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);
100 gst_fdkaacenc_set_property (GObject * object, guint prop_id,
101 const GValue * value, GParamSpec * pspec)
103 GstFdkAacEnc *self = GST_FDKAACENC (object);
107 self->bitrate = g_value_get_int (value);
110 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
117 gst_fdkaacenc_get_property (GObject * object, guint prop_id,
118 GValue * value, GParamSpec * pspec)
120 GstFdkAacEnc *self = GST_FDKAACENC (object);
124 g_value_set_int (value, self->bitrate);
127 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
134 gst_fdkaacenc_start (GstAudioEncoder * enc)
136 GstFdkAacEnc *self = GST_FDKAACENC (enc);
138 GST_DEBUG_OBJECT (self, "start");
144 gst_fdkaacenc_stop (GstAudioEncoder * enc)
146 GstFdkAacEnc *self = GST_FDKAACENC (enc);
148 GST_DEBUG_OBJECT (self, "stop");
151 aacEncClose (&self->enc);
155 self->is_drained = TRUE;
160 gst_fdkaacenc_get_caps (GstAudioEncoder * enc, GstCaps * filter)
162 const GstFdkAacChannelLayout *layout;
165 caps = gst_caps_new_empty ();
167 for (layout = channel_layouts; layout->channels; layout++) {
168 gint channels = layout->channels;
170 gst_caps_make_writable (gst_pad_get_pad_template_caps
171 (GST_AUDIO_ENCODER_SINK_PAD (enc)));
174 gst_caps_set_simple (tmp, "channels", G_TYPE_INT, channels, NULL);
176 guint64 channel_mask;
177 gst_audio_channel_positions_to_mask (layout->positions, channels, FALSE,
179 gst_caps_set_simple (tmp, "channels", G_TYPE_INT, channels,
180 "channel-mask", GST_TYPE_BITMASK, channel_mask, NULL);
183 gst_caps_append (caps, tmp);
186 res = gst_audio_encoder_proxy_getcaps (enc, caps, filter);
187 gst_caps_unref (caps);
193 gst_fdkaacenc_set_format (GstAudioEncoder * enc, GstAudioInfo * info)
195 GstFdkAacEnc *self = GST_FDKAACENC (enc);
196 gboolean ret = FALSE;
197 GstCaps *allowed_caps;
200 gint transmux = 0, aot = AOT_AAC_LC;
201 gint mpegversion = 4;
202 CHANNEL_MODE channel_mode;
203 AACENC_InfoStruct enc_info = { 0 };
206 if (self->enc && !self->is_drained) {
208 gst_fdkaacenc_handle_frame (enc, NULL);
209 aacEncClose (&self->enc);
210 self->is_drained = TRUE;
213 allowed_caps = gst_pad_get_allowed_caps (GST_AUDIO_ENCODER_SRC_PAD (self));
215 GST_DEBUG_OBJECT (self, "allowed caps: %" GST_PTR_FORMAT, allowed_caps);
217 if (allowed_caps && gst_caps_get_size (allowed_caps) > 0) {
218 GstStructure *s = gst_caps_get_structure (allowed_caps, 0);
219 const gchar *str = NULL;
221 if ((str = gst_structure_get_string (s, "stream-format"))) {
222 if (strcmp (str, "adts") == 0) {
223 GST_DEBUG_OBJECT (self, "use ADTS format for output");
225 } else if (strcmp (str, "adif") == 0) {
226 GST_DEBUG_OBJECT (self, "use ADIF format for output");
228 } else if (strcmp (str, "raw") == 0) {
229 GST_DEBUG_OBJECT (self, "use RAW format for output");
234 gst_structure_get_int (s, "mpegversion", &mpegversion);
237 gst_caps_unref (allowed_caps);
239 err = aacEncOpen (&self->enc, 0, GST_AUDIO_INFO_CHANNELS (info));
240 if (err != AACENC_OK) {
241 GST_ERROR_OBJECT (self, "Unable to open encoder: %d", err);
