3 * Copyright (C) 2008 Konstantin Shishkov
5 * This file is part of Libav.
7 * Libav 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.
12 * Libav 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.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
27 /***********************************
29 * add sane pulse detection
30 * add temporal noise shaping
31 ***********************************/
33 #include "libavutil/opt.h"
37 #include "mpeg4audio.h"
47 #define AAC_MAX_CHANNELS 6
49 static const uint8_t swb_size_1024_96[] = {
50 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8,
51 12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44,
52 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
55 static const uint8_t swb_size_1024_64[] = {
56 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8,
57 12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36,
58 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40
61 static const uint8_t swb_size_1024_48[] = {
62 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
63 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
64 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32,
68 static const uint8_t swb_size_1024_32[] = {
69 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
70 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
71 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
74 static const uint8_t swb_size_1024_24[] = {
75 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
76 12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28,
77 32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64
80 static const uint8_t swb_size_1024_16[] = {
81 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
82 12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28,
83 32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64
86 static const uint8_t swb_size_1024_8[] = {
87 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
88 16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28,
89 32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80
92 static const uint8_t *swb_size_1024[] = {
93 swb_size_1024_96, swb_size_1024_96, swb_size_1024_64,
94 swb_size_1024_48, swb_size_1024_48, swb_size_1024_32,
95 swb_size_1024_24, swb_size_1024_24, swb_size_1024_16,
96 swb_size_1024_16, swb_size_1024_16, swb_size_1024_8
99 static const uint8_t swb_size_128_96[] = {
100 4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36
103 static const uint8_t swb_size_128_48[] = {
104 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16
107 static const uint8_t swb_size_128_24[] = {
108 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20
111 static const uint8_t swb_size_128_16[] = {
112 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20
115 static const uint8_t swb_size_128_8[] = {
116 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20
119 static const uint8_t *swb_size_128[] = {
120 /* the last entry on the following row is swb_size_128_64 but is a
121 duplicate of swb_size_128_96 */
122 swb_size_128_96, swb_size_128_96, swb_size_128_96,
123 swb_size_128_48, swb_size_128_48, swb_size_128_48,
124 swb_size_128_24, swb_size_128_24, swb_size_128_16,
125 swb_size_128_16, swb_size_128_16, swb_size_128_8
128 /** default channel configurations */
129 static const uint8_t aac_chan_configs[6][5] = {
130 {1, TYPE_SCE}, // 1 channel - single channel element
131 {1, TYPE_CPE}, // 2 channels - channel pair
132 {2, TYPE_SCE, TYPE_CPE}, // 3 channels - center + stereo
133 {3, TYPE_SCE, TYPE_CPE, TYPE_SCE}, // 4 channels - front center + stereo + back center
134 {3, TYPE_SCE, TYPE_CPE, TYPE_CPE}, // 5 channels - front center + stereo + back stereo
135 {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE
139 * Make AAC audio config object.
