3 * Copyright (c) 2012 Martin Storsjo
5 * This file is part of Libav.
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"
29 #define FDKENC_VER_AT_LEAST(vl0, vl1) \
30 (defined(AACENCODER_LIB_VL0) && \
31 ((AACENCODER_LIB_VL0 > vl0) || \
32 (AACENCODER_LIB_VL0 == vl0 && AACENCODER_LIB_VL1 >= vl1)))
34 typedef struct AACContext {
36 HANDLE_AACENCODER handle;
48 static const AVOption aac_enc_options[] = {
49 { "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 },
50 { "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 },
51 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
52 { "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 },
54 { "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" },
55 { "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" },
56 { "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" },
57 { "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" },
58 { "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" },
59 { "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 },
60 { "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 },
61 { "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 },
65 static const AVClass aac_enc_class = {
66 .class_name = "libfdk_aac",
67 .item_name = av_default_item_name,
68 .option = aac_enc_options,
69 .version = LIBAVUTIL_VERSION_INT,
72 static const char *aac_get_error(AACENC_ERROR err)
77 case AACENC_INVALID_HANDLE:
78 return "Invalid handle";
79 case AACENC_MEMORY_ERROR:
80 return "Memory allocation error";
81 case AACENC_UNSUPPORTED_PARAMETER:
82 return "Unsupported parameter";
83 case AACENC_INVALID_CONFIG:
84 return "Invalid config";
85 case AACENC_INIT_ERROR:
86 return "Initialization error";
87 case AACENC_INIT_AAC_ERROR:
88 return "AAC library initialization error";
89 case AACENC_INIT_SBR_ERROR:
90 return "SBR library initialization error";
91 case AACENC_INIT_TP_ERROR:
92 return "Transport library initialization error";
93 case AACENC_INIT_META_ERROR:
94 return "Metadata library initialization error";
95 case AACENC_ENCODE_ERROR:
96 return "Encoding error";
97 case AACENC_ENCODE_EOF:
100 return "Unknown error";
104 static int aac_encode_close(AVCodecContext *avctx)
106 AACContext *s = avctx->priv_data;
109 aacEncClose(&s->handle);
110 av_freep(&avctx->extradata);
111 ff_af_queue_close(&s->afq);
116 static av_cold int aac_encode_init(AVCodecContext *avctx)
118 AACContext *s = avctx->priv_data;
119 int ret = AVERROR(EINVAL);
120 AACENC_InfoStruct info = { 0 };
123 int aot = FF_PROFILE_AAC_LOW + 1;
124 int sce = 0, cpe = 0;
126 if ((err = aacEncOpen(&s->handle, 0, avctx->channels)) != AACENC_OK) {
127 av_log(avctx, AV_LOG_ERROR, "Unable to open the encoder: %s\n",
132 if (avctx->profile != FF_PROFILE_UNKNOWN)
133 aot = avctx->profile + 1;
135 if ((err = aacEncoder_SetParam(s->handle, AACENC_AOT, aot)) != AACENC_OK) {
136 av_log(avctx, AV_LOG_ERROR, "Unable to set the AOT %d: %s\n",
137 aot, aac_get_error(err));
141 if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_sbr) {
142 if ((err = aacEncoder_SetParam(s->handle, AACENC_SBR_MODE,
144 av_log(avctx, AV_LOG_ERROR, "Unable to enable SBR for ELD: %s\n",
150 if ((err = aacEncoder_SetParam(s->handle, AACENC_SAMPLERATE,
151 avctx->sample_rate)) != AACENC_OK) {
152 av_log(avctx, AV_LOG_ERROR, "Unable to set the sample rate %d: %s\n",
153 avctx->sample_rate, aac_get_error(err));
157 switch (avctx->channels) {
158 case 1: mode = MODE_1; sce = 1; cpe = 0; break;
160 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
161 // (profile + 1) to map from profile range to AOT range
162 if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_v2) {
163 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
164 128)) != AACENC_OK) {
165 av_log(avctx, AV_LOG_ERROR, "Unable to enable ELDv2: %s\n",
181 case 3: mode = MODE_1_2; sce = 1; cpe = 1; break;
182 case 4: mode = MODE_1_2_1; sce = 2; cpe = 1; break;
183 case 5: mode = MODE_1_2_2; sce = 1; cpe = 2; break;
184 case 6: mode = MODE_1_2_2_1; sce = 2; cpe = 2; break;