247 if ((err = aacEncoder_SetParam (self->enc, AACENC_AOT, aot)) != AACENC_OK) {
248 GST_ERROR_OBJECT (self, "Unable to set AOT %d: %d", aot, err);
252 if ((err = aacEncoder_SetParam (self->enc, AACENC_SAMPLERATE,
253 GST_AUDIO_INFO_RATE (info))) != AACENC_OK) {
254 GST_ERROR_OBJECT (self, "Unable to set sample rate %d: %d",
255 GST_AUDIO_INFO_RATE (info), err);
259 if (GST_AUDIO_INFO_CHANNELS (info) == 1) {
260 channel_mode = MODE_1;
261 self->need_reorder = FALSE;
262 self->aac_positions = NULL;
264 gint in_channels = GST_AUDIO_INFO_CHANNELS (info);
265 const GstAudioChannelPosition *in_positions =
266 &GST_AUDIO_INFO_POSITION (info, 0);
267 guint64 in_channel_mask;
268 const GstFdkAacChannelLayout *layout;
270 gst_audio_channel_positions_to_mask (in_positions, in_channels, FALSE,
273 for (layout = channel_layouts; layout->channels; layout++) {
274 gint channels = layout->channels;
275 const GstAudioChannelPosition *positions = layout->positions;
276 guint64 channel_mask;
278 if (channels != in_channels)
281 gst_audio_channel_positions_to_mask (positions, channels, FALSE,
283 if (channel_mask != in_channel_mask)
286 channel_mode = layout->mode;
287 self->need_reorder = memcmp (positions, in_positions,
288 channels * sizeof *positions) != 0;
289 self->aac_positions = positions;
293 if (!layout->channels) {
294 GST_ERROR_OBJECT (self, "Couldn't find a valid channel layout");
299 if ((err = aacEncoder_SetParam (self->enc, AACENC_CHANNELMODE,
300 channel_mode)) != AACENC_OK) {
301 GST_ERROR_OBJECT (self, "Unable to set channel mode %d: %d", channel_mode,
306 /* MPEG channel order */
307 if ((err = aacEncoder_SetParam (self->enc, AACENC_CHANNELORDER,
309 GST_ERROR_OBJECT (self, "Unable to set channel order %d: %d", channel_mode,
314 bitrate = self->bitrate;
316 * http://wiki.hydrogenaud.io/index.php?title=Fraunhofer_FDK_AAC#Recommended_Sampling_Rate_and_Bitrate_Combinations
319 if (GST_AUDIO_INFO_CHANNELS (info) == 1) {
320 if (GST_AUDIO_INFO_RATE (info) < 16000) {
322 } else if (GST_AUDIO_INFO_RATE (info) == 16000) {
324 } else if (GST_AUDIO_INFO_RATE (info) < 32000) {
326 } else if (GST_AUDIO_INFO_RATE (info) == 32000) {
328 } else if (GST_AUDIO_INFO_RATE (info) <= 44100) {
333 } else if (GST_AUDIO_INFO_CHANNELS (info) == 2) {
334 if (GST_AUDIO_INFO_RATE (info) < 16000) {
336 } else if (GST_AUDIO_INFO_RATE (info) == 16000) {
338 } else if (GST_AUDIO_INFO_RATE (info) < 22050) {
340 } else if (GST_AUDIO_INFO_RATE (info) < 32000) {
342 } else if (GST_AUDIO_INFO_RATE (info) == 32000) {
344 } else if (GST_AUDIO_INFO_RATE (info) <= 44100) {
351 if (GST_AUDIO_INFO_RATE (info) < 32000) {
353 } else if (GST_AUDIO_INFO_RATE (info) <= 44100) {
361 if ((err = aacEncoder_SetParam (self->enc, AACENC_TRANSMUX,
362 transmux)) != AACENC_OK) {
363 GST_ERROR_OBJECT (self, "Unable to set transmux %d: %d", transmux, err);
367 if ((err = aacEncoder_SetParam (self->enc, AACENC_BITRATE,
368 bitrate)) != AACENC_OK) {
369 GST_ERROR_OBJECT (self, "Unable to set bitrate %d: %d", bitrate, err);
