1 /* GStreamer Opus Encoder
2 * Copyright (C) <1999> Erik Walthinsen <omega@cse.ogi.edu>
3 * Copyright (C) <2008> Sebastian Dröge <sebastian.droege@collabora.co.uk>
4 * Copyright (C) <2011> Vincent Penquerc'h <vincent.penquerch@collabora.co.uk>
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Library General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Library General Public License for more details.
16 * You should have received a copy of the GNU Library General Public
17 * License along with this library; if not, write to the
18 * Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
19 * Boston, MA 02110-1301, USA.
23 * Based on the speexenc element
27 * SECTION:element-opusenc
29 * @see_also: opusdec, oggmux
31 * This element encodes raw audio to OPUS.
33 * ## Example pipelines
35 * gst-launch-1.0 -v audiotestsrc wave=sine num-buffers=100 ! audioconvert ! opusenc ! oggmux ! filesink location=sine.ogg
37 * Encode a test sine signal to Ogg/OPUS.
50 #include <gst/gsttagsetter.h>
51 #include <gst/audio/audio.h>
52 #include <gst/pbutils/pbutils.h>
53 #include <gst/tag/tag.h>
54 #include <gst/glib-compat-private.h>
56 #include "gstopuselements.h"
57 #include "gstopusheader.h"
58 #include "gstopuscommon.h"
59 #include "gstopusenc.h"
61 GST_DEBUG_CATEGORY_STATIC (opusenc_debug);
62 #define GST_CAT_DEFAULT opusenc_debug
64 /* Some arbitrary bounds beyond which it really doesn't make sense.
65 The spec mentions 6 kb/s to 510 kb/s, so 4000 and 650000 ought to be
66 safe as property bounds. */
67 #define LOWEST_BITRATE 4000
68 #define HIGHEST_BITRATE 650000
70 #define GST_OPUS_ENC_TYPE_BANDWIDTH (gst_opus_enc_bandwidth_get_type())
72 gst_opus_enc_bandwidth_get_type (void)
74 static const GEnumValue values[] = {
75 {OPUS_BANDWIDTH_NARROWBAND, "Narrow band", "narrowband"},
76 {OPUS_BANDWIDTH_MEDIUMBAND, "Medium band", "mediumband"},
77 {OPUS_BANDWIDTH_WIDEBAND, "Wide band", "wideband"},
78 {OPUS_BANDWIDTH_SUPERWIDEBAND, "Super wide band", "superwideband"},
79 {OPUS_BANDWIDTH_FULLBAND, "Full band", "fullband"},
80 {OPUS_AUTO, "Auto", "auto"},
85 if (g_once_init_enter ((gsize *) & id)) {
88 _id = g_enum_register_static ("GstOpusEncBandwidth", values);
90 g_once_init_leave ((gsize *) & id, _id);
96 #define GST_OPUS_ENC_TYPE_FRAME_SIZE (gst_opus_enc_frame_size_get_type())
98 gst_opus_enc_frame_size_get_type (void)
100 static const GEnumValue values[] = {
111 if (g_once_init_enter ((gsize *) & id)) {
114 _id = g_enum_register_static ("GstOpusEncFrameSize", values);
116 g_once_init_leave ((gsize *) & id, _id);
122 #define GST_OPUS_ENC_TYPE_AUDIO_TYPE (gst_opus_enc_audio_type_get_type())
124 gst_opus_enc_audio_type_get_type (void)
126 static const GEnumValue values[] = {
127 {OPUS_APPLICATION_AUDIO, "Generic audio", "generic"},
128 {OPUS_APPLICATION_VOIP, "Voice", "voice"},
129 {OPUS_APPLICATION_RESTRICTED_LOWDELAY, "Restricted low delay",
130 "restricted-lowdelay"},
135 if (g_once_init_enter ((gsize *) & id)) {
138 _id = g_enum_register_static ("GstOpusEncAudioType", values);
140 g_once_init_leave ((gsize *) & id, _id);
146 #define GST_OPUS_ENC_TYPE_BITRATE_TYPE (gst_opus_enc_bitrate_type_get_type())
148 gst_opus_enc_bitrate_type_get_type (void)
150 static const GEnumValue values[] = {
151 {BITRATE_TYPE_CBR, "CBR", "cbr"},
152 {BITRATE_TYPE_VBR, "VBR", "vbr"},
153 {BITRATE_TYPE_CONSTRAINED_VBR, "Constrained VBR", "constrained-vbr"},
158 if (g_once_init_enter ((gsize *) & id)) {
161 _id = g_enum_register_static ("GstOpusEncBitrateType", values);
163 g_once_init_leave ((gsize *) & id, _id);
169 #define FORMAT_STR GST_AUDIO_NE(S16)
170 static GstStaticPadTemplate sink_factory = GST_STATIC_PAD_TEMPLATE ("sink",
173 GST_STATIC_CAPS ("audio/x-raw, "
174 "format = (string) " FORMAT_STR ", "
175 "layout = (string) interleaved, "
176 "rate = (int) 48000, "
177 "channels = (int) [ 1, 8 ]; "
