3 * Copyright (C) 2008 Konstantin Shishkov
5 * This file is part of FFmpeg.
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.
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.
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
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
138 static const uint8_t channel_maps[][AAC_MAX_CHANNELS] = {
144 { 2, 0, 1, 4, 5, 3 },
148 * Make AAC audio config object.
149 * @see 1.6.2.1 "Syntax - AudioSpecificConfig"
151 static void put_audio_specific_config(AVCodecContext *avctx)
154 AACEncContext *s = avctx->priv_data;
156 init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8);
157 put_bits(&pb, 5, 2); //object type - AAC-LC
158 put_bits(&pb, 4, s->samplerate_index); //sample rate index
159 put_bits(&pb, 4, avctx->channels);
161 put_bits(&pb, 1, 0); //frame length - 1024 samples
162 put_bits(&pb, 1, 0); //does not depend on core coder
163 put_bits(&pb, 1, 0); //is not extension
165 //Explicitly Mark SBR absent
166 put_bits(&pb, 11, 0x2b7); //sync extension
167 put_bits(&pb, 5, AOT_SBR);
172 static av_cold int aac_encode_init(AVCodecContext *avctx)
174 AACEncContext *s = avctx->priv_data;
176 const uint8_t *sizes[2];
179 avctx->frame_size = 1024;
181 for (i = 0; i < 16; i++)
182 if (avctx->sample_rate == ff_mpeg4audio_sample_rates[i])
185 av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate);
188 if (avctx->channels > AAC_MAX_CHANNELS) {
189 av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels);
192 if (avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW) {
193 av_log(avctx, AV_LOG_ERROR, "Unsupported profile %d\n", avctx->profile);
196 if (1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * avctx->channels) {
197 av_log(avctx, AV_LOG_ERROR, "Too many bits per frame requested\n");
200 s->samplerate_index = i;
202 dsputil_init(&s->dsp, avctx);
203 ff_mdct_init(&s->mdct1024, 11, 0, 1.0);
204 ff_mdct_init(&s->mdct128, 8, 0, 1.0);
206 ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
207 ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
208 ff_init_ff_sine_windows(10);
209 ff_init_ff_sine_windows(7);
211 s->chan_map = aac_chan_configs[avctx->channels-1];
212 s->samples = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0]));
213 s->cpe = av_mallocz(sizeof(ChannelElement) * s->chan_map[0]);
214 avctx->extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
215 avctx->extradata_size = 5;
216 put_audio_specific_config(avctx);
218 sizes[0] = swb_size_1024[i];
219 sizes[1] = swb_size_128[i];
220 lengths[0] = ff_aac_num_swb_1024[i];
221 lengths[1] = ff_aac_num_swb_128[i];
222 ff_psy_init(&s->psy, avctx, 2, sizes, lengths, s->chan_map[0], &s->chan_map[1]);
223 s->psypp = ff_psy_preprocess_init(avctx);
224 s->coder = &ff_aac_coders[2];
226 s->lambda = avctx->global_quality ? avctx->global_quality : 120;
233 static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s,
234 SingleChannelElement *sce, short *audio)
237 const int chans = avctx->channels;
238 const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
239 const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
240 const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
241 float *output = sce->ret;
243 if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
244 memcpy(output, sce->saved, sizeof(float)*1024);
245 if (sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE) {
246 memset(output, 0, sizeof(output[0]) * 448);
247 for (i = 448; i < 576; i++)
248 output[i] = sce->saved[i] * pwindow[i - 448];
249 for (i = 576; i < 704; i++)
250 output[i] = sce->saved[i];
252 if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) {
253 for (i = 0; i < 1024; i++) {
254 output[i+1024] = audio[i * chans] * lwindow[1024 - i - 1];
255 sce->saved[i] = audio[i * chans] * lwindow[i];
258 for (i = 0; i < 448; i++)
259 output[i+1024] = audio[i * chans];
261 output[i+1024] = audio[i * chans] * swindow[576 - i - 1];
262 memset(output+1024+576, 0, sizeof(output[0]) * 448);
263 for (i = 0; i < 1024; i++)
264 sce->saved[i] = audio[i * chans];
266 s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
268 for (k = 0; k < 1024; k += 128) {
269 for (i = 448 + k; i < 448 + k + 256; i++)
270 output[i - 448 - k] = (i < 1024)
272 : audio[(i-1024)*chans];
273 s->dsp.vector_fmul (output, output, k ? swindow : pwindow, 128);
274 s->dsp.vector_fmul_reverse(output+128, output+128, swindow, 128);
275 s->mdct128.mdct_calc(&s->mdct128, sce->coeffs + k, output);
277 for (i = 0; i < 1024; i++)
278 sce->saved[i] = audio[i * chans];
283 * Encode ics_info element.