185 /* The version macro is introduced the same time as the 7.1 support, so this
187 #if FDKENC_VER_AT_LEAST(3, 4) // 3.4.12
191 if (avctx->channel_layout == AV_CH_LAYOUT_7POINT1) {
192 mode = MODE_7_1_REAR_SURROUND;
194 // MODE_1_2_2_2_1 and MODE_7_1_FRONT_CENTER use the same channel layout
195 mode = MODE_7_1_FRONT_CENTER;
200 av_log(avctx, AV_LOG_ERROR,
201 "Unsupported number of channels %d\n", avctx->channels);
205 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
206 mode)) != AACENC_OK) {
207 av_log(avctx, AV_LOG_ERROR,
208 "Unable to set channel mode %d: %s\n", mode, aac_get_error(err));
212 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELORDER,
214 av_log(avctx, AV_LOG_ERROR,
215 "Unable to set wav channel order %d: %s\n",
216 mode, aac_get_error(err));
220 if (avctx->flags & AV_CODEC_FLAG_QSCALE || s->vbr) {
221 int mode = s->vbr ? s->vbr : avctx->global_quality;
222 if (mode < 1 || mode > 5) {
223 av_log(avctx, AV_LOG_WARNING,
224 "VBR quality %d out of range, should be 1-5\n", mode);
225 mode = av_clip(mode, 1, 5);
227 av_log(avctx, AV_LOG_WARNING,
228 "Note, the VBR setting is unsupported and only works with "
229 "some parameter combinations\n");
230 if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATEMODE,
231 mode)) != AACENC_OK) {
232 av_log(avctx, AV_LOG_ERROR, "Unable to set the VBR bitrate mode %d: %s\n",
233 mode, aac_get_error(err));
237 if (avctx->bit_rate <= 0) {
238 if (avctx->profile == FF_PROFILE_AAC_HE_V2) {
242 avctx->bit_rate = (96*sce + 128*cpe) * avctx->sample_rate / 44;
243 if (avctx->profile == FF_PROFILE_AAC_HE ||
244 avctx->profile == FF_PROFILE_AAC_HE_V2 ||
245 avctx->profile == FF_PROFILE_MPEG2_AAC_HE ||
247 avctx->bit_rate /= 2;
249 if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATE,
250 avctx->bit_rate)) != AACENC_OK) {
251 av_log(avctx, AV_LOG_ERROR, "Unable to set the bitrate %d: %s\n",
252 avctx->bit_rate, aac_get_error(err));
257 /* Choose bitstream format - if global header is requested, use
258 * raw access units, otherwise use ADTS. */
259 if ((err = aacEncoder_SetParam(s->handle, AACENC_TRANSMUX,
260 avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? TT_MP4_RAW :
261 s->latm ? TT_MP4_LOAS : TT_MP4_ADTS)) != AACENC_OK) {
262 av_log(avctx, AV_LOG_ERROR, "Unable to set the transmux format: %s\n",
267 if (s->latm && s->header_period) {
268 if ((err = aacEncoder_SetParam(s->handle, AACENC_HEADER_PERIOD,
269 s->header_period)) != AACENC_OK) {
270 av_log(avctx, AV_LOG_ERROR, "Unable to set header period: %s\n",
276 /* If no signaling mode is chosen, use explicit hierarchical signaling
277 * if using mp4 mode (raw access units, with global header) and
278 * implicit signaling if using ADTS. */
279 if (s->signaling < 0)
280 s->signaling = avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? 2 : 0;
282 if ((err = aacEncoder_SetParam(s->handle, AACENC_SIGNALING_MODE,
283 s->signaling)) != AACENC_OK) {
284 av_log(avctx, AV_LOG_ERROR, "Unable to set signaling mode %d: %s\n",
285 s->signaling, aac_get_error(err));
289 if ((err = aacEncoder_SetParam(s->handle, AACENC_AFTERBURNER,
290 s->afterburner)) != AACENC_OK) {
291 av_log(avctx, AV_LOG_ERROR, "Unable to set afterburner to %d: %s\n",
292 s->afterburner, aac_get_error(err));
296 if (avctx->cutoff > 0) {
297 if (avctx->cutoff < (avctx->sample_rate + 255) >> 8 || avctx->cutoff > 20000) {
298 av_log(avctx, AV_LOG_ERROR, "cutoff valid range is %d-20000\n",
299 (avctx->sample_rate + 255) >> 8);
302 if ((err = aacEncoder_SetParam(s->handle, AACENC_BANDWIDTH,
303 avctx->cutoff)) != AACENC_OK) {
304 av_log(avctx, AV_LOG_ERROR, "Unable to set the encoder bandwidth to %d: %s\n",
305 avctx->cutoff, aac_get_error(err));
310 if ((err = aacEncEncode(s->handle, NULL, NULL, NULL, NULL)) != AACENC_OK) {
311 av_log(avctx, AV_LOG_ERROR, "Unable to initialize the encoder: %s\n",
313 return AVERROR(EINVAL);
316 if ((err = aacEncInfo(s->handle, &info)) != AACENC_OK) {
317 av_log(avctx, AV_LOG_ERROR, "Unable to get encoder info: %s\n",
322 avctx->frame_size = info.frameLength;