179 "format = (string) " FORMAT_STR ", "
180 "layout = (string) interleaved, "
181 "rate = (int) { 8000, 12000, 16000, 24000 }, "
182 "channels = (int) [ 1, 8 ] ")
185 static GstStaticPadTemplate src_factory = GST_STATIC_PAD_TEMPLATE ("src",
188 GST_STATIC_CAPS ("audio/x-opus")
191 #define DEFAULT_AUDIO TRUE
192 #define DEFAULT_AUDIO_TYPE OPUS_APPLICATION_AUDIO
193 #define DEFAULT_BITRATE 64000
194 #define DEFAULT_BANDWIDTH OPUS_BANDWIDTH_FULLBAND
195 #define DEFAULT_FRAMESIZE 20
196 #define DEFAULT_CBR TRUE
197 #define DEFAULT_CONSTRAINED_VBR TRUE
198 #define DEFAULT_BITRATE_TYPE BITRATE_TYPE_CONSTRAINED_VBR
199 #define DEFAULT_COMPLEXITY 10
200 #define DEFAULT_INBAND_FEC FALSE
201 #define DEFAULT_DTX FALSE
202 #define DEFAULT_PACKET_LOSS_PERCENT 0
203 #define DEFAULT_MAX_PAYLOAD_SIZE 4000
216 PROP_PACKET_LOSS_PERCENT,
217 PROP_MAX_PAYLOAD_SIZE
220 static void gst_opus_enc_finalize (GObject * object);
222 static gboolean gst_opus_enc_sink_event (GstAudioEncoder * benc,
224 static GstCaps *gst_opus_enc_sink_getcaps (GstAudioEncoder * benc,
226 static gboolean gst_opus_enc_setup (GstOpusEnc * enc);
228 static void gst_opus_enc_get_property (GObject * object, guint prop_id,
229 GValue * value, GParamSpec * pspec);
230 static void gst_opus_enc_set_property (GObject * object, guint prop_id,
231 const GValue * value, GParamSpec * pspec);
233 static void gst_opus_enc_set_tags (GstOpusEnc * enc);
234 static gboolean gst_opus_enc_start (GstAudioEncoder * benc);
235 static gboolean gst_opus_enc_stop (GstAudioEncoder * benc);
236 static gboolean gst_opus_enc_set_format (GstAudioEncoder * benc,
237 GstAudioInfo * info);
238 static GstFlowReturn gst_opus_enc_handle_frame (GstAudioEncoder * benc,
240 static gint64 gst_opus_enc_get_latency (GstOpusEnc * enc);
242 static GstFlowReturn gst_opus_enc_encode (GstOpusEnc * enc, GstBuffer * buffer);
244 #define gst_opus_enc_parent_class parent_class
245 G_DEFINE_TYPE_WITH_CODE (GstOpusEnc, gst_opus_enc, GST_TYPE_AUDIO_ENCODER,
246 G_IMPLEMENT_INTERFACE (GST_TYPE_TAG_SETTER, NULL);
247 G_IMPLEMENT_INTERFACE (GST_TYPE_PRESET, NULL));
248 GST_ELEMENT_REGISTER_DEFINE_WITH_CODE (opusenc, "opusenc",
249 GST_RANK_PRIMARY, GST_TYPE_OPUS_ENC, opus_element_init (plugin));
252 gst_opus_enc_set_tags (GstOpusEnc * enc)
256 /* create a taglist and add a bitrate tag to it */
257 taglist = gst_tag_list_new_empty ();
258 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
259 GST_TAG_BITRATE, enc->bitrate, NULL);
261 gst_audio_encoder_merge_tags (GST_AUDIO_ENCODER (enc), taglist,
262 GST_TAG_MERGE_REPLACE);
264 gst_tag_list_unref (taglist);
268 gst_opus_enc_class_init (GstOpusEncClass * klass)
270 GObjectClass *gobject_class;
271 GstAudioEncoderClass *base_class;
272 GstElementClass *gstelement_class;
274 gobject_class = (GObjectClass *) klass;
275 base_class = (GstAudioEncoderClass *) klass;
276 gstelement_class = (GstElementClass *) klass;
278 gobject_class->set_property = gst_opus_enc_set_property;
279 gobject_class->get_property = gst_opus_enc_get_property;
281 gst_element_class_add_static_pad_template (gstelement_class, &src_factory);
282 gst_element_class_add_static_pad_template (gstelement_class, &sink_factory);
283 gst_element_class_set_static_metadata (gstelement_class, "Opus audio encoder",
284 "Codec/Encoder/Audio",
285 "Encodes audio in Opus format",
286 "Vincent Penquerc'h <vincent.penquerch@collabora.co.uk>");
288 base_class->start = GST_DEBUG_FUNCPTR (gst_opus_enc_start);
289 base_class->stop = GST_DEBUG_FUNCPTR (gst_opus_enc_stop);
290 base_class->set_format = GST_DEBUG_FUNCPTR (gst_opus_enc_set_format);
291 base_class->handle_frame = GST_DEBUG_FUNCPTR (gst_opus_enc_handle_frame);
292 base_class->sink_event = GST_DEBUG_FUNCPTR (gst_opus_enc_sink_event);
293 base_class->getcaps = GST_DEBUG_FUNCPTR (gst_opus_enc_sink_getcaps);
