2 * Copyright (C) 2016 Sebastian Dröge <sebastian@centricular.com>
4 * This library is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Library General Public
6 * License as published by the Free Software Foundation; either
7 * version 2 of the License, or (at your option) any later version.
9 * This library is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Library General Public License for more details.
14 * You should have received a copy of the GNU Library General Public
15 * License along with this library; if not, write to the
16 * Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
17 * Boston, MA 02110-1301, USA.
24 #include "gstfdkaac.h"
25 #include "gstfdkaacenc.h"
27 #include <gst/pbutils/pbutils.h>
32 * - Add support for other AOT / profiles
33 * - Expose more properties, e.g. afterburner and vbr
34 * - Signal encoder delay
35 * - LOAS / LATM support
44 #define DEFAULT_BITRATE (0)
46 #define SAMPLE_RATES " 8000, " \
59 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
62 GST_STATIC_CAPS ("audio/x-raw, "
63 "format = (string) " GST_AUDIO_NE (S16) ", "
64 "layout = (string) interleaved, "
65 "rate = (int) { " SAMPLE_RATES " }, "
66 "channels = (int) {1, 2, 3, 4, 5, 6, 8}")
69 static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
72 GST_STATIC_CAPS ("audio/mpeg, "
73 "mpegversion = (int) 4, "
74 "rate = (int) { " SAMPLE_RATES " }, "
75 "channels = (int) {1, 2, 3, 4, 5, 6, 8}, "
76 "stream-format = (string) { adts, adif, raw }, "
77 "base-profile = (string) lc, " "framed = (boolean) true")
80 GST_DEBUG_CATEGORY_STATIC (gst_fdkaacenc_debug);
81 #define GST_CAT_DEFAULT gst_fdkaacenc_debug
83 static void gst_fdkaacenc_set_property (GObject * object, guint prop_id,
84 const GValue * value, GParamSpec * pspec);
85 static void gst_fdkaacenc_get_property (GObject * object, guint prop_id,
86 GValue * value, GParamSpec * pspec);
87 static gboolean gst_fdkaacenc_start (GstAudioEncoder * enc);
88 static gboolean gst_fdkaacenc_stop (GstAudioEncoder * enc);
89 static gboolean gst_fdkaacenc_set_format (GstAudioEncoder * enc,
91 static GstFlowReturn gst_fdkaacenc_handle_frame (GstAudioEncoder * enc,
93 static GstCaps *gst_fdkaacenc_get_caps (GstAudioEncoder * enc,
95 static void gst_fdkaacenc_flush (GstAudioEncoder * enc);
97 G_DEFINE_TYPE (GstFdkAacEnc, gst_fdkaacenc, GST_TYPE_AUDIO_ENCODER);
98 GST_ELEMENT_REGISTER_DEFINE (fdkaacenc, "fdkaacenc", GST_RANK_PRIMARY,
102 gst_fdkaacenc_set_property (GObject * object, guint prop_id,
103 const GValue * value, GParamSpec * pspec)
105 GstFdkAacEnc *self = GST_FDKAACENC (object);
109 self->bitrate = g_value_get_int (value);
112 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
119 gst_fdkaacenc_get_property (GObject * object, guint prop_id,
120 GValue * value, GParamSpec * pspec)
122 GstFdkAacEnc *self = GST_FDKAACENC (object);
126 g_value_set_int (value, self->bitrate);
129 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
136 gst_fdkaacenc_start (GstAudioEncoder * enc)
138 GstFdkAacEnc *self = GST_FDKAACENC (enc);
140 GST_DEBUG_OBJECT (self, "start");
146 gst_fdkaacenc_stop (GstAudioEncoder * enc)
148 GstFdkAacEnc *self = GST_FDKAACENC (enc);
150 GST_DEBUG_OBJECT (self, "stop");
153 aacEncClose (&self->enc);
157 self->is_drained = TRUE;
162 gst_fdkaacenc_get_caps (GstAudioEncoder * enc, GstCaps * filter)
164 const GstFdkAacChannelLayout *layout;
167 caps = gst_caps_new_empty ();
169 for (layout = channel_layouts; layout->channels; layout++) {
170 gint channels = layout->channels;
172 gst_caps_make_writable (gst_pad_get_pad_template_caps
173 (GST_AUDIO_ENCODER_SINK_PAD (enc)));
