1 /*
2 * AAC encoder wrapper
3 * Copyright (c) 2012 Martin Storsjo
4 *
5 * This file is part of FFmpeg.
6 *
7 * Permission to use, copy, modify, and/or distribute this software for any
8 * purpose with or without fee is hereby granted, provided that the above
9 * copyright notice and this permission notice appear in all copies.
10 *
11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18 */
19
20 #include <fdk-aac/aacenc_lib.h>
21
22 #include "libavutil/channel_layout.h"
23 #include "libavutil/common.h"
24 #include "libavutil/opt.h"
25 #include "avcodec.h"
26 #include "audio_frame_queue.h"
27 #include "codec_internal.h"
28 #include "encode.h"
29 #include "profiles.h"
30
31 #ifdef AACENCODER_LIB_VL0
32 #define FDKENC_VER_AT_LEAST(vl0, vl1) \
33 ((AACENCODER_LIB_VL0 > vl0) || \
34 (AACENCODER_LIB_VL0 == vl0 && AACENCODER_LIB_VL1 >= vl1))
35 #else
36 #define FDKENC_VER_AT_LEAST(vl0, vl1) 0
37 #endif
38
39 typedef struct AACContext {
40 const AVClass *class;
41 HANDLE_AACENCODER handle;
42 int afterburner;
43 int eld_sbr;
44 int eld_v2;
45 int signaling;
46 int latm;
47 int header_period;
48 int vbr;
49
50 AudioFrameQueue afq;
51 } AACContext;
52
53 static const AVOption aac_enc_options[] = {
54 { "afterburner", "Afterburner (improved quality)", offsetof(AACContext, afterburner), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
55 { "eld_sbr", "Enable SBR for ELD (for SBR in other configurations, use the -profile parameter)", offsetof(AACContext, eld_sbr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
56 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
57 { "eld_v2", "Enable ELDv2 (LD-MPS extension for ELD stereo signals)", offsetof(AACContext, eld_v2), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
58 #endif
59 { "signaling", "SBR/PS signaling style", offsetof(AACContext, signaling), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 2, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
60 { "default", "Choose signaling implicitly (explicit hierarchical by default, implicit if global header is disabled)", 0, AV_OPT_TYPE_CONST, { .i64 = -1 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
61 { "implicit", "Implicit backwards compatible signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
62 { "explicit_sbr", "Explicit SBR, implicit PS signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
63 { "explicit_hierarchical", "Explicit hierarchical signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 2 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
64 { "latm", "Output LATM/LOAS encapsulated data", offsetof(AACContext, latm), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
65 { "header_period", "StreamMuxConfig and PCE repetition period (in frames)", offsetof(AACContext, header_period), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 0xffff, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
66 { "vbr", "VBR mode (1-5)", offsetof(AACContext, vbr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 5, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
67 FF_AAC_PROFILE_OPTS
68 { NULL }
69 };
70
71 static const AVClass aac_enc_class = {
72 .class_name = "libfdk_aac",
73 .item_name = av_default_item_name,
74 .option = aac_enc_options,
75 .version = LIBAVUTIL_VERSION_INT,
76 };
77
aac_get_error(AACENC_ERROR err)78 static const char *aac_get_error(AACENC_ERROR err)