295 g_object_class_install_property (gobject_class, PROP_AUDIO_TYPE,
296 g_param_spec_enum ("audio-type", "What type of audio to optimize for",
297 "What type of audio to optimize for", GST_OPUS_ENC_TYPE_AUDIO_TYPE,
298 DEFAULT_AUDIO_TYPE, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
299 g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_BITRATE,
300 g_param_spec_int ("bitrate", "Encoding Bit-rate",
301 "Specify an encoding bit-rate (in bps).", LOWEST_BITRATE,
302 HIGHEST_BITRATE, DEFAULT_BITRATE,
303 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
304 GST_PARAM_MUTABLE_PLAYING));
305 g_object_class_install_property (gobject_class, PROP_BANDWIDTH,
306 g_param_spec_enum ("bandwidth", "Band Width", "Audio Band Width",
307 GST_OPUS_ENC_TYPE_BANDWIDTH, DEFAULT_BANDWIDTH,
308 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
309 GST_PARAM_MUTABLE_PLAYING));
310 g_object_class_install_property (gobject_class, PROP_FRAME_SIZE,
311 g_param_spec_enum ("frame-size", "Frame Size",
312 "The duration of an audio frame, in ms", GST_OPUS_ENC_TYPE_FRAME_SIZE,
314 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
315 GST_PARAM_MUTABLE_PLAYING));
316 g_object_class_install_property (gobject_class, PROP_BITRATE_TYPE,
317 g_param_spec_enum ("bitrate-type", "Bitrate type", "Bitrate type",
318 GST_OPUS_ENC_TYPE_BITRATE_TYPE, DEFAULT_BITRATE_TYPE,
319 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
320 GST_PARAM_MUTABLE_PLAYING));
321 g_object_class_install_property (gobject_class, PROP_COMPLEXITY,
322 g_param_spec_int ("complexity", "Complexity", "Complexity", 0, 10,
324 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
325 GST_PARAM_MUTABLE_PLAYING));
326 g_object_class_install_property (gobject_class, PROP_INBAND_FEC,
327 g_param_spec_boolean ("inband-fec", "In-band FEC",
328 "Enable forward error correction", DEFAULT_INBAND_FEC,
329 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
330 GST_PARAM_MUTABLE_PLAYING));
331 g_object_class_install_property (gobject_class, PROP_DTX,
332 g_param_spec_boolean ("dtx", "DTX", "DTX", DEFAULT_DTX,
333 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
334 GST_PARAM_MUTABLE_PLAYING));
335 g_object_class_install_property (G_OBJECT_CLASS (klass),
336 PROP_PACKET_LOSS_PERCENT, g_param_spec_int ("packet-loss-percentage",
337 "Loss percentage", "Packet loss percentage", 0, 100,
338 DEFAULT_PACKET_LOSS_PERCENT,
339 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
340 GST_PARAM_MUTABLE_PLAYING));
341 g_object_class_install_property (G_OBJECT_CLASS (klass),
342 PROP_MAX_PAYLOAD_SIZE, g_param_spec_uint ("max-payload-size",
343 "Max payload size", "Maximum payload size in bytes", 2, 4000,
344 DEFAULT_MAX_PAYLOAD_SIZE,
345 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS |
346 GST_PARAM_MUTABLE_PLAYING));
348 gobject_class->finalize = GST_DEBUG_FUNCPTR (gst_opus_enc_finalize);
350 GST_DEBUG_CATEGORY_INIT (opusenc_debug, "opusenc", 0, "Opus encoder");
352 gst_type_mark_as_plugin_api (GST_OPUS_ENC_TYPE_AUDIO_TYPE, 0);
353 gst_type_mark_as_plugin_api (GST_OPUS_ENC_TYPE_BANDWIDTH, 0);
354 gst_type_mark_as_plugin_api (GST_OPUS_ENC_TYPE_BITRATE_TYPE, 0);
355 gst_type_mark_as_plugin_api (GST_OPUS_ENC_TYPE_FRAME_SIZE, 0);
359 gst_opus_enc_finalize (GObject * object)
363 enc = GST_OPUS_ENC (object);
365 g_mutex_clear (&enc->property_lock);
367 G_OBJECT_CLASS (parent_class)->finalize (object);
371 gst_opus_enc_init (GstOpusEnc * enc)
373 GST_DEBUG_OBJECT (enc, "init");
375 GST_PAD_SET_ACCEPT_TEMPLATE (GST_AUDIO_ENCODER_SINK_PAD (enc));
377 g_mutex_init (&enc->property_lock);
379 enc->n_channels = -1;
380 enc->sample_rate = -1;
381 enc->frame_samples = 0;
382 enc->unpositioned = FALSE;
384 enc->bitrate = DEFAULT_BITRATE;
385 enc->bandwidth = DEFAULT_BANDWIDTH;
386 enc->frame_size = DEFAULT_FRAMESIZE;
387 enc->bitrate_type = DEFAULT_BITRATE_TYPE;