140 * @see 1.6.2.1 "Syntax - AudioSpecificConfig"
142 static void put_audio_specific_config(AVCodecContext *avctx)
145 AACEncContext *s = avctx->priv_data;
147 init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8);
148 put_bits(&pb, 5, 2); //object type - AAC-LC
149 put_bits(&pb, 4, s->samplerate_index); //sample rate index
150 put_bits(&pb, 4, avctx->channels);
152 put_bits(&pb, 1, 0); //frame length - 1024 samples
153 put_bits(&pb, 1, 0); //does not depend on core coder
154 put_bits(&pb, 1, 0); //is not extension
156 //Explicitly Mark SBR absent
157 put_bits(&pb, 11, 0x2b7); //sync extension
158 put_bits(&pb, 5, AOT_SBR);
163 static av_cold int aac_encode_init(AVCodecContext *avctx)
165 AACEncContext *s = avctx->priv_data;
167 const uint8_t *sizes[2];
170 avctx->frame_size = 1024;
172 for (i = 0; i < 16; i++)
173 if (avctx->sample_rate == ff_mpeg4audio_sample_rates[i])
176 av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate);
179 if (avctx->channels > AAC_MAX_CHANNELS) {
180 av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels);
183 if (avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW) {
184 av_log(avctx, AV_LOG_ERROR, "Unsupported profile %d\n", avctx->profile);
187 if (1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * avctx->channels) {
188 av_log(avctx, AV_LOG_ERROR, "Too many bits per frame requested\n");
191 s->samplerate_index = i;
193 dsputil_init(&s->dsp, avctx);
194 ff_mdct_init(&s->mdct1024, 11, 0, 1.0);
195 ff_mdct_init(&s->mdct128, 8, 0, 1.0);
197 ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
198 ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
199 ff_init_ff_sine_windows(10);
200 ff_init_ff_sine_windows(7);
202 s->chan_map = aac_chan_configs[avctx->channels-1];
203 s->samples = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0]));
204 s->cpe = av_mallocz(sizeof(ChannelElement) * s->chan_map[0]);
205 avctx->extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
206 avctx->extradata_size = 5;
207 put_audio_specific_config(avctx);
209 sizes[0] = swb_size_1024[i];
210 sizes[1] = swb_size_128[i];
211 lengths[0] = ff_aac_num_swb_1024[i];
212 lengths[1] = ff_aac_num_swb_128[i];
213 ff_psy_init(&s->psy, avctx, 2, sizes, lengths, s->chan_map[0], &s->chan_map[1]);
214 s->psypp = ff_psy_preprocess_init(avctx);
215 s->coder = &ff_aac_coders[2];
217 s->lambda = avctx->global_quality ? avctx->global_quality : 120;
224 static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s,
225 SingleChannelElement *sce, short *audio)
228 const int chans = avctx->channels;
229 const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
230 const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
231 const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
232 float *output = sce->ret;
234 if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
235 memcpy(output, sce->saved, sizeof(float)*1024);
236 if (sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE) {
237 memset(output, 0, sizeof(output[0]) * 448);
238 for (i = 448; i < 576; i++)
239 output[i] = sce->saved[i] * pwindow[i - 448];
240 for (i = 576; i < 704; i++)
241 output[i] = sce->saved[i];
243 if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) {
244 for (i = 0; i < 1024; i++) {
245 output[i+1024] = audio[i * chans] * lwindow[1024 - i - 1];
246 sce->saved[i] = audio[i * chans] * lwindow[i];
249 for (i = 0; i < 448; i++)
250 output[i+1024] = audio[i * chans];
252 output[i+1024] = audio[i * chans] * swindow[576 - i - 1];
253 memset(output+1024+576, 0, sizeof(output[0]) * 448);
254 for (i = 0; i < 1024; i++)
255 sce->saved[i] = audio[i * chans];
257 s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
259 for (k = 0; k < 1024; k += 128) {
260 for (i = 448 + k; i < 448 + k + 256; i++)
261 output[i - 448 - k] = (i < 1024)
263 : audio[(i-1024)*chans];
264 s->dsp.vector_fmul (output, output, k ? swindow : pwindow, 128);
265 s->dsp.vector_fmul_reverse(output+128, output+128, swindow, 128);
266 s->mdct128.mdct_calc(&s->mdct128, sce->coeffs + k, output);
268 for (i = 0; i < 1024; i++)
269 sce->saved[i] = audio[i * chans];
274 * Encode ics_info element.
275 * @see Table 4.6 (syntax of ics_info)
277 static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
281 put_bits(&s->pb, 1, 0); // ics_reserved bit
282 put_bits(&s->pb, 2, info->window_sequence[0]);
283 put_bits(&s->pb, 1, info->use_kb_window[0]);
284 if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
285 put_bits(&s->pb, 6, info->max_sfb);
286 put_bits(&s->pb, 1, 0); // no prediction
288 put_bits(&s->pb, 4, info->max_sfb);
289 for (w = 1; w < 8; w++)
290 put_bits(&s->pb, 1, !info->group_len[w]);
296 * @see 4.6.8.1 "Joint Coding - M/S Stereo"
298 static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
302 put_bits(pb, 2, cpe->ms_mode);
303 if (cpe->ms_mode == 1)
304 for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w])
305 for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
306 put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
310 * Produce integer coefficients from scalefactors provided by the model.