79 {
80 switch (err) {
81 case AACENC_OK:
82 return "No error";
83 case AACENC_INVALID_HANDLE:
84 return "Invalid handle";
85 case AACENC_MEMORY_ERROR:
86 return "Memory allocation error";
87 case AACENC_UNSUPPORTED_PARAMETER:
88 return "Unsupported parameter";
89 case AACENC_INVALID_CONFIG:
90 return "Invalid config";
91 case AACENC_INIT_ERROR:
92 return "Initialization error";
93 case AACENC_INIT_AAC_ERROR:
94 return "AAC library initialization error";
95 case AACENC_INIT_SBR_ERROR:
96 return "SBR library initialization error";
97 case AACENC_INIT_TP_ERROR:
98 return "Transport library initialization error";
99 case AACENC_INIT_META_ERROR:
100 return "Metadata library initialization error";
101 case AACENC_ENCODE_ERROR:
102 return "Encoding error";
103 case AACENC_ENCODE_EOF:
104 return "End of file";
105 default:
106 return "Unknown error";
107 }
108 }
109
aac_encode_close(AVCodecContext * avctx)110 static int aac_encode_close(AVCodecContext *avctx)
111 {
112 AACContext *s = avctx->priv_data;
113
114 if (s->handle)
115 aacEncClose(&s->handle);
116 ff_af_queue_close(&s->afq);
117
118 return 0;
119 }
120
aac_encode_init(AVCodecContext * avctx)121 static av_cold int aac_encode_init(AVCodecContext *avctx)
122 {
123 AACContext *s = avctx->priv_data;
124 int ret = AVERROR(EINVAL);
125 AACENC_InfoStruct info = { 0 };
126 CHANNEL_MODE mode;
127 AACENC_ERROR err;
128 int aot = FF_PROFILE_AAC_LOW + 1;
129 int sce = 0, cpe = 0;
130
131 if ((err = aacEncOpen(&s->handle, 0, avctx->ch_layout.nb_channels)) != AACENC_OK) {
132 av_log(avctx, AV_LOG_ERROR, "Unable to open the encoder: %s\n",
133 aac_get_error(err));
134 goto error;
135 }
136
137 if (avctx->profile != FF_PROFILE_UNKNOWN)
138 aot = avctx->profile + 1;
139
140 if ((err = aacEncoder_SetParam(s->handle, AACENC_AOT, aot)) != AACENC_OK) {
141 av_log(avctx, AV_LOG_ERROR, "Unable to set the AOT %d: %s\n",
142 aot, aac_get_error(err));
143 goto error;
144 }
145
146 if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_sbr) {
147 if ((err = aacEncoder_SetParam(s->handle, AACENC_SBR_MODE,
148 1)) != AACENC_OK) {
149 av_log(avctx, AV_LOG_ERROR, "Unable to enable SBR for ELD: %s\n",
150 aac_get_error(err));
151 goto error;
152 }
153 }
154
155 if ((err = aacEncoder_SetParam(s->handle, AACENC_SAMPLERATE,
156 avctx->sample_rate)) != AACENC_OK) {
157 av_log(avctx, AV_LOG_ERROR, "Unable to set the sample rate %d: %s\n",
158 avctx->sample_rate, aac_get_error(err));
159 goto error;
160 }
161
162 switch (avctx->ch_layout.nb_channels) {
163 case 1: mode = MODE_1; sce = 1; cpe = 0; break;
164 case 2:
165 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
166 // (profile + 1) to map from profile range to AOT range
167 if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_v2) {
168 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
169 128)) != AACENC_OK) {
170 av_log(avctx, AV_LOG_ERROR, "Unable to enable ELDv2: %s\n",
171 aac_get_error(err));
172 goto error;
173 } else {
174 mode = MODE_212;
175 sce = 1;
176 cpe = 0;
177 }
178 } else
179 #endif
180 {
181 mode = MODE_2;
182 sce = 0;
183 cpe = 1;
184 }
185 break;
186 case 3: mode = MODE_1_2; sce = 1; cpe = 1; break;
187 case 4: mode = MODE_1_2_1; sce = 2; cpe = 1; break;
188 case 5: mode = MODE_1_2_2; sce = 1; cpe = 2; break;
189 case 6: mode = MODE_1_2_2_1; sce = 2; cpe = 2; break;