176 gst_caps_set_simple (tmp, "channels", G_TYPE_INT, channels, NULL);
178 guint64 channel_mask;
179 gst_audio_channel_positions_to_mask (layout->positions, channels, FALSE,
181 gst_caps_set_simple (tmp, "channels", G_TYPE_INT, channels,
182 "channel-mask", GST_TYPE_BITMASK, channel_mask, NULL);
185 gst_caps_append (caps, tmp);
188 res = gst_audio_encoder_proxy_getcaps (enc, caps, filter);
189 gst_caps_unref (caps);
195 gst_fdkaacenc_set_format (GstAudioEncoder * enc, GstAudioInfo * info)
197 GstFdkAacEnc *self = GST_FDKAACENC (enc);
198 gboolean ret = FALSE;
199 GstCaps *allowed_caps;
202 gint transmux = 0, aot = AOT_AAC_LC;
203 gint mpegversion = 4;
204 CHANNEL_MODE channel_mode;
205 AACENC_InfoStruct enc_info = { 0 };
208 if (self->enc && !self->is_drained) {
210 gst_fdkaacenc_handle_frame (enc, NULL);
211 aacEncClose (&self->enc);
212 self->is_drained = TRUE;
215 allowed_caps = gst_pad_get_allowed_caps (GST_AUDIO_ENCODER_SRC_PAD (self));
217 GST_DEBUG_OBJECT (self, "allowed caps: %" GST_PTR_FORMAT, allowed_caps);
219 if (allowed_caps && gst_caps_get_size (allowed_caps) > 0) {
220 GstStructure *s = gst_caps_get_structure (allowed_caps, 0);
221 const gchar *str = NULL;
223 if ((str = gst_structure_get_string (s, "stream-format"))) {
224 if (strcmp (str, "adts") == 0) {
225 GST_DEBUG_OBJECT (self, "use ADTS format for output");
227 } else if (strcmp (str, "adif") == 0) {
228 GST_DEBUG_OBJECT (self, "use ADIF format for output");
230 } else if (strcmp (str, "raw") == 0) {
231 GST_DEBUG_OBJECT (self, "use RAW format for output");
236 gst_structure_get_int (s, "mpegversion", &mpegversion);
239 gst_caps_unref (allowed_caps);
241 err = aacEncOpen (&self->enc, 0, GST_AUDIO_INFO_CHANNELS (info));
242 if (err != AACENC_OK) {
243 GST_ERROR_OBJECT (self, "Unable to open encoder: %d", err);
249 if ((err = aacEncoder_SetParam (self->enc, AACENC_AOT, aot)) != AACENC_OK) {
250 GST_ERROR_OBJECT (self, "Unable to set AOT %d: %d", aot, err);
254 if ((err = aacEncoder_SetParam (self->enc, AACENC_SAMPLERATE,
255 GST_AUDIO_INFO_RATE (info))) != AACENC_OK) {
256 GST_ERROR_OBJECT (self, "Unable to set sample rate %d: %d",
257 GST_AUDIO_INFO_RATE (info), err);
261 if (GST_AUDIO_INFO_CHANNELS (info) == 1) {
262 channel_mode = MODE_1;
263 self->need_reorder = FALSE;
264 self->aac_positions = NULL;
266 gint in_channels = GST_AUDIO_INFO_CHANNELS (info);
267 const GstAudioChannelPosition *in_positions =
268 &GST_AUDIO_INFO_POSITION (info, 0);
269 guint64 in_channel_mask;
270 const GstFdkAacChannelLayout *layout;
272 gst_audio_channel_positions_to_mask (in_positions, in_channels, FALSE,
275 for (layout = channel_layouts; layout->channels; layout++) {
276 gint channels = layout->channels;
277 const GstAudioChannelPosition *positions = layout->positions;
278 guint64 channel_mask;
280 if (channels != in_channels)
283 gst_audio_channel_positions_to_mask (positions, channels, FALSE,
285 if (channel_mask != in_channel_mask)
288 channel_mode = layout->mode;
289 self->need_reorder = memcmp (positions, in_positions,
290 channels * sizeof *positions) != 0;
291 self->aac_positions = positions;
295 if (!layout->channels) {
296 GST_ERROR_OBJECT (self, "Couldn't find a valid channel layout");
301 if ((err = aacEncoder_SetParam (self->enc, AACENC_CHANNELMODE,
302 channel_mode)) != AACENC_OK) {
303 GST_ERROR_OBJECT (self, "Unable to set channel mode %d: %d", channel_mode,
308 /* MPEG channel order */
309 if ((err = aacEncoder_SetParam (self->enc, AACENC_CHANNELORDER,
311 GST_ERROR_OBJECT (self, "Unable to set channel order %d: %d", channel_mode,
316 bitrate = self->bitrate;