323 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
324 avctx->initial_padding = info.nDelay;
326 avctx->initial_padding = info.encoderDelay;
328 ff_af_queue_init(avctx, &s->afq);
330 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
331 avctx->extradata_size = info.confSize;
332 avctx->extradata = av_mallocz(avctx->extradata_size +
333 AV_INPUT_BUFFER_PADDING_SIZE);
334 if (!avctx->extradata) {
335 ret = AVERROR(ENOMEM);
339 memcpy(avctx->extradata, info.confBuf, info.confSize);
343 aac_encode_close(avctx);
347 static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
348 const AVFrame *frame, int *got_packet_ptr)
350 AACContext *s = avctx->priv_data;
351 AACENC_BufDesc in_buf = { 0 }, out_buf = { 0 };
352 AACENC_InArgs in_args = { 0 };
353 AACENC_OutArgs out_args = { 0 };
354 int in_buffer_identifier = IN_AUDIO_DATA;
355 int in_buffer_size, in_buffer_element_size;
356 int out_buffer_identifier = OUT_BITSTREAM_DATA;
357 int out_buffer_size, out_buffer_element_size;
358 void *in_ptr, *out_ptr;
360 uint8_t dummy_buf[1];
363 /* handle end-of-stream small frame and flushing */
365 /* Must be a non-null pointer, even if it's a dummy. We could use
366 * the address of anything else on the stack as well. */
370 in_args.numInSamples = -1;
372 in_ptr = frame->data[0];
373 in_buffer_size = 2 * avctx->channels * frame->nb_samples;
375 in_args.numInSamples = avctx->channels * frame->nb_samples;
377 /* add current frame to the queue */
378 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
382 in_buffer_element_size = 2;
384 in_buf.bufs = &in_ptr;
385 in_buf.bufferIdentifiers = &in_buffer_identifier;
386 in_buf.bufSizes = &in_buffer_size;
387 in_buf.bufElSizes = &in_buffer_element_size;
389 /* The maximum packet size is 6144 bits aka 768 bytes per channel. */
390 if ((ret = ff_alloc_packet(avpkt, FFMAX(8192, 768 * avctx->channels)))) {
391 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
395 out_ptr = avpkt->data;
396 out_buffer_size = avpkt->size;
397 out_buffer_element_size = 1;
399 out_buf.bufs = &out_ptr;
400 out_buf.bufferIdentifiers = &out_buffer_identifier;
401 out_buf.bufSizes = &out_buffer_size;
402 out_buf.bufElSizes = &out_buffer_element_size;
404 if ((err = aacEncEncode(s->handle, &in_buf, &out_buf, &in_args,
405 &out_args)) != AACENC_OK) {
406 if (!frame && err == AACENC_ENCODE_EOF)
408 av_log(avctx, AV_LOG_ERROR, "Unable to encode frame: %s\n",
410 return AVERROR(EINVAL);
413 if (!out_args.numOutBytes)
416 /* Get the next frame pts & duration */
417 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
420 avpkt->size = out_args.numOutBytes;
425 static const AVProfile profiles[] = {
426 { FF_PROFILE_AAC_LOW, "LC" },
427 { FF_PROFILE_AAC_HE, "HE-AAC" },
428 { FF_PROFILE_AAC_HE_V2, "HE-AACv2" },
429 { FF_PROFILE_AAC_LD, "LD" },
430 { FF_PROFILE_AAC_ELD, "ELD" },
431 { FF_PROFILE_UNKNOWN },
434 static const AVCodecDefault aac_encode_defaults[] = {
439 static const uint64_t aac_channel_layout[] = {
442 AV_CH_LAYOUT_SURROUND,
443 AV_CH_LAYOUT_4POINT0,
444 AV_CH_LAYOUT_5POINT0_BACK,
445 AV_CH_LAYOUT_5POINT1_BACK,
446 #if FDKENC_VER_AT_LEAST(3, 4) // 3.4.12
447 AV_CH_LAYOUT_7POINT1_WIDE_BACK,
448 AV_CH_LAYOUT_7POINT1,
453 static const int aac_sample_rates[] = {
454 96000, 88200, 64000, 48000, 44100, 32000,
455 24000, 22050, 16000, 12000, 11025, 8000, 0
458 AVCodec ff_libfdk_aac_encoder = {
459 .name = "libfdk_aac",
460 .long_name = NULL_IF_CONFIG_SMALL("Fraunhofer FDK AAC"),
461 .type = AVMEDIA_TYPE_AUDIO,
462 .id = AV_CODEC_ID_AAC,
463 .priv_data_size = sizeof(AACContext),
464 .init = aac_encode_init,
465 .encode2 = aac_encode_frame,
466 .close = aac_encode_close,
467 .capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY,
468 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
469 AV_SAMPLE_FMT_NONE },
470 .priv_class = &aac_enc_class,
471 .defaults = aac_encode_defaults,
472 .profiles = profiles,
473 .supported_samplerates = aac_sample_rates,
474 .channel_layouts = aac_channel_layout,
475 .wrapper_name = "libfdk",