3 * Copyright (c) 2012 Martin Storsjo
5 * This file is part of FFmpeg.
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.
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.
20 #include <fdk-aac/aacenc_lib.h>
22 #include "libavutil/channel_layout.h"
23 #include "libavutil/common.h"
24 #include "libavutil/opt.h"
26 #include "audio_frame_queue.h"
30 #ifdef AACENCODER_LIB_VL0
31 #define FDKENC_VER_AT_LEAST(vl0, vl1) \
32 ((AACENCODER_LIB_VL0 > vl0) || \
33 (AACENCODER_LIB_VL0 == vl0 && AACENCODER_LIB_VL1 >= vl1))
35 #define FDKENC_VER_AT_LEAST(vl0, vl1) 0
38 typedef struct AACContext {
40 HANDLE_AACENCODER handle;
52 static const AVOption aac_enc_options[] = {
53 { "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 },
54 { "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 },
55 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
56 { "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 { "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" },
59 { "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" },
60 { "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" },
61 { "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" },
62 { "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" },
63 { "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 },
64 { "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 },
65 { "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 },
70 static const AVClass aac_enc_class = {
71 .class_name = "libfdk_aac",
72 .item_name = av_default_item_name,
73 .option = aac_enc_options,
74 .version = LIBAVUTIL_VERSION_INT,
77 static const char *aac_get_error(AACENC_ERROR err)
82 case AACENC_INVALID_HANDLE:
83 return "Invalid handle";
84 case AACENC_MEMORY_ERROR:
85 return "Memory allocation error";
86 case AACENC_UNSUPPORTED_PARAMETER:
87 return "Unsupported parameter";
88 case AACENC_INVALID_CONFIG:
89 return "Invalid config";
90 case AACENC_INIT_ERROR:
91 return "Initialization error";
92 case AACENC_INIT_AAC_ERROR:
93 return "AAC library initialization error";
94 case AACENC_INIT_SBR_ERROR:
95 return "SBR library initialization error";
96 case AACENC_INIT_TP_ERROR:
97 return "Transport library initialization error";
98 case AACENC_INIT_META_ERROR:
99 return "Metadata library initialization error";
100 case AACENC_ENCODE_ERROR:
101 return "Encoding error";
102 case AACENC_ENCODE_EOF:
103 return "End of file";
105 return "Unknown error";
109 static int aac_encode_close(AVCodecContext *avctx)
111 AACContext *s = avctx->priv_data;
114 aacEncClose(&s->handle);
115 ff_af_queue_close(&s->afq);
120 static av_cold int aac_encode_init(AVCodecContext *avctx)
122 AACContext *s = avctx->priv_data;
123 int ret = AVERROR(EINVAL);
124 AACENC_InfoStruct info = { 0 };
127 int aot = FF_PROFILE_AAC_LOW + 1;
128 int sce = 0, cpe = 0;
130 if ((err = aacEncOpen(&s->handle, 0, avctx->channels)) != AACENC_OK) {
131 av_log(avctx, AV_LOG_ERROR, "Unable to open the encoder: %s\n",
136 if (avctx->profile != FF_PROFILE_UNKNOWN)
137 aot = avctx->profile + 1;
139 if ((err = aacEncoder_SetParam(s->handle, AACENC_AOT, aot)) != AACENC_OK) {
140 av_log(avctx, AV_LOG_ERROR, "Unable to set the AOT %d: %s\n",
141 aot, aac_get_error(err));
145 if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_sbr) {
146 if ((err = aacEncoder_SetParam(s->handle, AACENC_SBR_MODE,
148 av_log(avctx, AV_LOG_ERROR, "Unable to enable SBR for ELD: %s\n",
154 if ((err = aacEncoder_SetParam(s->handle, AACENC_SAMPLERATE,
155 avctx->sample_rate)) != AACENC_OK) {
156 av_log(avctx, AV_LOG_ERROR, "Unable to set the sample rate %d: %s\n",
157 avctx->sample_rate, aac_get_error(err));
161 switch (avctx->channels) {
162 case 1: mode = MODE_1; sce = 1; cpe = 0; break;
164 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
165 // (profile + 1) to map from profile range to AOT range
166 if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_v2) {
167 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
168 128)) != AACENC_OK) {
169 av_log(avctx, AV_LOG_ERROR, "Unable to enable ELDv2: %s\n",
185 case 3: mode = MODE_1_2; sce = 1; cpe = 1; break;