318 * http://wiki.hydrogenaud.io/index.php?title=Fraunhofer_FDK_AAC#Recommended_Sampling_Rate_and_Bitrate_Combinations
321 if (GST_AUDIO_INFO_CHANNELS (info) == 1) {
322 if (GST_AUDIO_INFO_RATE (info) < 16000) {
324 } else if (GST_AUDIO_INFO_RATE (info) == 16000) {
326 } else if (GST_AUDIO_INFO_RATE (info) < 32000) {
328 } else if (GST_AUDIO_INFO_RATE (info) == 32000) {
330 } else if (GST_AUDIO_INFO_RATE (info) <= 44100) {
335 } else if (GST_AUDIO_INFO_CHANNELS (info) == 2) {
336 if (GST_AUDIO_INFO_RATE (info) < 16000) {
338 } else if (GST_AUDIO_INFO_RATE (info) == 16000) {
340 } else if (GST_AUDIO_INFO_RATE (info) < 22050) {
342 } else if (GST_AUDIO_INFO_RATE (info) < 32000) {
344 } else if (GST_AUDIO_INFO_RATE (info) == 32000) {
346 } else if (GST_AUDIO_INFO_RATE (info) <= 44100) {
353 if (GST_AUDIO_INFO_RATE (info) < 32000) {
355 } else if (GST_AUDIO_INFO_RATE (info) <= 44100) {
363 if ((err = aacEncoder_SetParam (self->enc, AACENC_TRANSMUX,
364 transmux)) != AACENC_OK) {
365 GST_ERROR_OBJECT (self, "Unable to set transmux %d: %d", transmux, err);
369 if ((err = aacEncoder_SetParam (self->enc, AACENC_BITRATE,
370 bitrate)) != AACENC_OK) {
371 GST_ERROR_OBJECT (self, "Unable to set bitrate %d: %d", bitrate, err);
375 if ((err = aacEncEncode (self->enc, NULL, NULL, NULL, NULL)) != AACENC_OK) {
376 GST_ERROR_OBJECT (self, "Unable to initialize encoder: %d", err);
380 if ((err = aacEncInfo (self->enc, &enc_info)) != AACENC_OK) {
381 GST_ERROR_OBJECT (self, "Unable to get encoder info: %d", err);
385 gst_audio_encoder_set_frame_max (enc, 1);
386 gst_audio_encoder_set_frame_samples_min (enc, enc_info.frameLength);
387 gst_audio_encoder_set_frame_samples_max (enc, enc_info.frameLength);
388 gst_audio_encoder_set_hard_min (enc, FALSE);
389 self->outbuf_size = enc_info.maxOutBufBytes;
390 self->samples_per_frame = enc_info.frameLength;
392 src_caps = gst_caps_new_simple ("audio/mpeg",
393 "mpegversion", G_TYPE_INT, mpegversion,
394 "channels", G_TYPE_INT, GST_AUDIO_INFO_CHANNELS (info),
395 "framed", G_TYPE_BOOLEAN, TRUE,
396 "rate", G_TYPE_INT, GST_AUDIO_INFO_RATE (info), NULL);
400 GstBuffer *codec_data =
401 gst_buffer_new_memdup (enc_info.confBuf, enc_info.confSize);
402 gst_caps_set_simple (src_caps, "codec_data", GST_TYPE_BUFFER, codec_data,
403 "stream-format", G_TYPE_STRING, "raw", NULL);
404 gst_buffer_unref (codec_data);
405 } else if (transmux == 1) {
406 gst_caps_set_simple (src_caps, "stream-format", G_TYPE_STRING, "adif",
408 } else if (transmux == 2) {
409 gst_caps_set_simple (src_caps, "stream-format", G_TYPE_STRING, "adts",
412 g_assert_not_reached ();
415 gst_codec_utils_aac_caps_set_level_and_profile (src_caps, enc_info.confBuf,
418 ret = gst_audio_encoder_set_output_format (enc, src_caps);
419 gst_caps_unref (src_caps);
425 gst_fdkaacenc_handle_frame (GstAudioEncoder * enc, GstBuffer * inbuf)
427 GstFdkAacEnc *self = GST_FDKAACENC (enc);
428 GstFlowReturn ret = GST_FLOW_OK;
430 GstMapInfo imap, omap;
432 AACENC_BufDesc in_desc = { 0 };
433 AACENC_BufDesc out_desc = { 0 };
434 AACENC_InArgs in_args = { 0 };
435 AACENC_OutArgs out_args = { 0 };
436 gint in_id = IN_AUDIO_DATA, out_id = OUT_BITSTREAM_DATA;
437 gint in_sizes, out_sizes;
438 gint in_el_sizes, out_el_sizes;
441 info = gst_audio_encoder_get_audio_info (enc);
444 if (self->need_reorder) {
445 inbuf = gst_buffer_copy (inbuf);
446 gst_buffer_map (inbuf, &imap, GST_MAP_READWRITE);
447 gst_audio_reorder_channels (imap.data, imap.size,
448 GST_AUDIO_INFO_FORMAT (info), GST_AUDIO_INFO_CHANNELS (info),
449 &GST_AUDIO_INFO_POSITION (info, 0), self->aac_positions);
451 gst_buffer_map (inbuf, &imap, GST_MAP_READ);