312 static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans)
315 int start, maxsfb, cmaxsfb;
317 for (ch = 0; ch < chans; ch++) {
318 IndividualChannelStream *ics = &cpe->ch[ch].ics;
321 cpe->ch[ch].pulse.num_pulse = 0;
322 for (w = 0; w < ics->num_windows*16; w += 16) {
323 for (g = 0; g < ics->num_swb; g++) {
325 if (cpe->common_window && !ch && cpe->ms_mask[w + g]) {
326 for (i = 0; i < ics->swb_sizes[g]; i++) {
327 cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
328 cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
331 start += ics->swb_sizes[g];
333 for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--)
335 maxsfb = FFMAX(maxsfb, cmaxsfb);
337 ics->max_sfb = maxsfb;
339 //adjust zero bands for window groups
340 for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
341 for (g = 0; g < ics->max_sfb; g++) {
343 for (w2 = w; w2 < w + ics->group_len[w]; w2++) {
344 if (!cpe->ch[ch].zeroes[w2*16 + g]) {
349 cpe->ch[ch].zeroes[w*16 + g] = i;
354 if (chans > 1 && cpe->common_window) {
355 IndividualChannelStream *ics0 = &cpe->ch[0].ics;
356 IndividualChannelStream *ics1 = &cpe->ch[1].ics;
358 ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
359 ics1->max_sfb = ics0->max_sfb;
360 for (w = 0; w < ics0->num_windows*16; w += 16)
361 for (i = 0; i < ics0->max_sfb; i++)
362 if (cpe->ms_mask[w+i])
364 if (msc == 0 || ics0->max_sfb == 0)
367 cpe->ms_mode = msc < ics0->max_sfb * ics0->num_windows ? 1 : 2;
372 * Encode scalefactor band coding type.
374 static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
378 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
379 s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
383 * Encode scalefactors.
385 static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s,
386 SingleChannelElement *sce)
388 int off = sce->sf_idx[0], diff;
391 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
392 for (i = 0; i < sce->ics.max_sfb; i++) {
393 if (!sce->zeroes[w*16 + i]) {
394 diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO;
395 if (diff < 0 || diff > 120)
396 av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n");
397 off = sce->sf_idx[w*16 + i];
398 put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
407 static void encode_pulses(AACEncContext *s, Pulse *pulse)
411 put_bits(&s->pb, 1, !!pulse->num_pulse);
412 if (!pulse->num_pulse)
415 put_bits(&s->pb, 2, pulse->num_pulse - 1);
416 put_bits(&s->pb, 6, pulse->start);
417 for (i = 0; i < pulse->num_pulse; i++) {
418 put_bits(&s->pb, 5, pulse->pos[i]);
419 put_bits(&s->pb, 4, pulse->amp[i]);
424 * Encode spectral coefficients processed by psychoacoustic model.
426 static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
430 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
432 for (i = 0; i < sce->ics.max_sfb; i++) {
433 if (sce->zeroes[w*16 + i]) {
434 start += sce->ics.swb_sizes[i];
437 for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++)
438 s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128,
439 sce->ics.swb_sizes[i],
440 sce->sf_idx[w*16 + i],
441 sce->band_type[w*16 + i],
443 start += sce->ics.swb_sizes[i];
449 * Encode one channel of audio data.
451 static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
452 SingleChannelElement *sce,
455 put_bits(&s->pb, 8, sce->sf_idx[0]);
457 put_ics_info(s, &sce->ics);
458 encode_band_info(s, sce);
459 encode_scale_factors(avctx, s, sce);
460 encode_pulses(s, &sce->pulse);
461 put_bits(&s->pb, 1, 0); //tns
462 put_bits(&s->pb, 1, 0); //ssr
463 encode_spectral_coeffs(s, sce);
468 * Write some auxiliary information about the created AAC file.