190 /* The version macro is introduced the same time as the 7.1 support, so this
191 should suffice. */
192 #if FDKENC_VER_AT_LEAST(3, 4) // 3.4.12
193 case 8:
194 sce = 2;
195 cpe = 3;
196 if (!av_channel_layout_compare(&avctx->ch_layout, &(AVChannelLayout)AV_CHANNEL_LAYOUT_7POINT1)) {
197 mode = MODE_7_1_REAR_SURROUND;
198 } else {
199 // MODE_1_2_2_2_1 and MODE_7_1_FRONT_CENTER use the same channel layout
200 mode = MODE_7_1_FRONT_CENTER;
201 }
202 break;
203 #endif
204 default:
205 av_log(avctx, AV_LOG_ERROR,
206 "Unsupported number of channels %d\n", avctx->ch_layout.nb_channels);
207 goto error;
208 }
209
210 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
211 mode)) != AACENC_OK) {
212 av_log(avctx, AV_LOG_ERROR,
213 "Unable to set channel mode %d: %s\n", mode, aac_get_error(err));
214 goto error;
215 }
216
217 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELORDER,
218 1)) != AACENC_OK) {
219 av_log(avctx, AV_LOG_ERROR,
220 "Unable to set wav channel order %d: %s\n",
221 mode, aac_get_error(err));
222 goto error;
223 }
224
225 if (avctx->flags & AV_CODEC_FLAG_QSCALE || s->vbr) {
226 int mode = s->vbr ? s->vbr : avctx->global_quality;
227 if (mode < 1 || mode > 5) {
228 av_log(avctx, AV_LOG_WARNING,
229 "VBR quality %d out of range, should be 1-5\n", mode);
230 mode = av_clip(mode, 1, 5);
231 }
232 av_log(avctx, AV_LOG_WARNING,
233 "Note, the VBR setting is unsupported and only works with "
234 "some parameter combinations\n");
235 if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATEMODE,
236 mode)) != AACENC_OK) {
237 av_log(avctx, AV_LOG_ERROR, "Unable to set the VBR bitrate mode %d: %s\n",
238 mode, aac_get_error(err));
239 goto error;
240 }
241 } else {
242 if (avctx->bit_rate <= 0) {
243 if (avctx->profile == FF_PROFILE_AAC_HE_V2) {
244 sce = 1;
245 cpe = 0;
246 }
247 avctx->bit_rate = (96*sce + 128*cpe) * avctx->sample_rate / 44;
248 if (avctx->profile == FF_PROFILE_AAC_HE ||
249 avctx->profile == FF_PROFILE_AAC_HE_V2 ||
250 avctx->profile == FF_PROFILE_MPEG2_AAC_HE ||
251 s->eld_sbr)
252 avctx->bit_rate /= 2;
253 }
254 if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATE,
255 avctx->bit_rate)) != AACENC_OK) {
256 av_log(avctx, AV_LOG_ERROR, "Unable to set the bitrate %"PRId64": %s\n",
257 avctx->bit_rate, aac_get_error(err));
258 goto error;
259 }
260 }
261
262 /* Choose bitstream format - if global header is requested, use
263 * raw access units, otherwise use ADTS. */
264 if ((err = aacEncoder_SetParam(s->handle, AACENC_TRANSMUX,
265 avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? TT_MP4_RAW :
266 s->latm ? TT_MP4_LOAS : TT_MP4_ADTS)) != AACENC_OK) {
267 av_log(avctx, AV_LOG_ERROR, "Unable to set the transmux format: %s\n",
268 aac_get_error(err));
269 goto error;
270 }
271
272 if (s->latm && s->header_period) {
273 if ((err = aacEncoder_SetParam(s->handle, AACENC_HEADER_PERIOD,
274 s->header_period)) != AACENC_OK) {
275 av_log(avctx, AV_LOG_ERROR, "Unable to set header period: %s\n",
276 aac_get_error(err));
277 goto error;
278 }
279 }
280
281 /* If no signaling mode is chosen, use explicit hierarchical signaling
282 * if using mp4 mode (raw access units, with global header) and
283 * implicit signaling if using ADTS. */
284 if (s->signaling < 0)
285 s->signaling = avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? 2 : 0;
286