284 * @see Table 4.6 (syntax of ics_info)
286 static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
290 put_bits(&s->pb, 1, 0); // ics_reserved bit
291 put_bits(&s->pb, 2, info->window_sequence[0]);
292 put_bits(&s->pb, 1, info->use_kb_window[0]);
293 if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
294 put_bits(&s->pb, 6, info->max_sfb);
295 put_bits(&s->pb, 1, 0); // no prediction
297 put_bits(&s->pb, 4, info->max_sfb);
298 for (w = 1; w < 8; w++)
299 put_bits(&s->pb, 1, !info->group_len[w]);
305 * @see 4.6.8.1 "Joint Coding - M/S Stereo"
307 static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
311 put_bits(pb, 2, cpe->ms_mode);
312 if (cpe->ms_mode == 1)
313 for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w])
314 for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
315 put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
319 * Produce integer coefficients from scalefactors provided by the model.
321 static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans)
324 int start, maxsfb, cmaxsfb;
326 for (ch = 0; ch < chans; ch++) {
327 IndividualChannelStream *ics = &cpe->ch[ch].ics;
330 cpe->ch[ch].pulse.num_pulse = 0;
331 for (w = 0; w < ics->num_windows*16; w += 16) {
332 for (g = 0; g < ics->num_swb; g++) {
334 if (cpe->common_window && !ch && cpe->ms_mask[w + g]) {
335 for (i = 0; i < ics->swb_sizes[g]; i++) {
336 cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
337 cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
340 start += ics->swb_sizes[g];
342 for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--)
344 maxsfb = FFMAX(maxsfb, cmaxsfb);
346 ics->max_sfb = maxsfb;
348 //adjust zero bands for window groups
349 for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
350 for (g = 0; g < ics->max_sfb; g++) {
352 for (w2 = w; w2 < w + ics->group_len[w]; w2++) {
353 if (!cpe->ch[ch].zeroes[w2*16 + g]) {
358 cpe->ch[ch].zeroes[w*16 + g] = i;
363 if (chans > 1 && cpe->common_window) {
364 IndividualChannelStream *ics0 = &cpe->ch[0].ics;
365 IndividualChannelStream *ics1 = &cpe->ch[1].ics;
367 ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
368 ics1->max_sfb = ics0->max_sfb;
369 for (w = 0; w < ics0->num_windows*16; w += 16)
370 for (i = 0; i < ics0->max_sfb; i++)
371 if (cpe->ms_mask[w+i])
373 if (msc == 0 || ics0->max_sfb == 0)
376 cpe->ms_mode = msc < ics0->max_sfb * ics0->num_windows ? 1 : 2;
381 * Encode scalefactor band coding type.
383 static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
387 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
388 s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
392 * Encode scalefactors.
394 static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s,
395 SingleChannelElement *sce)
397 int off = sce->sf_idx[0], diff;
400 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
401 for (i = 0; i < sce->ics.max_sfb; i++) {
402 if (!sce->zeroes[w*16 + i]) {
403 diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO;
404 if (diff < 0 || diff > 120)
405 av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n");
406 off = sce->sf_idx[w*16 + i];
407 put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
416 static void encode_pulses(AACEncContext *s, Pulse *pulse)
420 put_bits(&s->pb, 1, !!pulse->num_pulse);
421 if (!pulse->num_pulse)
424 put_bits(&s->pb, 2, pulse->num_pulse - 1);
425 put_bits(&s->pb, 6, pulse->start);
426 for (i = 0; i < pulse->num_pulse; i++) {
427 put_bits(&s->pb, 5, pulse->pos[i]);
428 put_bits(&s->pb, 4, pulse->amp[i]);
433 * Encode spectral coefficients processed by psychoacoustic model.
435 static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
439 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
441 for (i = 0; i < sce->ics.max_sfb; i++) {
442 if (sce->zeroes[w*16 + i]) {
443 start += sce->ics.swb_sizes[i];
446 for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++)
447 s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128,
448 sce->ics.swb_sizes[i],
449 sce->sf_idx[w*16 + i],
450 sce->band_type[w*16 + i],
452 start += sce->ics.swb_sizes[i];
458 * Encode one channel of audio data.