388 enc->complexity = DEFAULT_COMPLEXITY;
389 enc->inband_fec = DEFAULT_INBAND_FEC;
390 enc->dtx = DEFAULT_DTX;
391 enc->packet_loss_percentage = DEFAULT_PACKET_LOSS_PERCENT;
392 enc->max_payload_size = DEFAULT_MAX_PAYLOAD_SIZE;
393 enc->audio_type = DEFAULT_AUDIO_TYPE;
397 gst_opus_enc_start (GstAudioEncoder * benc)
399 GstOpusEnc *enc = GST_OPUS_ENC (benc);
401 GST_DEBUG_OBJECT (enc, "start");
402 enc->encoded_samples = 0;
403 enc->consumed_samples = 0;
409 gst_opus_enc_stop (GstAudioEncoder * benc)
411 GstOpusEnc *enc = GST_OPUS_ENC (benc);
413 GST_DEBUG_OBJECT (enc, "stop");
415 opus_multistream_encoder_destroy (enc->state);
418 gst_tag_setter_reset_tags (GST_TAG_SETTER (enc));
424 gst_opus_enc_get_latency (GstOpusEnc * enc)
426 gint64 latency = gst_util_uint64_scale (enc->frame_samples, GST_SECOND,
428 GST_DEBUG_OBJECT (enc, "Latency: %" GST_TIME_FORMAT, GST_TIME_ARGS (latency));
433 gst_opus_enc_setup_base_class (GstOpusEnc * enc, GstAudioEncoder * benc)
435 gst_audio_encoder_set_latency (benc,
436 gst_opus_enc_get_latency (enc), gst_opus_enc_get_latency (enc));
437 gst_audio_encoder_set_frame_samples_min (benc, enc->frame_samples);
438 gst_audio_encoder_set_frame_samples_max (benc, enc->frame_samples);
439 gst_audio_encoder_set_frame_max (benc, 1);
443 gst_opus_enc_get_frame_samples (GstOpusEnc * enc)
445 gint frame_samples = 0;
446 switch (enc->frame_size) {
448 frame_samples = enc->sample_rate / 400;
451 frame_samples = enc->sample_rate / 200;
454 frame_samples = enc->sample_rate / 100;
457 frame_samples = enc->sample_rate / 50;
460 frame_samples = enc->sample_rate / 25;
463 frame_samples = 3 * enc->sample_rate / 50;
466 GST_WARNING_OBJECT (enc, "Unsupported frame size: %d", enc->frame_size);
470 return frame_samples;
474 gst_opus_enc_setup_trivial_mapping (GstOpusEnc * enc, guint8 mapping[256])
478 for (n = 0; n < 255; ++n)
483 gst_opus_enc_find_channel_position (GstOpusEnc * enc, const GstAudioInfo * info,
484 GstAudioChannelPosition position)
487 for (n = 0; n < enc->n_channels; ++n) {
488 if (GST_AUDIO_INFO_POSITION (info, n) == position) {
496 gst_opus_enc_find_channel_position_in_vorbis_order (GstOpusEnc * enc,
497 GstAudioChannelPosition position)
501 for (c = 0; c < enc->n_channels; ++c) {
502 if (gst_opus_channel_positions[enc->n_channels - 1][c] == position) {
503 GST_INFO_OBJECT (enc,
504 "Channel position %s maps to index %d in Vorbis order",
505 gst_opus_channel_names[position], c);
509 GST_WARNING_OBJECT (enc,
510 "Channel position %s is not representable in Vorbis order",
511 gst_opus_channel_names[position]);
516 gst_opus_enc_setup_channel_mappings (GstOpusEnc * enc,
517 const GstAudioInfo * info)
519 #define MAPS(idx,pos) (GST_AUDIO_INFO_POSITION (info, (idx)) == GST_AUDIO_CHANNEL_POSITION_##pos)
523 GST_DEBUG_OBJECT (enc, "Setting up channel mapping for %d channels",
526 /* Start by setting up a default trivial mapping */
527 enc->n_stereo_streams = 0;
528 gst_opus_enc_setup_trivial_mapping (enc, enc->encoding_channel_mapping);
529 gst_opus_enc_setup_trivial_mapping (enc, enc->decoding_channel_mapping);
531 /* For one channel, use the basic RTP mapping */
532 if (enc->n_channels == 1 && !enc->unpositioned) {
533 GST_INFO_OBJECT (enc, "Mono, trivial RTP mapping");
534 enc->channel_mapping_family = 0;
535 /* implicit mapping for family 0 */
539 /* For two channels, use the basic RTP mapping if the channels are
540 mapped as left/right. */
541 if (enc->n_channels == 2 && !enc->unpositioned) {
542 GST_INFO_OBJECT (enc, "Stereo, trivial RTP mapping");
543 enc->channel_mapping_family = 0;
544 enc->n_stereo_streams = 1;
545 /* implicit mapping for family 0 */
549 /* For channels between 3 and 8, we use the Vorbis mapping if we can
550 find a permutation that matches it. Mono and stereo will have been taken
551 care of earlier, but this code also handles it. There are two mappings.