287 if ((err = aacEncoder_SetParam(s->handle, AACENC_SIGNALING_MODE,
288 s->signaling)) != AACENC_OK) {
289 av_log(avctx, AV_LOG_ERROR, "Unable to set signaling mode %d: %s\n",
290 s->signaling, aac_get_error(err));
291 goto error;
292 }
293
294 if ((err = aacEncoder_SetParam(s->handle, AACENC_AFTERBURNER,
295 s->afterburner)) != AACENC_OK) {
296 av_log(avctx, AV_LOG_ERROR, "Unable to set afterburner to %d: %s\n",
297 s->afterburner, aac_get_error(err));
298 goto error;
299 }
300
301 if (avctx->cutoff > 0) {
302 if (avctx->cutoff < (avctx->sample_rate + 255) >> 8 || avctx->cutoff > 20000) {
303 av_log(avctx, AV_LOG_ERROR, "cutoff valid range is %d-20000\n",
304 (avctx->sample_rate + 255) >> 8);
305 goto error;
306 }
307 if ((err = aacEncoder_SetParam(s->handle, AACENC_BANDWIDTH,
308 avctx->cutoff)) != AACENC_OK) {
309 av_log(avctx, AV_LOG_ERROR, "Unable to set the encoder bandwidth to %d: %s\n",
310 avctx->cutoff, aac_get_error(err));
311 goto error;
312 }
313 }
314
315 if ((err = aacEncEncode(s->handle, NULL, NULL, NULL, NULL)) != AACENC_OK) {
316 av_log(avctx, AV_LOG_ERROR, "Unable to initialize the encoder: %s\n",
317 aac_get_error(err));
318 return AVERROR(EINVAL);
319 }
320
321 if ((err = aacEncInfo(s->handle, &info)) != AACENC_OK) {
322 av_log(avctx, AV_LOG_ERROR, "Unable to get encoder info: %s\n",
323 aac_get_error(err));
324 goto error;
325 }
326
327 avctx->frame_size = info.frameLength;
328 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
329 avctx->initial_padding = info.nDelay;
330 #else
331 avctx->initial_padding = info.encoderDelay;
332 #endif
333 ff_af_queue_init(avctx, &s->afq);
334
335 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
336 avctx->extradata_size = info.confSize;
337 avctx->extradata = av_mallocz(avctx->extradata_size +
338 AV_INPUT_BUFFER_PADDING_SIZE);
339 if (!avctx->extradata) {
340 ret = AVERROR(ENOMEM);
341 goto error;
342 }
343
344 memcpy(avctx->extradata, info.confBuf, info.confSize);
345 }
346 return 0;
347 error:
348 aac_encode_close(avctx);
349 return ret;
350 }
351
aac_encode_frame(AVCodecContext * avctx,AVPacket * avpkt,const AVFrame * frame,int * got_packet_ptr)352 static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
353 const AVFrame *frame, int *got_packet_ptr)
354 {
355 AACContext *s = avctx->priv_data;
356 AACENC_BufDesc in_buf = { 0 }, out_buf = { 0 };
357 AACENC_InArgs in_args = { 0 };
358 AACENC_OutArgs out_args = { 0 };
359 int in_buffer_identifier = IN_AUDIO_DATA;
360 int in_buffer_size, in_buffer_element_size;
361 int out_buffer_identifier = OUT_BITSTREAM_DATA;
362 int out_buffer_size, out_buffer_element_size;
363 void *in_ptr, *out_ptr;
364 int ret;
365 uint8_t dummy_buf[1];
366 AACENC_ERROR err;
367
368 /* handle end-of-stream small frame and flushing */
369 if (!frame) {
370 /* Must be a non-null pointer, even if it's a dummy. We could use
371 * the address of anything else on the stack as well. */
372 in_ptr = dummy_buf;
373 in_buffer_size = 0;
374
375 in_args.numInSamples = -1;
376 } else {
377 in_ptr = frame->data[0];
378 in_buffer_size = 2 * avctx->ch_layout.nb_channels * frame->nb_samples;
379
380 in_args.numInSamples = avctx->ch_layout.nb_channels * frame->nb_samples;
381
382 /* add current frame to the queue */
383 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
384 return ret;
385 }
386
387 in_buffer_element_size = 2;
388 in_buf.numBufs = 1;