373 if ((err = aacEncEncode (self->enc, NULL, NULL, NULL, NULL)) != AACENC_OK) {
374 GST_ERROR_OBJECT (self, "Unable to initialize encoder: %d", err);
378 if ((err = aacEncInfo (self->enc, &enc_info)) != AACENC_OK) {
379 GST_ERROR_OBJECT (self, "Unable to get encoder info: %d", err);
383 gst_audio_encoder_set_frame_max (enc, 1);
384 gst_audio_encoder_set_frame_samples_min (enc, enc_info.frameLength);
385 gst_audio_encoder_set_frame_samples_max (enc, enc_info.frameLength);
386 gst_audio_encoder_set_hard_min (enc, FALSE);
387 self->outbuf_size = enc_info.maxOutBufBytes;
388 self->samples_per_frame = enc_info.frameLength;
390 src_caps = gst_caps_new_simple ("audio/mpeg",
391 "mpegversion", G_TYPE_INT, mpegversion,
392 "channels", G_TYPE_INT, GST_AUDIO_INFO_CHANNELS (info),
393 "framed", G_TYPE_BOOLEAN, TRUE,
394 "rate", G_TYPE_INT, GST_AUDIO_INFO_RATE (info), NULL);
398 GstBuffer *codec_data =
399 gst_buffer_new_wrapped (g_memdup (enc_info.confBuf, enc_info.confSize),
401 gst_caps_set_simple (src_caps, "codec_data", GST_TYPE_BUFFER, codec_data,
402 "stream-format", G_TYPE_STRING, "raw", NULL);
403 gst_buffer_unref (codec_data);
404 } else if (transmux == 1) {
405 gst_caps_set_simple (src_caps, "stream-format", G_TYPE_STRING, "adif",
407 } else if (transmux == 2) {
408 gst_caps_set_simple (src_caps, "stream-format", G_TYPE_STRING, "adts",
411 g_assert_not_reached ();
414 gst_codec_utils_aac_caps_set_level_and_profile (src_caps, enc_info.confBuf,
417 ret = gst_audio_encoder_set_output_format (enc, src_caps);
418 gst_caps_unref (src_caps);
424 gst_fdkaacenc_handle_frame (GstAudioEncoder * enc, GstBuffer * inbuf)
426 GstFdkAacEnc *self = GST_FDKAACENC (enc);
427 GstFlowReturn ret = GST_FLOW_OK;
429 GstMapInfo imap, omap;
431 AACENC_BufDesc in_desc = { 0 };
432 AACENC_BufDesc out_desc = { 0 };
433 AACENC_InArgs in_args = { 0 };
434 AACENC_OutArgs out_args = { 0 };
435 gint in_id = IN_AUDIO_DATA, out_id = OUT_BITSTREAM_DATA;
436 gint in_sizes, out_sizes;
437 gint in_el_sizes, out_el_sizes;
440 info = gst_audio_encoder_get_audio_info (enc);
443 if (self->need_reorder) {
444 inbuf = gst_buffer_copy (inbuf);
445 gst_buffer_map (inbuf, &imap, GST_MAP_READWRITE);
446 gst_audio_reorder_channels (imap.data, imap.size,
447 GST_AUDIO_INFO_FORMAT (info), GST_AUDIO_INFO_CHANNELS (info),
448 &GST_AUDIO_INFO_POSITION (info, 0), self->aac_positions);
450 gst_buffer_map (inbuf, &imap, GST_MAP_READ);
453 in_args.numInSamples = imap.size / GST_AUDIO_INFO_BPS (info);
455 in_sizes = imap.size;
456 in_el_sizes = GST_AUDIO_INFO_BPS (info);
459 in_args.numInSamples = -1;
465 /* We unset is_drained even if there's no inbuf. Basically this is a
466 * workaround for aacEncEncode always producing 1024 bytes even without any
467 * input, thus messing up with the base class counting */
468 self->is_drained = FALSE;
470 in_desc.bufferIdentifiers = &in_id;
471 in_desc.bufs = (void *) &imap.data;
472 in_desc.bufSizes = &in_sizes;
473 in_desc.bufElSizes = &in_el_sizes;
475 outbuf = gst_audio_encoder_allocate_output_buffer (enc, self->outbuf_size);
