unsafety may seem nit picking but Optimizing compilers have in fact optimized code on the assumption that no undefined Behavior occurs Optimizing code based on wrong assumptions can and has in some cases lead to effects beyond the output of computations The signed integer overflow problem in speed critical code Code which is highly optimized and works with signed integers sometimes has the problem that often the output of the computation does not c
Definition: undefined.txt:32
opusenc_defaults
static const FFCodecDefault opusenc_defaults[]
Definition: enc.c:729
float_dsp.h
opus_encode_frame
static int opus_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, const AVFrame *frame, int *got_packet_ptr)
Definition: enc.c:547
OpusEncContext::channels
int channels
Definition: enc.c:56
ff_celt_beta_coef
const float ff_celt_beta_coef[]
Definition: tab.c:804
f
f
Definition: af_crystalizer.c:122
init
int(* init)(AVBSFContext *ctx)
Definition: dts2pts.c:368
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
enc.h
AVPacket::size
int size
Definition: packet.h:540
ff_opus_default_coupled_streams
const uint8_t ff_opus_default_coupled_streams[]
Definition: tab.c:27
AVChannelLayout
An AVChannelLayout holds information about the channel layout of audio data.
Definition: channel_layout.h:317
codec_internal.h
DECLARE_ALIGNED
#define DECLARE_ALIGNED(n, t, v)
Definition: mem_internal.h:102
dst
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition: dsp.h:83
ff_celt_window
static const float *const ff_celt_window
Definition: tab.h:162
exp_quant_coarse
static void exp_quant_coarse(OpusRangeCoder *rc, CeltFrame *f, float last_energy[][CELT_MAX_BANDS], int intra)
Definition: enc.c:327
for
for(k=2;k<=8;++k)
Definition: h264pred_template.c:425
bps
unsigned bps
Definition: movenc.c:1880
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: enc.c:51
OpusRangeCoder
Definition: rc.h:41
ff_opus_encoder
const FFCodec ff_opus_encoder
Definition: enc.c:735
OpusEncContext::psyctx
OpusPsyContext psyctx
Definition: enc.c:42
AV_CODEC_ID_OPUS
@ AV_CODEC_ID_OPUS
Definition: codec_id.h:506
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:476
ff_opus_psy_process
int ff_opus_psy_process(OpusPsyContext *s, OpusPacketInfo *p)
Definition: enc_psy.c:227
diff
static av_always_inline int diff(const struct color_info *a, const struct color_info *b, const int trans_thresh)
Definition: vf_paletteuse.c:166
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
OpusEncContext
Definition: enc.c:39
ff_opus_psy_celt_frame_process
int ff_opus_psy_celt_frame_process(OpusPsyContext *s, CeltFrame *f, int index)
Definition: enc_psy.c:459
ff_celt_quant_bands
void ff_celt_quant_bands(CeltFrame *f, OpusRangeCoder *rc)
Definition: celt.c:28
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
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:295
celt_apply_preemph_filter
static void celt_apply_preemph_filter(OpusEncContext *s, CeltFrame *f)
Definition: enc.c:163
AV_OPT_TYPE_FLOAT
@ AV_OPT_TYPE_FLOAT
Underlying C type is float.
Definition: opt.h:271
OpusEncContext::last_quantized_energy
float last_quantized_energy[OPUS_MAX_CHANNELS][CELT_MAX_BANDS]
Definition: enc.c:62
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:469
tab.h
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
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:256
AVPacket::pts
int64_t pts
Presentation timestamp in AVStream->time_base units; the time at which the decompressed packet will b...
Definition: packet.h:532
celt_encode_frame
static void celt_encode_frame(OpusEncContext *s, OpusRangeCoder *rc, CeltFrame *f, int index)
Definition: enc.c:425
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:529
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:450
src2
const pixel * src2
Definition: h264pred_template.c:422
AVSampleFormat
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:55
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
AVCodec::name
const char * name
Name of the codec implementation.
Definition: codec.h:194
write_opuslacing
static int write_opuslacing(uint8_t *dst, int v)
Definition: enc.c:490
len
int len
Definition: vorbis_enc_data.h:426
ff_celt_model_energy_small
const uint16_t ff_celt_model_energy_small[]
Definition: tab.c:766
spawn_empty_frame
static AVFrame * spawn_empty_frame(OpusEncContext *s)
Definition: enc.c:523
opus_write_extradata
static void opus_write_extradata(AVCodecContext *avctx)
Definition: enc.c:67
OpusEncContext::tx
AVTXContext * tx[CELT_BLOCK_NB]
Definition: enc.c:46
ff_opus_rc_enc_cdf
void ff_opus_rc_enc_cdf(OpusRangeCoder *rc, int val, const uint16_t *cdf)
Definition: rc.c:109
AVCodecContext::frame_num
int64_t frame_num
Frame counter, set by libavcodec.