552 One maps the input channels to an ordering which has the natural pairs
553 first so they can benefit from the Opus stereo channel coupling, and the
554 other maps this ordering to the Vorbis ordering. */
555 if (enc->n_channels >= 3 && enc->n_channels <= 8 && !enc->unpositioned) {
556 int c0, c1, c0v, c1v;
558 gboolean positions_done[256];
559 static const GstAudioChannelPosition pairs[][2] = {
560 {GST_AUDIO_CHANNEL_POSITION_FRONT_LEFT,
561 GST_AUDIO_CHANNEL_POSITION_FRONT_RIGHT},
562 {GST_AUDIO_CHANNEL_POSITION_REAR_LEFT,
563 GST_AUDIO_CHANNEL_POSITION_REAR_RIGHT},
564 {GST_AUDIO_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER,
565 GST_AUDIO_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER},
566 {GST_AUDIO_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER,
567 GST_AUDIO_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER},
568 {GST_AUDIO_CHANNEL_POSITION_SIDE_LEFT,
569 GST_AUDIO_CHANNEL_POSITION_SIDE_RIGHT},
570 {GST_AUDIO_CHANNEL_POSITION_FRONT_CENTER,
571 GST_AUDIO_CHANNEL_POSITION_REAR_CENTER},
575 GST_DEBUG_OBJECT (enc,
576 "In range for the Vorbis mapping, building channel mapping tables");
578 enc->n_stereo_streams = 0;
580 for (n = 0; n < 256; ++n)
581 positions_done[n] = FALSE;
583 /* First, find any natural pairs, and move them to the front */
584 for (pair = 0; pair < G_N_ELEMENTS (pairs); ++pair) {
585 GstAudioChannelPosition p0 = pairs[pair][0];
586 GstAudioChannelPosition p1 = pairs[pair][1];
587 c0 = gst_opus_enc_find_channel_position (enc, info, p0);
588 c1 = gst_opus_enc_find_channel_position (enc, info, p1);
589 if (c0 >= 0 && c1 >= 0) {
590 /* We found a natural pair */
591 GST_DEBUG_OBJECT (enc, "Natural pair '%s/%s' found at %d %d",
592 gst_opus_channel_names[p0], gst_opus_channel_names[p1], c0, c1);
593 /* Find where they map in Vorbis order */
594 c0v = gst_opus_enc_find_channel_position_in_vorbis_order (enc, p0);
595 c1v = gst_opus_enc_find_channel_position_in_vorbis_order (enc, p1);
596 if (c0v < 0 || c1v < 0) {
597 GST_WARNING_OBJECT (enc,
598 "Cannot map channel positions to Vorbis order, using unknown mapping");
599 enc->channel_mapping_family = 255;
600 enc->n_stereo_streams = 0;
604 enc->encoding_channel_mapping[mapped] = c0;
605 enc->encoding_channel_mapping[mapped + 1] = c1;
606 enc->decoding_channel_mapping[c0v] = mapped;
607 enc->decoding_channel_mapping[c1v] = mapped + 1;
608 enc->n_stereo_streams++;
610 positions_done[p0] = positions_done[p1] = TRUE;
614 /* Now add all other input channels as mono streams */
615 for (n = 0; n < enc->n_channels; ++n) {
616 GstAudioChannelPosition position = GST_AUDIO_INFO_POSITION (info, n);
618 /* if we already mapped it while searching for pairs, nothing else
620 if (!positions_done[position]) {
622 GST_DEBUG_OBJECT (enc, "Channel position %s is not mapped yet, adding",
623 gst_opus_channel_names[position]);
624 cv = gst_opus_enc_find_channel_position_in_vorbis_order (enc, position);
626 g_assert_not_reached ();
627 enc->encoding_channel_mapping[mapped] = n;
628 enc->decoding_channel_mapping[cv] = mapped;
633 #ifndef GST_DISABLE_GST_DEBUG
634 GST_INFO_OBJECT (enc,
635 "Mapping tables built: %d channels, %d stereo streams", enc->n_channels,
636 enc->n_stereo_streams);
637 gst_opus_common_log_channel_mapping_table (GST_ELEMENT (enc), opusenc_debug,
638 "Encoding mapping table", enc->n_channels,
639 enc->encoding_channel_mapping);
640 gst_opus_common_log_channel_mapping_table (GST_ELEMENT (enc), opusenc_debug,
641 "Decoding mapping table", enc->n_channels,
642 enc->decoding_channel_mapping);
645 enc->channel_mapping_family = 1;
649 /* More than 8 channels, if future mappings are added for those */
651 /* For other cases, we use undefined, with the default trivial mapping
652 and all mono streams */
653 if (!enc->unpositioned)
654 GST_WARNING_OBJECT (enc, "Unknown mapping");
656 GST_INFO_OBJECT (enc, "Unpositioned mapping, all channels mono");
658 enc->channel_mapping_family = 255;
659 enc->n_stereo_streams = 0;
665 gst_opus_enc_set_format (GstAudioEncoder * benc, GstAudioInfo * info)
669 enc = GST_OPUS_ENC (benc);
671 g_mutex_lock (&enc->property_lock);
673 enc->n_channels = GST_AUDIO_INFO_CHANNELS (info);
674 enc->unpositioned = GST_AUDIO_INFO_IS_UNPOSITIONED (info);
675 enc->sample_rate = GST_AUDIO_INFO_RATE (info);
676 gst_opus_enc_setup_channel_mappings (enc, info);
677 GST_DEBUG_OBJECT (benc, "Setup with %d channels, %d Hz", enc->n_channels,
680 /* handle reconfigure */
682 opus_multistream_encoder_destroy (enc->state);
685 if (!gst_opus_enc_setup (enc)) {
686 g_mutex_unlock (&enc->property_lock);