477 ret = GST_FLOW_ERROR;
481 gst_buffer_map (outbuf, &omap, GST_MAP_WRITE);
482 out_sizes = omap.size;
484 out_desc.bufferIdentifiers = &out_id;
485 out_desc.numBufs = 1;
486 out_desc.bufs = (void *) &omap.data;
487 out_desc.bufSizes = &out_sizes;
488 out_desc.bufElSizes = &out_el_sizes;
490 err = aacEncEncode (self->enc, &in_desc, &out_desc, &in_args, &out_args);
491 if (err == AACENC_ENCODE_EOF && !inbuf)
493 else if (err != AACENC_OK) {
494 GST_ERROR_OBJECT (self, "Failed to encode data: %d", err);
495 ret = GST_FLOW_ERROR;
500 gst_buffer_unmap (inbuf, &imap);
501 if (self->need_reorder)
502 gst_buffer_unref (inbuf);
506 if (!out_args.numOutBytes)
509 gst_buffer_unmap (outbuf, &omap);
510 gst_buffer_set_size (outbuf, out_args.numOutBytes);
512 ret = gst_audio_encoder_finish_frame (enc, outbuf, self->samples_per_frame);
517 gst_buffer_unmap (outbuf, &omap);
518 gst_buffer_unref (outbuf);
521 gst_buffer_unmap (inbuf, &imap);
522 if (self->need_reorder)
523 gst_buffer_unref (inbuf);
530 gst_fdkaacenc_flush (GstAudioEncoder * enc)
532 GstFdkAacEnc *self = GST_FDKAACENC (enc);
533 GstAudioInfo *info = gst_audio_encoder_get_audio_info (enc);
535 aacEncClose (&self->enc);
537 self->is_drained = TRUE;
539 if (GST_AUDIO_INFO_IS_VALID (info))
540 gst_fdkaacenc_set_format (enc, info);
544 gst_fdkaacenc_init (GstFdkAacEnc * self)
546 self->bitrate = DEFAULT_BITRATE;
548 self->is_drained = TRUE;
550 gst_audio_encoder_set_drainable (GST_AUDIO_ENCODER (self), TRUE);
554 gst_fdkaacenc_class_init (GstFdkAacEncClass * klass)
556 GObjectClass *object_class = G_OBJECT_CLASS (klass);
557 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
558 GstAudioEncoderClass *base_class = GST_AUDIO_ENCODER_CLASS (klass);
560 object_class->set_property = GST_DEBUG_FUNCPTR (gst_fdkaacenc_set_property);
561 object_class->get_property = GST_DEBUG_FUNCPTR (gst_fdkaacenc_get_property);
563 base_class->start = GST_DEBUG_FUNCPTR (gst_fdkaacenc_start);
564 base_class->stop = GST_DEBUG_FUNCPTR (gst_fdkaacenc_stop);
565 base_class->set_format = GST_DEBUG_FUNCPTR (gst_fdkaacenc_set_format);
566 base_class->getcaps = GST_DEBUG_FUNCPTR (gst_fdkaacenc_get_caps);
567 base_class->handle_frame = GST_DEBUG_FUNCPTR (gst_fdkaacenc_handle_frame);
568 base_class->flush = GST_DEBUG_FUNCPTR (gst_fdkaacenc_flush);
570 g_object_class_install_property (object_class, PROP_BITRATE,
571 g_param_spec_int ("bitrate",
573 "Target Audio Bitrate (0 = fixed value based on "
574 " sample rate and channel count)",
575 0, G_MAXINT, DEFAULT_BITRATE,
576 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
578 gst_element_class_add_static_pad_template (element_class, &sink_template);
579 gst_element_class_add_static_pad_template (element_class, &src_template);
581 gst_element_class_set_static_metadata (element_class, "FDK AAC audio encoder",
582 "Codec/Encoder/Audio", "FDK AAC audio encoder",
583 "Sebastian Dröge <sebastian@centricular.com>");
585 GST_DEBUG_CATEGORY_INIT (gst_fdkaacenc_debug, "fdkaacenc", 0,