186 case 4: mode = MODE_1_2_1; sce = 2; cpe = 1; break;
187 case 5: mode = MODE_1_2_2; sce = 1; cpe = 2; break;
188 case 6: mode = MODE_1_2_2_1; sce = 2; cpe = 2; break;
189 /* The version macro is introduced the same time as the 7.1 support, so this
191 #if FDKENC_VER_AT_LEAST(3, 4) // 3.4.12
195 if (avctx->channel_layout == AV_CH_LAYOUT_7POINT1) {
196 mode = MODE_7_1_REAR_SURROUND;
198 // MODE_1_2_2_2_1 and MODE_7_1_FRONT_CENTER use the same channel layout
199 mode = MODE_7_1_FRONT_CENTER;
204 av_log(avctx, AV_LOG_ERROR,
205 "Unsupported number of channels %d\n", avctx->channels);
209 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
210 mode)) != AACENC_OK) {
211 av_log(avctx, AV_LOG_ERROR,
212 "Unable to set channel mode %d: %s\n", mode, aac_get_error(err));
216 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELORDER,
218 av_log(avctx, AV_LOG_ERROR,
219 "Unable to set wav channel order %d: %s\n",
220 mode, aac_get_error(err));
224 if (avctx->flags & AV_CODEC_FLAG_QSCALE || s->vbr) {
225 int mode = s->vbr ? s->vbr : avctx->global_quality;
226 if (mode < 1 || mode > 5) {
227 av_log(avctx, AV_LOG_WARNING,
228 "VBR quality %d out of range, should be 1-5\n", mode);
229 mode = av_clip(mode, 1, 5);
231 av_log(avctx, AV_LOG_WARNING,
232 "Note, the VBR setting is unsupported and only works with "
233 "some parameter combinations\n");
234 if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATEMODE,
235 mode)) != AACENC_OK) {
236 av_log(avctx, AV_LOG_ERROR, "Unable to set the VBR bitrate mode %d: %s\n",
237 mode, aac_get_error(err));
241 if (avctx->bit_rate <= 0) {
242 if (avctx->profile == FF_PROFILE_AAC_HE_V2) {
246 avctx->bit_rate = (96*sce + 128*cpe) * avctx->sample_rate / 44;
247 if (avctx->profile == FF_PROFILE_AAC_HE ||
248 avctx->profile == FF_PROFILE_AAC_HE_V2 ||
249 avctx->profile == FF_PROFILE_MPEG2_AAC_HE ||
251 avctx->bit_rate /= 2;
253 if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATE,
254 avctx->bit_rate)) != AACENC_OK) {
255 av_log(avctx, AV_LOG_ERROR, "Unable to set the bitrate %"PRId64": %s\n",
256 avctx->bit_rate, aac_get_error(err));
261 /* Choose bitstream format - if global header is requested, use
262 * raw access units, otherwise use ADTS. */
263 if ((err = aacEncoder_SetParam(s->handle, AACENC_TRANSMUX,
264 avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? TT_MP4_RAW :
265 s->latm ? TT_MP4_LOAS : TT_MP4_ADTS)) != AACENC_OK) {
266 av_log(avctx, AV_LOG_ERROR, "Unable to set the transmux format: %s\n",
271 if (s->latm && s->header_period) {
272 if ((err = aacEncoder_SetParam(s->handle, AACENC_HEADER_PERIOD,
273 s->header_period)) != AACENC_OK) {
274 av_log(avctx, AV_LOG_ERROR, "Unable to set header period: %s\n",
280 /* If no signaling mode is chosen, use explicit hierarchical signaling
281 * if using mp4 mode (raw access units, with global header) and
282 * implicit signaling if using ADTS. */
283 if (s->signaling < 0)
284 s->signaling = avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? 2 : 0;
286 if ((err = aacEncoder_SetParam(s->handle, AACENC_SIGNALING_MODE,
287 s->signaling)) != AACENC_OK) {
288 av_log(avctx, AV_LOG_ERROR, "Unable to set signaling mode %d: %s\n",
289 s->signaling, aac_get_error(err));
293 if ((err = aacEncoder_SetParam(s->handle, AACENC_AFTERBURNER,
294 s->afterburner)) != AACENC_OK) {
295 av_log(avctx, AV_LOG_ERROR, "Unable to set afterburner to %d: %s\n",
296 s->afterburner, aac_get_error(err));
300 if (avctx->cutoff > 0) {
301 if (avctx->cutoff < (avctx->sample_rate + 255) >> 8 || avctx->cutoff > 20000) {
302 av_log(avctx, AV_LOG_ERROR, "cutoff valid range is %d-20000\n",
303 (avctx->sample_rate + 255) >> 8);
306 if ((err = aacEncoder_SetParam(s->handle, AACENC_BANDWIDTH,
307 avctx->cutoff)) != AACENC_OK) {
308 av_log(avctx, AV_LOG_ERROR, "Unable to set the encoder bandwidth to %d: %s\n",
309 avctx->cutoff, aac_get_error(err));
314 if ((err = aacEncEncode(s->handle, NULL, NULL, NULL, NULL)) != AACENC_OK) {
315 av_log(avctx, AV_LOG_ERROR, "Unable to initialize the encoder: %s\n",
317 return AVERROR(EINVAL);
320 if ((err = aacEncInfo(s->handle, &info)) != AACENC_OK) {