470 static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s,
473 int i, namelen, padbits;
475 namelen = strlen(name) + 2;
476 put_bits(&s->pb, 3, TYPE_FIL);
477 put_bits(&s->pb, 4, FFMIN(namelen, 15));
479 put_bits(&s->pb, 8, namelen - 16);
480 put_bits(&s->pb, 4, 0); //extension type - filler
481 padbits = 8 - (put_bits_count(&s->pb) & 7);
482 align_put_bits(&s->pb);
483 for (i = 0; i < namelen - 2; i++)
484 put_bits(&s->pb, 8, name[i]);
485 put_bits(&s->pb, 12 - padbits, 0);
488 static int aac_encode_frame(AVCodecContext *avctx,
489 uint8_t *frame, int buf_size, void *data)
491 AACEncContext *s = avctx->priv_data;
492 int16_t *samples = s->samples, *samples2, *la;
494 int i, ch, w, g, chans, tag, start_ch;
495 int chan_el_counter[4];
496 FFPsyWindowInfo windows[AAC_MAX_CHANNELS];
502 memcpy(s->samples + 1024 * avctx->channels, data,
503 1024 * avctx->channels * sizeof(s->samples[0]));
506 samples2 = s->samples + 1024 * avctx->channels;
507 for (i = 0; i < s->chan_map[0]; i++) {
508 tag = s->chan_map[i+1];
509 chans = tag == TYPE_CPE ? 2 : 1;
510 ff_psy_preprocess(s->psypp, (uint16_t*)data + start_ch,
511 samples2 + start_ch, start_ch, chans);
516 if (!avctx->frame_number) {
517 memcpy(s->samples, s->samples + 1024 * avctx->channels,
518 1024 * avctx->channels * sizeof(s->samples[0]));
523 for (i = 0; i < s->chan_map[0]; i++) {
524 FFPsyWindowInfo* wi = windows + start_ch;
525 tag = s->chan_map[i+1];
526 chans = tag == TYPE_CPE ? 2 : 1;
528 for (ch = 0; ch < chans; ch++) {
529 IndividualChannelStream *ics = &cpe->ch[ch].ics;
530 int cur_channel = start_ch + ch;
531 samples2 = samples + cur_channel;
532 la = samples2 + (448+64) * avctx->channels;
535 if (tag == TYPE_LFE) {
536 wi[ch].window_type[0] = ONLY_LONG_SEQUENCE;
537 wi[ch].window_shape = 0;
538 wi[ch].num_windows = 1;
539 wi[ch].grouping[0] = 1;
541 wi[ch] = s->psy.model->window(&s->psy, samples2, la, cur_channel,
542 ics->window_sequence[0]);
544 ics->window_sequence[1] = ics->window_sequence[0];
545 ics->window_sequence[0] = wi[ch].window_type[0];
546 ics->use_kb_window[1] = ics->use_kb_window[0];
547 ics->use_kb_window[0] = wi[ch].window_shape;
548 ics->num_windows = wi[ch].num_windows;
549 ics->swb_sizes = s->psy.bands [ics->num_windows == 8];
550 ics->num_swb = tag == TYPE_LFE ? 12 : s->psy.num_bands[ics->num_windows == 8];
551 for (w = 0; w < ics->num_windows; w++)
552 ics->group_len[w] = wi[ch].grouping[w];
554 apply_window_and_mdct(avctx, s, &cpe->ch[ch], samples2);
560 init_put_bits(&s->pb, frame, buf_size*8);
561 if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT))
562 put_bitstream_info(avctx, s, LIBAVCODEC_IDENT);
564 memset(chan_el_counter, 0, sizeof(chan_el_counter));
565 for (i = 0; i < s->chan_map[0]; i++) {
566 FFPsyWindowInfo* wi = windows + start_ch;
567 const float *coeffs[2];
568 tag = s->chan_map[i+1];
569 chans = tag == TYPE_CPE ? 2 : 1;
571 put_bits(&s->pb, 3, tag);
572 put_bits(&s->pb, 4, chan_el_counter[tag]++);
573 for (ch = 0; ch < chans; ch++)
574 coeffs[ch] = cpe->ch[ch].coeffs;
575 s->psy.model->analyze(&s->psy, start_ch, coeffs, wi);