690 /* update the tags */
691 gst_opus_enc_set_tags (enc);
693 enc->frame_samples = gst_opus_enc_get_frame_samples (enc);
695 /* feedback to base class */
696 gst_opus_enc_setup_base_class (enc, benc);
698 g_mutex_unlock (&enc->property_lock);
704 gst_opus_enc_setup (GstOpusEnc * enc)
710 const GstTagList *tags;
711 GstTagList *empty_tags = NULL;
712 GstBuffer *header, *comments;
714 #ifndef GST_DISABLE_GST_DEBUG
715 GST_DEBUG_OBJECT (enc,
716 "setup: %d Hz, %d channels, %d stereo streams, family %d",
717 enc->sample_rate, enc->n_channels, enc->n_stereo_streams,
718 enc->channel_mapping_family);
719 GST_INFO_OBJECT (enc, "Mapping tables built: %d channels, %d stereo streams",
720 enc->n_channels, enc->n_stereo_streams);
721 gst_opus_common_log_channel_mapping_table (GST_ELEMENT (enc), opusenc_debug,
722 "Encoding mapping table", enc->n_channels, enc->encoding_channel_mapping);
723 gst_opus_common_log_channel_mapping_table (GST_ELEMENT (enc), opusenc_debug,
724 "Decoding mapping table", enc->n_channels, enc->decoding_channel_mapping);
727 enc->state = opus_multistream_encoder_create (enc->sample_rate,
728 enc->n_channels, enc->n_channels - enc->n_stereo_streams,
729 enc->n_stereo_streams, enc->encoding_channel_mapping,
730 enc->audio_type, &error);
731 if (!enc->state || error != OPUS_OK)
732 goto encoder_creation_failed;
734 opus_multistream_encoder_ctl (enc->state, OPUS_SET_BITRATE (enc->bitrate), 0);
735 opus_multistream_encoder_ctl (enc->state, OPUS_SET_BANDWIDTH (enc->bandwidth),
737 opus_multistream_encoder_ctl (enc->state,
738 OPUS_SET_VBR (enc->bitrate_type != BITRATE_TYPE_CBR), 0);
739 opus_multistream_encoder_ctl (enc->state,
740 OPUS_SET_VBR_CONSTRAINT (enc->bitrate_type ==
741 BITRATE_TYPE_CONSTRAINED_VBR), 0);
742 opus_multistream_encoder_ctl (enc->state,
743 OPUS_SET_COMPLEXITY (enc->complexity), 0);
744 opus_multistream_encoder_ctl (enc->state,
745 OPUS_SET_INBAND_FEC (enc->inband_fec), 0);
746 opus_multistream_encoder_ctl (enc->state, OPUS_SET_DTX (enc->dtx), 0);
747 opus_multistream_encoder_ctl (enc->state,
748 OPUS_SET_PACKET_LOSS_PERC (enc->packet_loss_percentage), 0);
750 opus_multistream_encoder_ctl (enc->state, OPUS_GET_LOOKAHEAD (&lookahead), 0);
752 GST_LOG_OBJECT (enc, "we have frame size %d, lookahead %d", enc->frame_size,
755 /* lookahead is samples, the Opus header wants it in 48kHz samples */
756 lookahead = lookahead * 48000 / enc->sample_rate;
757 enc->lookahead = enc->pending_lookahead = lookahead;
759 header = gst_codec_utils_opus_create_header (enc->sample_rate,
760 enc->n_channels, enc->channel_mapping_family,
761 enc->n_channels - enc->n_stereo_streams, enc->n_stereo_streams,
762 enc->decoding_channel_mapping, lookahead, 0);
763 tags = gst_tag_setter_get_tag_list (GST_TAG_SETTER (enc));
765 tags = empty_tags = gst_tag_list_new_empty ();
767 gst_tag_list_to_vorbiscomment_buffer (tags, (const guint8 *) "OpusTags",
768 8, "Encoded with GStreamer opusenc");
769 caps = gst_codec_utils_opus_create_caps_from_header (header, comments);
771 gst_tag_list_unref (empty_tags);
772 gst_buffer_unref (header);
773 gst_buffer_unref (comments);
775 /* negotiate with these caps */
776 GST_DEBUG_OBJECT (enc, "here are the caps: %" GST_PTR_FORMAT, caps);
778 ret = gst_audio_encoder_set_output_format (GST_AUDIO_ENCODER (enc), caps);
779 gst_caps_unref (caps);
783 encoder_creation_failed:
784 GST_ERROR_OBJECT (enc, "Encoder creation failed");
789 gst_opus_enc_sink_event (GstAudioEncoder * benc, GstEvent * event)
793 enc = GST_OPUS_ENC (benc);
795 GST_DEBUG_OBJECT (enc, "sink event: %s", GST_EVENT_TYPE_NAME (event));
796 switch (GST_EVENT_TYPE (event)) {
800 GstTagSetter *setter = GST_TAG_SETTER (enc);
801 const GstTagMergeMode mode = gst_tag_setter_get_tag_merge_mode (setter);
803 gst_event_parse_tag (event, &list);
804 gst_tag_setter_merge_tags (setter, list, mode);
807 case GST_EVENT_SEGMENT:
808 enc->encoded_samples = 0;
809 enc->consumed_samples = 0;
816 return GST_AUDIO_ENCODER_CLASS (parent_class)->sink_event (benc, event);
820 gst_opus_enc_get_sink_template_caps (void)
822 static gsize init = 0;
823 static GstCaps *caps = NULL;
825 if (g_once_init_enter (&init)) {
826 GValue rate_array = G_VALUE_INIT;
827 GValue v = G_VALUE_INIT;
828 GstStructure *s1, *s2, *s;
831 caps = gst_caps_new_empty ();
833 /* The caps is cached */
834 GST_MINI_OBJECT_FLAG_SET (caps, GST_MINI_OBJECT_FLAG_MAY_BE_LEAKED);
836 /* Generate our two template structures */
837 g_value_init (&rate_array, GST_TYPE_LIST);
838 g_value_init (&v, G_TYPE_INT);
839 g_value_set_int (&v, 8000);