Definition: avcodec.h:2041
ret
ret
Definition: filter_design.txt:187
OpusEncContext::avctx
AVCodecContext * avctx
Definition: enc.c:43
ff_celt_bitalloc
void ff_celt_bitalloc(CeltFrame *f, OpusRangeCoder *rc, int encode)
Definition: celt.c:137
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:80
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
CELT_OVERLAP
#define CELT_OVERLAP
Definition: celt.h:40
celt_quant_final
static void celt_quant_final(OpusEncContext *s, OpusRangeCoder *rc, CeltFrame *f)
Definition: enc.c:407
AV_INPUT_BUFFER_PADDING_SIZE
#define AV_INPUT_BUFFER_PADDING_SIZE
Definition: defs.h:40
OpusEncContext::enc_id_bits
int enc_id_bits
Definition: enc.c:52
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
AVCodecContext
main external API structure.
Definition: avcodec.h:451
ff_celt_alpha_coef
const float ff_celt_alpha_coef[]
Definition: tab.c:800
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: packet.c:231
celt_frame_setup_input
static void celt_frame_setup_input(OpusEncContext *s, CeltFrame *f)
Definition: enc.c:123
OpusEncContext::tx_fn
av_tx_fn tx_fn[CELT_BLOCK_NB]
Definition: enc.c:47
temp
else temp
Definition: vf_mcdeint.c:263
AV_PKT_DATA_SKIP_SAMPLES
@ AV_PKT_DATA_SKIP_SAMPLES
Recommmends skipping the specified number of samples.
Definition: packet.h:153
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:76
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:447
ff_celt_mean_energy
const float ff_celt_mean_energy[]
Definition: tab.c:792
OPUS_MAX_CHANNELS
#define OPUS_MAX_CHANNELS
Definition: enc.h:34
mem.h
AV_CODEC_FLAG_BITEXACT
#define AV_CODEC_FLAG_BITEXACT
Use only bitexact stuff (except (I)DCT).
Definition: avcodec.h:342
AV_CHANNEL_LAYOUT_MONO
#define AV_CHANNEL_LAYOUT_MONO
Definition: channel_layout.h:392
scale
static void scale(int *out, const int *in, const int w, const int h, const int shift)
Definition: intra.c:291
OpusEncOptions
Definition: enc.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:478
AVPacket
This structure stores compressed data.
Definition: packet.h:516
AV_OPT_TYPE_BOOL
@ AV_OPT_TYPE_BOOL
Underlying C type is int.
Definition: opt.h:327
OpusEncContext::pvq
CeltPVQ * pvq
Definition: enc.c:48
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:146
ff_celt_pvq_init
int av_cold ff_celt_pvq_init(CeltPVQ **pvq, int encode)
Definition: pvq.c:907
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: rc.c:161
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
OPUS_RC_CHECKPOINT_ROLLBACK
#define OPUS_RC_CHECKPOINT_ROLLBACK(rc)
Definition: rc.h:124
ff_celt_freq_bands
const uint8_t ff_celt_freq_bands[]
Definition: tab.c:768
ff_opus_psy_init
av_cold int ff_opus_psy_init(OpusPsyContext *s, AVCodecContext *avctx, struct FFBufQueue *bufqueue, OpusEncOptions *options)
Definition: enc_psy.c:516
OpusEncContext::bufqueue
struct FFBufQueue bufqueue
Definition: enc.c:49
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:81
ff_opus_rc_enc_log
void ff_opus_rc_enc_log(OpusRangeCoder *rc, int val, uint32_t bits)
Definition: rc.c:131
ff_alloc_packet
int ff_alloc_packet(AVCodecContext *avctx, AVPacket *avpkt, int64_t size)
Check AVPacket size and allocate data.
Definition: encode.c:62
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: rc.c:204
OPUS_RC_CHECKPOINT_BITS
#define OPUS_RC_CHECKPOINT_BITS(rc)
Definition: rc.h:121
OPUSENC_FLAGS
#define OPUSENC_FLAGS
Definition: enc.c:715
CeltFrame
Definition: celt.h:98
OpusEncContext::dsp
AVFloatDSPContext * dsp
Definition: enc.c:45