454 in_args.numInSamples = imap.size / GST_AUDIO_INFO_BPS (info);
456 in_sizes = imap.size;
457 in_el_sizes = GST_AUDIO_INFO_BPS (info);
460 in_args.numInSamples = -1;
466 /* We unset is_drained even if there's no inbuf. Basically this is a
467 * workaround for aacEncEncode always producing 1024 bytes even without any
468 * input, thus messing up with the base class counting */
469 self->is_drained = FALSE;
471 in_desc.bufferIdentifiers = &in_id;
472 in_desc.bufs = (void *) &imap.data;
473 in_desc.bufSizes = &in_sizes;
474 in_desc.bufElSizes = &in_el_sizes;
476 outbuf = gst_audio_encoder_allocate_output_buffer (enc, self->outbuf_size);
478 ret = GST_FLOW_ERROR;
482 gst_buffer_map (outbuf, &omap, GST_MAP_WRITE);
483 out_sizes = omap.size;
485 out_desc.bufferIdentifiers = &out_id;
486 out_desc.numBufs = 1;
487 out_desc.bufs = (void *) &omap.data;
488 out_desc.bufSizes = &out_sizes;
489 out_desc.bufElSizes = &out_el_sizes;
491 err = aacEncEncode (self->enc, &in_desc, &out_desc, &in_args, &out_args);
492 if (err == AACENC_ENCODE_EOF && !inbuf)
494 else if (err != AACENC_OK) {
495 GST_ERROR_OBJECT (self, "Failed to encode data: %d", err);
496 ret = GST_FLOW_ERROR;
501 gst_buffer_unmap (inbuf, &imap);
502 if (self->need_reorder)
503 gst_buffer_unref (inbuf);
507 if (!out_args.numOutBytes)
510 gst_buffer_unmap (outbuf, &omap);
511 gst_buffer_set_size (outbuf, out_args.numOutBytes);
513 ret = gst_audio_encoder_finish_frame (enc, outbuf, self->samples_per_frame);
518 gst_buffer_unmap (outbuf, &omap);
519 gst_buffer_unref (outbuf);
522 gst_buffer_unmap (inbuf, &imap);
523 if (self->need_reorder)
524 gst_buffer_unref (inbuf);
531 gst_fdkaacenc_flush (GstAudioEncoder * enc)
533 GstFdkAacEnc *self = GST_FDKAACENC (enc);
534 GstAudioInfo *info = gst_audio_encoder_get_audio_info (enc);
536 aacEncClose (&self->enc);
538 self->is_drained = TRUE;
540 if (GST_AUDIO_INFO_IS_VALID (info))
541 gst_fdkaacenc_set_format (enc, info);
545 gst_fdkaacenc_init (GstFdkAacEnc * self)
547 self->bitrate = DEFAULT_BITRATE;
549 self->is_drained = TRUE;
551 gst_audio_encoder_set_drainable (GST_AUDIO_ENCODER (self), TRUE);
555 gst_fdkaacenc_class_init (GstFdkAacEncClass * klass)
557 GObjectClass *object_class = G_OBJECT_CLASS (klass);
558 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
559 GstAudioEncoderClass *base_class = GST_AUDIO_ENCODER_CLASS (klass);
561 object_class->set_property = GST_DEBUG_FUNCPTR (gst_fdkaacenc_set_property);
562 object_class->get_property = GST_DEBUG_FUNCPTR (gst_fdkaacenc_get_property);
564 base_class->start = GST_DEBUG_FUNCPTR (gst_fdkaacenc_start);
565 base_class->stop = GST_DEBUG_FUNCPTR (gst_fdkaacenc_stop);
566 base_class->set_format = GST_DEBUG_FUNCPTR (gst_fdkaacenc_set_format);
567 base_class->getcaps = GST_DEBUG_FUNCPTR (gst_fdkaacenc_get_caps);
568 base_class->handle_frame = GST_DEBUG_FUNCPTR (gst_fdkaacenc_handle_frame);
569 base_class->flush = GST_DEBUG_FUNCPTR (gst_fdkaacenc_flush);
571 g_object_class_install_property (object_class, PROP_BITRATE,
572 g_param_spec_int ("bitrate",
574 "Target Audio Bitrate (0 = fixed value based on "
575 " sample rate and channel count)",
576 0, G_MAXINT, DEFAULT_BITRATE,
577 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
579 gst_element_class_add_static_pad_template (element_class, &sink_template);
580 gst_element_class_add_static_pad_template (element_class, &src_template);
582 gst_element_class_set_static_metadata (element_class, "FDK AAC audio encoder",
583 "Codec/Encoder/Audio/Converter", "FDK AAC audio encoder",
584 "Sebastian Dröge <sebastian@centricular.com>");
586 GST_DEBUG_CATEGORY_INIT (gst_fdkaacenc_debug, "fdkaacenc", 0,