840 gst_value_list_append_value (&rate_array, &v);
841 g_value_set_int (&v, 12000);
842 gst_value_list_append_value (&rate_array, &v);
843 g_value_set_int (&v, 16000);
844 gst_value_list_append_value (&rate_array, &v);
845 g_value_set_int (&v, 24000);
846 gst_value_list_append_value (&rate_array, &v);
848 s1 = gst_structure_new ("audio/x-raw",
849 "format", G_TYPE_STRING, GST_AUDIO_NE (S16),
850 "layout", G_TYPE_STRING, "interleaved",
851 "rate", G_TYPE_INT, 48000, NULL);
852 s2 = gst_structure_new ("audio/x-raw",
853 "format", G_TYPE_STRING, GST_AUDIO_NE (S16),
854 "layout", G_TYPE_STRING, "interleaved", NULL);
855 gst_structure_set_value (s2, "rate", &rate_array);
856 g_value_unset (&rate_array);
860 s = gst_structure_copy (s1);
861 gst_structure_set (s, "channels", G_TYPE_INT, 1, NULL);
862 gst_caps_append_structure (caps, s);
864 s = gst_structure_copy (s2);
865 gst_structure_set (s, "channels", G_TYPE_INT, 1, NULL);
866 gst_caps_append_structure (caps, s);
868 /* Stereo and further */
869 for (i = 2; i <= 8; i++) {
870 guint64 channel_mask = 0;
871 const GstAudioChannelPosition *pos = gst_opus_channel_positions[i - 1];
873 for (c = 0; c < i; c++) {
874 channel_mask |= G_GUINT64_CONSTANT (1) << pos[c];
877 s = gst_structure_copy (s1);
878 gst_structure_set (s, "channels", G_TYPE_INT, i, "channel-mask",
879 GST_TYPE_BITMASK, channel_mask, NULL);
880 gst_caps_append_structure (caps, s);
882 s = gst_structure_copy (s2);
883 gst_structure_set (s, "channels", G_TYPE_INT, i, "channel-mask",
884 GST_TYPE_BITMASK, channel_mask, NULL);
885 gst_caps_append_structure (caps, s);
887 /* We also allow unpositioned channels, input will be
888 * treated as a set of individual mono channels */
889 s = gst_structure_copy (s2);
890 gst_structure_set (s, "channels", G_TYPE_INT, i, "channel-mask",
891 GST_TYPE_BITMASK, G_GUINT64_CONSTANT (0), NULL);
892 gst_caps_append_structure (caps, s);
894 s = gst_structure_copy (s1);
895 gst_structure_set (s, "channels", G_TYPE_INT, i, "channel-mask",
896 GST_TYPE_BITMASK, G_GUINT64_CONSTANT (0), NULL);
897 gst_caps_append_structure (caps, s);
900 gst_structure_free (s1);
901 gst_structure_free (s2);
903 g_once_init_leave (&init, 1);
910 gst_opus_enc_sink_getcaps (GstAudioEncoder * benc, GstCaps * filter)
915 enc = GST_OPUS_ENC (benc);
917 GST_DEBUG_OBJECT (enc, "sink getcaps");
919 caps = gst_opus_enc_get_sink_template_caps ();
920 caps = gst_audio_encoder_proxy_getcaps (benc, caps, filter);
922 GST_DEBUG_OBJECT (enc, "Returning caps: %" GST_PTR_FORMAT, caps);
928 gst_opus_enc_encode (GstOpusEnc * enc, GstBuffer * buf)
930 guint8 *bdata = NULL, *data, *mdata = NULL;
933 gint ret = GST_FLOW_OK;
938 guint64 trim_start = 0, trim_end = 0;
940 guint max_payload_size;
941 gint frame_samples, input_samples, output_samples;
943 g_mutex_lock (&enc->property_lock);
945 bytes = enc->frame_samples * enc->n_channels * 2;
946 max_payload_size = enc->max_payload_size;
947 frame_samples = input_samples = enc->frame_samples;
949 g_mutex_unlock (&enc->property_lock);
951 if (G_LIKELY (buf)) {
952 gst_buffer_map (buf, &map, GST_MAP_READ);
956 if (G_UNLIKELY (bsize % bytes)) {
959 GST_DEBUG_OBJECT (enc, "draining; adding silence samples");
960 g_assert (bsize < bytes);
962 input_samples = bsize / (enc->n_channels * 2);
964 (enc->encoded_samples + frame_samples) - (enc->consumed_samples +
967 GST_DEBUG_OBJECT (enc,
968 "%" G_GINT64_FORMAT " extra samples of padding in this frame",
970 output_samples = frame_samples - diff;
971 trim_end = diff * 48000 / enc->sample_rate;
973 GST_DEBUG_OBJECT (enc,
974 "Need to add %" G_GINT64_FORMAT " extra samples in the next frame",
976 output_samples = frame_samples;
979 size = ((bsize / bytes) + 1) * bytes;
980 mdata = g_malloc0 (size);
981 /* FIXME: Instead of silence, use LPC with the last real samples.
982 * Otherwise we will create a discontinuity here, which will distort the
983 * last few encoded samples
985 memcpy (mdata, bdata, bsize);
991 /* Adjust for lookahead here */
992 if (enc->pending_lookahead) {
993 guint scaled_lookahead =
994 enc->pending_lookahead * enc->sample_rate / 48000;
996 if (input_samples > scaled_lookahead) {
997 output_samples = input_samples - scaled_lookahead;
998 trim_start = enc->pending_lookahead;
999 enc->pending_lookahead = 0;
1001 trim_start = ((guint64) input_samples) * 48000 / enc->sample_rate;
1002 enc->pending_lookahead -= trim_start;
1006 output_samples = input_samples;
1010 if (enc->encoded_samples < enc->consumed_samples) {
1011 /* FIXME: Instead of silence, use LPC with the last real samples.