321 av_log(avctx, AV_LOG_ERROR, "Unable to get encoder info: %s\n",
326 avctx->frame_size = info.frameLength;
327 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
328 avctx->initial_padding = info.nDelay;
330 avctx->initial_padding = info.encoderDelay;
332 ff_af_queue_init(avctx, &s->afq);
334 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
335 avctx->extradata_size = info.confSize;
336 avctx->extradata = av_mallocz(avctx->extradata_size +
337 AV_INPUT_BUFFER_PADDING_SIZE);
338 if (!avctx->extradata) {
339 ret = AVERROR(ENOMEM);
343 memcpy(avctx->extradata, info.confBuf, info.confSize);
347 aac_encode_close(avctx);
351 static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
352 const AVFrame *frame, int *got_packet_ptr)
354 AACContext *s = avctx->priv_data;
355 AACENC_BufDesc in_buf = { 0 }, out_buf = { 0 };
356 AACENC_InArgs in_args = { 0 };
357 AACENC_OutArgs out_args = { 0 };
358 int in_buffer_identifier = IN_AUDIO_DATA;
359 int in_buffer_size, in_buffer_element_size;
360 int out_buffer_identifier = OUT_BITSTREAM_DATA;
361 int out_buffer_size, out_buffer_element_size;
362 void *in_ptr, *out_ptr;
364 uint8_t dummy_buf[1];
367 /* handle end-of-stream small frame and flushing */
369 /* Must be a non-null pointer, even if it's a dummy. We could use
370 * the address of anything else on the stack as well. */
374 in_args.numInSamples = -1;
376 in_ptr = frame->data[0];
377 in_buffer_size = 2 * avctx->channels * frame->nb_samples;
379 in_args.numInSamples = avctx->channels * frame->nb_samples;
381 /* add current frame to the queue */
382 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
386 in_buffer_element_size = 2;
388 in_buf.bufs = &in_ptr;
389 in_buf.bufferIdentifiers = &in_buffer_identifier;
390 in_buf.bufSizes = &in_buffer_size;
391 in_buf.bufElSizes = &in_buffer_element_size;
393 /* The maximum packet size is 6144 bits aka 768 bytes per channel. */
394 if ((ret = ff_alloc_packet2(avctx, avpkt, FFMAX(8192, 768 * avctx->channels), 0)) < 0)
397 out_ptr = avpkt->data;
398 out_buffer_size = avpkt->size;
399 out_buffer_element_size = 1;
401 out_buf.bufs = &out_ptr;
402 out_buf.bufferIdentifiers = &out_buffer_identifier;
403 out_buf.bufSizes = &out_buffer_size;
404 out_buf.bufElSizes = &out_buffer_element_size;
406 if ((err = aacEncEncode(s->handle, &in_buf, &out_buf, &in_args,
407 &out_args)) != AACENC_OK) {
408 if (!frame && err == AACENC_ENCODE_EOF)
410 av_log(avctx, AV_LOG_ERROR, "Unable to encode frame: %s\n",
412 return AVERROR(EINVAL);
415 if (!out_args.numOutBytes)
418 /* Get the next frame pts & duration */
419 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
422 avpkt->size = out_args.numOutBytes;
427 static const AVProfile profiles[] = {
428 { FF_PROFILE_AAC_LOW, "LC" },
429 { FF_PROFILE_AAC_HE, "HE-AAC" },
430 { FF_PROFILE_AAC_HE_V2, "HE-AACv2" },
431 { FF_PROFILE_AAC_LD, "LD" },
432 { FF_PROFILE_AAC_ELD, "ELD" },
433 { FF_PROFILE_UNKNOWN },
436 static const AVCodecDefault aac_encode_defaults[] = {
441 static const uint64_t aac_channel_layout[] = {
444 AV_CH_LAYOUT_SURROUND,
445 AV_CH_LAYOUT_4POINT0,
446 AV_CH_LAYOUT_5POINT0_BACK,
447 AV_CH_LAYOUT_5POINT1_BACK,
448 #if FDKENC_VER_AT_LEAST(3, 4) // 3.4.12
449 AV_CH_LAYOUT_7POINT1_WIDE_BACK,
450 AV_CH_LAYOUT_7POINT1,
455 static const int aac_sample_rates[] = {
456 96000, 88200, 64000, 48000, 44100, 32000,
457 24000, 22050, 16000, 12000, 11025, 8000, 0
460 const AVCodec ff_libfdk_aac_encoder = {
461 .name = "libfdk_aac",
462 .long_name = NULL_IF_CONFIG_SMALL("Fraunhofer FDK AAC"),
463 .type = AVMEDIA_TYPE_AUDIO,
464 .id = AV_CODEC_ID_AAC,
465 .priv_data_size = sizeof(AACContext),
466 .init = aac_encode_init,
467 .encode2 = aac_encode_frame,
468 .close = aac_encode_close,
469 .capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY,
470 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
471 AV_SAMPLE_FMT_NONE },
472 .priv_class = &aac_enc_class,
473 .defaults = aac_encode_defaults,
474 .profiles = profiles,
475 .supported_samplerates = aac_sample_rates,
476 .channel_layouts = aac_channel_layout,
477 .wrapper_name = "libfdk",