389 in_buf.bufs = &in_ptr;
390 in_buf.bufferIdentifiers = &in_buffer_identifier;
391 in_buf.bufSizes = &in_buffer_size;
392 in_buf.bufElSizes = &in_buffer_element_size;
393
394 /* The maximum packet size is 6144 bits aka 768 bytes per channel. */
395 ret = ff_alloc_packet(avctx, avpkt, FFMAX(8192, 768 * avctx->ch_layout.nb_channels));
396 if (ret < 0)
397 return ret;
398
399 out_ptr = avpkt->data;
400 out_buffer_size = avpkt->size;
401 out_buffer_element_size = 1;
402 out_buf.numBufs = 1;
403 out_buf.bufs = &out_ptr;
404 out_buf.bufferIdentifiers = &out_buffer_identifier;
405 out_buf.bufSizes = &out_buffer_size;
406 out_buf.bufElSizes = &out_buffer_element_size;
407
408 if ((err = aacEncEncode(s->handle, &in_buf, &out_buf, &in_args,
409 &out_args)) != AACENC_OK) {
410 if (!frame && err == AACENC_ENCODE_EOF)
411 return 0;
412 av_log(avctx, AV_LOG_ERROR, "Unable to encode frame: %s\n",
413 aac_get_error(err));
414 return AVERROR(EINVAL);
415 }
416
417 if (!out_args.numOutBytes)
418 return 0;
419
420 /* Get the next frame pts & duration */
421 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
422 &avpkt->duration);
423
424 avpkt->size = out_args.numOutBytes;
425 *got_packet_ptr = 1;
426 return 0;
427 }
428
429 static const AVProfile profiles[] = {
430 { FF_PROFILE_AAC_LOW, "LC" },
431 { FF_PROFILE_AAC_HE, "HE-AAC" },
432 { FF_PROFILE_AAC_HE_V2, "HE-AACv2" },
433 { FF_PROFILE_AAC_LD, "LD" },
434 { FF_PROFILE_AAC_ELD, "ELD" },
435 { FF_PROFILE_UNKNOWN },
436 };
437
438 static const FFCodecDefault aac_encode_defaults[] = {
439 { "b", "0" },
440 { NULL }
441 };
442
443 #if FF_API_OLD_CHANNEL_LAYOUT
444 static const uint64_t aac_channel_layout[] = {
445 AV_CH_LAYOUT_MONO,
446 AV_CH_LAYOUT_STEREO,
447 AV_CH_LAYOUT_SURROUND,
448 AV_CH_LAYOUT_4POINT0,
449 AV_CH_LAYOUT_5POINT0_BACK,
450 AV_CH_LAYOUT_5POINT1_BACK,
451 #if FDKENC_VER_AT_LEAST(3, 4) // 3.4.12
452 AV_CH_LAYOUT_7POINT1_WIDE_BACK,
453 AV_CH_LAYOUT_7POINT1,
454 #endif
455 0,
456 };
457 #endif /* FF_API_OLD_CHANNEL_LAYOUT */
458
459 static const AVChannelLayout aac_ch_layouts[16] = {
460 AV_CHANNEL_LAYOUT_MONO,
461 AV_CHANNEL_LAYOUT_STEREO,
462 AV_CHANNEL_LAYOUT_SURROUND,
463 AV_CHANNEL_LAYOUT_4POINT0,
464 AV_CHANNEL_LAYOUT_5POINT0_BACK,
465 AV_CHANNEL_LAYOUT_5POINT1_BACK,
466 #ifdef AACENCODER_LIB_VL0
467 AV_CHANNEL_LAYOUT_7POINT1_WIDE_BACK,
468 AV_CHANNEL_LAYOUT_7POINT1,
469 #endif
470 { 0 },
471 };
472
473 static const int aac_sample_rates[] = {
474 96000, 88200, 64000, 48000, 44100, 32000,
475 24000, 22050, 16000, 12000, 11025, 8000, 0
476 };
477
478 const FFCodec ff_libfdk_aac_encoder = {
479 .p.name = "libfdk_aac",
480 .p.long_name = NULL_IF_CONFIG_SMALL("Fraunhofer FDK AAC"),
481 .p.type = AVMEDIA_TYPE_AUDIO,
482 .p.id = AV_CODEC_ID_AAC,
483 .priv_data_size = sizeof(AACContext),
484 .init = aac_encode_init,
485 FF_CODEC_ENCODE_CB(aac_encode_frame),
486 .close = aac_encode_close,
487 .p.capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY,
488 .p.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
489 AV_SAMPLE_FMT_NONE },
490 .p.priv_class = &aac_enc_class,
491 .defaults = aac_encode_defaults,
492 .p.profiles = profiles,
493 .p.supported_samplerates = aac_sample_rates,
494 .p.wrapper_name = "libfdk",
495 #if FF_API_OLD_CHANNEL_LAYOUT
496 .p.channel_layouts = aac_channel_layout,
497 #endif
498 .p.ch_layouts = aac_ch_layouts,
499 };
500