460 static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
461 SingleChannelElement *sce,
464 put_bits(&s->pb, 8, sce->sf_idx[0]);
466 put_ics_info(s, &sce->ics);
467 encode_band_info(s, sce);
468 encode_scale_factors(avctx, s, sce);
469 encode_pulses(s, &sce->pulse);
470 put_bits(&s->pb, 1, 0); //tns
471 put_bits(&s->pb, 1, 0); //ssr
472 encode_spectral_coeffs(s, sce);
477 * Write some auxiliary information about the created AAC file.
479 static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s,
482 int i, namelen, padbits;
484 namelen = strlen(name) + 2;
485 put_bits(&s->pb, 3, TYPE_FIL);
486 put_bits(&s->pb, 4, FFMIN(namelen, 15));
488 put_bits(&s->pb, 8, namelen - 16);
489 put_bits(&s->pb, 4, 0); //extension type - filler
490 padbits = 8 - (put_bits_count(&s->pb) & 7);
491 align_put_bits(&s->pb);
492 for (i = 0; i < namelen - 2; i++)
493 put_bits(&s->pb, 8, name[i]);
494 put_bits(&s->pb, 12 - padbits, 0);
497 static int aac_encode_frame(AVCodecContext *avctx,
498 uint8_t *frame, int buf_size, void *data)
500 AACEncContext *s = avctx->priv_data;
501 int16_t *samples = s->samples, *samples2, *la;
503 int i, ch, w, g, chans, tag, start_ch;
504 int chan_el_counter[4];
505 FFPsyWindowInfo windows[AAC_MAX_CHANNELS];
511 if (avctx->channels <= 2) {
512 memcpy(s->samples + 1024 * avctx->channels, data,
513 1024 * avctx->channels * sizeof(s->samples[0]));
515 for (i = 0; i < 1024; i++)
516 for (ch = 0; ch < avctx->channels; ch++)
517 s->samples[(i + 1024) * avctx->channels + ch] =
518 ((int16_t*)data)[i * avctx->channels +
519 channel_maps[avctx->channels-1][ch]];
523 samples2 = s->samples + 1024 * avctx->channels;
524 for (i = 0; i < s->chan_map[0]; i++) {
525 tag = s->chan_map[i+1];
526 chans = tag == TYPE_CPE ? 2 : 1;
527 ff_psy_preprocess(s->psypp,
528 (uint16_t*)data + channel_maps[avctx->channels-1][start_ch],
529 samples2 + start_ch, start_ch, chans);
534 if (!avctx->frame_number) {
535 memcpy(s->samples, s->samples + 1024 * avctx->channels,
536 1024 * avctx->channels * sizeof(s->samples[0]));
541 for (i = 0; i < s->chan_map[0]; i++) {
542 FFPsyWindowInfo* wi = windows + start_ch;
543 tag = s->chan_map[i+1];
544 chans = tag == TYPE_CPE ? 2 : 1;
546 for (ch = 0; ch < chans; ch++) {
547 IndividualChannelStream *ics = &cpe->ch[ch].ics;
548 int cur_channel = start_ch + ch;
549 samples2 = samples + cur_channel;
550 la = samples2 + (448+64) * avctx->channels;
553 if (tag == TYPE_LFE) {
554 wi[ch].window_type[0] = ONLY_LONG_SEQUENCE;
555 wi[ch].window_shape = 0;
556 wi[ch].num_windows = 1;
557 wi[ch].grouping[0] = 1;
559 wi[ch] = s->psy.model->window(&s->psy, samples2, la, cur_channel,
560 ics->window_sequence[0]);
562 ics->window_sequence[1] = ics->window_sequence[0];
563 ics->window_sequence[0] = wi[ch].window_type[0];
564 ics->use_kb_window[1] = ics->use_kb_window[0];
565 ics->use_kb_window[0] = wi[ch].window_shape;
566 ics->num_windows = wi[ch].num_windows;
567 ics->swb_sizes = s->psy.bands [ics->num_windows == 8];
568 ics->num_swb = tag == TYPE_LFE ? 12 : s->psy.num_bands[ics->num_windows == 8];
569 for (w = 0; w < ics->num_windows; w++)
570 ics->group_len[w] = wi[ch].grouping[w];
572 apply_window_and_mdct(avctx, s, &cpe->ch[ch], samples2);
578 init_put_bits(&s->pb, frame, buf_size*8);
579 if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT))
580 put_bitstream_info(avctx, s, LIBAVCODEC_IDENT);
582 memset(chan_el_counter, 0, sizeof(chan_el_counter));