576 for (ch = 0; ch < chans; ch++) {
577 s->cur_channel = start_ch * 2 + ch;
578 s->coder->search_for_quantizers(avctx, s, &cpe->ch[ch], s->lambda);
580 cpe->common_window = 0;
582 && wi[0].window_type[0] == wi[1].window_type[0]
583 && wi[0].window_shape == wi[1].window_shape) {
585 cpe->common_window = 1;
586 for (w = 0; w < wi[0].num_windows; w++) {
587 if (wi[0].grouping[w] != wi[1].grouping[w]) {
588 cpe->common_window = 0;
593 s->cur_channel = start_ch * 2;
594 if (s->options.stereo_mode && cpe->common_window) {
595 if (s->options.stereo_mode > 0) {
596 IndividualChannelStream *ics = &cpe->ch[0].ics;
597 for (w = 0; w < ics->num_windows; w += ics->group_len[w])
598 for (g = 0; g < ics->num_swb; g++)
599 cpe->ms_mask[w*16+g] = 1;
600 } else if (s->coder->search_for_ms) {
601 s->coder->search_for_ms(s, cpe, s->lambda);
604 adjust_frame_information(s, cpe, chans);
606 put_bits(&s->pb, 1, cpe->common_window);
607 if (cpe->common_window) {
608 put_ics_info(s, &cpe->ch[0].ics);
609 encode_ms_info(&s->pb, cpe);
612 for (ch = 0; ch < chans; ch++) {
613 s->cur_channel = start_ch + ch;
614 encode_individual_channel(avctx, s, &cpe->ch[ch], cpe->common_window);
619 frame_bits = put_bits_count(&s->pb);
620 if (frame_bits <= 6144 * avctx->channels - 3) {
621 s->psy.bitres.bits = frame_bits / avctx->channels;
625 s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
629 put_bits(&s->pb, 3, TYPE_END);
630 flush_put_bits(&s->pb);
631 avctx->frame_bits = put_bits_count(&s->pb);
633 // rate control stuff
634 if (!(avctx->flags & CODEC_FLAG_QSCALE)) {
635 float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
637 s->lambda = FFMIN(s->lambda, 65536.f);
642 memcpy(s->samples, s->samples + 1024 * avctx->channels,
643 1024 * avctx->channels * sizeof(s->samples[0]));
644 return put_bits_count(&s->pb)>>3;
647 static av_cold int aac_encode_end(AVCodecContext *avctx)
649 AACEncContext *s = avctx->priv_data;
651 ff_mdct_end(&s->mdct1024);
652 ff_mdct_end(&s->mdct128);
654 ff_psy_preprocess_end(s->psypp);
655 av_freep(&s->samples);
660 #define AACENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
661 static const AVOption aacenc_options[] = {
662 {"stereo_mode", "Stereo coding method", offsetof(AACEncContext, options.stereo_mode), FF_OPT_TYPE_INT, {.dbl = 0}, -1, 1, AACENC_FLAGS, "stereo_mode"},
663 {"auto", "Selected by the Encoder", 0, FF_OPT_TYPE_CONST, {.dbl = -1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
664 {"ms_off", "Disable Mid/Side coding", 0, FF_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
665 {"ms_force", "Force Mid/Side for the whole frame if possible", 0, FF_OPT_TYPE_CONST, {.dbl = 1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
669 static const AVClass aacenc_class = {
671 av_default_item_name,
673 LIBAVUTIL_VERSION_INT,
676 AVCodec ff_aac_encoder = {
680 sizeof(AACEncContext),
684 .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL,
685 .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE},
686 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
687 .priv_class = &aacenc_class,