583 for (i = 0; i < s->chan_map[0]; i++) {
584 FFPsyWindowInfo* wi = windows + start_ch;
585 const float *coeffs[2];
586 tag = s->chan_map[i+1];
587 chans = tag == TYPE_CPE ? 2 : 1;
589 put_bits(&s->pb, 3, tag);
590 put_bits(&s->pb, 4, chan_el_counter[tag]++);
591 for (ch = 0; ch < chans; ch++)
592 coeffs[ch] = cpe->ch[ch].coeffs;
593 s->psy.model->analyze(&s->psy, start_ch, coeffs, wi);
594 for (ch = 0; ch < chans; ch++) {
595 s->cur_channel = start_ch * 2 + ch;
596 s->coder->search_for_quantizers(avctx, s, &cpe->ch[ch], s->lambda);
598 cpe->common_window = 0;
600 && wi[0].window_type[0] == wi[1].window_type[0]
601 && wi[0].window_shape == wi[1].window_shape) {
603 cpe->common_window = 1;
604 for (w = 0; w < wi[0].num_windows; w++) {
605 if (wi[0].grouping[w] != wi[1].grouping[w]) {
606 cpe->common_window = 0;
611 s->cur_channel = start_ch * 2;
612 if (s->options.stereo_mode && cpe->common_window) {
613 if (s->options.stereo_mode > 0) {
614 IndividualChannelStream *ics = &cpe->ch[0].ics;
615 for (w = 0; w < ics->num_windows; w += ics->group_len[w])
616 for (g = 0; g < ics->num_swb; g++)
617 cpe->ms_mask[w*16+g] = 1;
618 } else if (s->coder->search_for_ms) {
619 s->coder->search_for_ms(s, cpe, s->lambda);
622 adjust_frame_information(s, cpe, chans);
624 put_bits(&s->pb, 1, cpe->common_window);
625 if (cpe->common_window) {
626 put_ics_info(s, &cpe->ch[0].ics);
627 encode_ms_info(&s->pb, cpe);
630 for (ch = 0; ch < chans; ch++) {
631 s->cur_channel = start_ch + ch;
632 encode_individual_channel(avctx, s, &cpe->ch[ch], cpe->common_window);
637 frame_bits = put_bits_count(&s->pb);
638 if (frame_bits <= 6144 * avctx->channels - 3) {
639 s->psy.bitres.bits = frame_bits / avctx->channels;
643 s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
647 put_bits(&s->pb, 3, TYPE_END);
648 flush_put_bits(&s->pb);
649 avctx->frame_bits = put_bits_count(&s->pb);
651 // rate control stuff
652 if (!(avctx->flags & CODEC_FLAG_QSCALE)) {
653 float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
655 s->lambda = FFMIN(s->lambda, 65536.f);
660 memcpy(s->samples, s->samples + 1024 * avctx->channels,
661 1024 * avctx->channels * sizeof(s->samples[0]));
662 return put_bits_count(&s->pb)>>3;
665 static av_cold int aac_encode_end(AVCodecContext *avctx)
667 AACEncContext *s = avctx->priv_data;
669 ff_mdct_end(&s->mdct1024);
670 ff_mdct_end(&s->mdct128);
672 ff_psy_preprocess_end(s->psypp);
673 av_freep(&s->samples);
678 #define AACENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
679 static const AVOption aacenc_options[] = {
680 {"stereo_mode", "Stereo coding method", offsetof(AACEncContext, options.stereo_mode), FF_OPT_TYPE_INT, {.dbl = 0}, -1, 1, AACENC_FLAGS, "stereo_mode"},
681 {"auto", "Selected by the Encoder", 0, FF_OPT_TYPE_CONST, {.dbl = -1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
682 {"ms_off", "Disable Mid/Side coding", 0, FF_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
683 {"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"},
687 static const AVClass aacenc_class = {
689 av_default_item_name,
691 LIBAVUTIL_VERSION_INT,
694 AVCodec ff_aac_encoder = {
698 sizeof(AACEncContext),
702 .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL,
703 .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE},
704 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
705 .priv_class = &aacenc_class,