1012 * Otherwise we will create a discontinuity here, which will distort the
1013 * last few encoded samples
1015 data = mdata = g_malloc0 (bytes);
1017 output_samples = enc->consumed_samples - enc->encoded_samples;
1019 GST_DEBUG_OBJECT (enc, "draining %d samples", output_samples);
1021 ((guint64) frame_samples - output_samples) * 48000 / enc->sample_rate;
1022 } else if (enc->encoded_samples == enc->consumed_samples) {
1023 GST_DEBUG_OBJECT (enc, "nothing to drain");
1026 g_assert_not_reached ();
1031 g_assert (size == bytes);
1034 gst_audio_encoder_allocate_output_buffer (GST_AUDIO_ENCODER (enc),
1035 max_payload_size * enc->n_channels);
1039 GST_DEBUG_OBJECT (enc, "encoding %d samples (%d bytes)",
1040 frame_samples, (int) bytes);
1042 if (trim_start || trim_end) {
1043 GST_DEBUG_OBJECT (enc,
1044 "Adding trim-start %" G_GUINT64_FORMAT " trim-end %" G_GUINT64_FORMAT,
1045 trim_start, trim_end);
1046 gst_buffer_add_audio_clipping_meta (outbuf, GST_FORMAT_DEFAULT, trim_start,
1050 gst_buffer_map (outbuf, &omap, GST_MAP_WRITE);
1053 opus_multistream_encode (enc->state, (const gint16 *) data,
1054 frame_samples, omap.data, max_payload_size * enc->n_channels);
1056 gst_buffer_unmap (outbuf, &omap);
1059 GST_ELEMENT_ERROR (enc, STREAM, ENCODE, (NULL),
1060 ("Encoding failed (%d): %s", outsize, opus_strerror (outsize)));
1061 ret = GST_FLOW_ERROR;
1063 } else if (outsize > max_payload_size) {
1064 GST_ELEMENT_ERROR (enc, STREAM, ENCODE, (NULL),
1065 ("Encoded size %d is higher than max payload size (%d bytes)",
1066 outsize, max_payload_size));
1067 ret = GST_FLOW_ERROR;
1071 GST_DEBUG_OBJECT (enc, "Output packet is %u bytes", outsize);
1072 gst_buffer_set_size (outbuf, outsize);
1076 gst_audio_encoder_finish_frame (GST_AUDIO_ENCODER (enc), outbuf,
1078 enc->encoded_samples += output_samples;
1079 enc->consumed_samples += input_samples;
1084 gst_buffer_unmap (buf, &map);
1091 static GstFlowReturn
1092 gst_opus_enc_handle_frame (GstAudioEncoder * benc, GstBuffer * buf)
1095 GstFlowReturn ret = GST_FLOW_OK;
1097 enc = GST_OPUS_ENC (benc);
1098 GST_DEBUG_OBJECT (enc, "handle_frame");
1099 GST_DEBUG_OBJECT (enc, "received buffer %p of %" G_GSIZE_FORMAT " bytes", buf,
1100 buf ? gst_buffer_get_size (buf) : 0);
1102 ret = gst_opus_enc_encode (enc, buf);
1108 gst_opus_enc_get_property (GObject * object, guint prop_id, GValue * value,
1113 enc = GST_OPUS_ENC (object);
1115 g_mutex_lock (&enc->property_lock);
1118 case PROP_AUDIO_TYPE:
1119 g_value_set_enum (value, enc->audio_type);
1122 g_value_set_int (value, enc->bitrate);
1124 case PROP_BANDWIDTH:
1125 g_value_set_enum (value, enc->bandwidth);
1127 case PROP_FRAME_SIZE:
1128 g_value_set_enum (value, enc->frame_size);
1130 case PROP_BITRATE_TYPE:
1131 g_value_set_enum (value, enc->bitrate_type);
1133 case PROP_COMPLEXITY:
1134 g_value_set_int (value, enc->complexity);
1136 case PROP_INBAND_FEC:
1137 g_value_set_boolean (value, enc->inband_fec);
1140 g_value_set_boolean (value, enc->dtx);
1142 case PROP_PACKET_LOSS_PERCENT:
1143 g_value_set_int (value, enc->packet_loss_percentage);
1145 case PROP_MAX_PAYLOAD_SIZE:
1146 g_value_set_uint (value, enc->max_payload_size);
1149 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
1153 g_mutex_unlock (&enc->property_lock);
1157 gst_opus_enc_set_property (GObject * object, guint prop_id,
1158 const GValue * value, GParamSpec * pspec)
1162 enc = GST_OPUS_ENC (object);
1164 #define GST_OPUS_UPDATE_PROPERTY(prop,type,ctl) do { \
1165 g_mutex_lock (&enc->property_lock); \
1166 enc->prop = g_value_get_##type (value); \
1168 opus_multistream_encoder_ctl (enc->state, OPUS_SET_##ctl (enc->prop)); \
1170 g_mutex_unlock (&enc->property_lock); \
1174 case PROP_AUDIO_TYPE:
1175 enc->audio_type = g_value_get_enum (value);
1178 GST_OPUS_UPDATE_PROPERTY (bitrate, int, BITRATE);
1180 case PROP_BANDWIDTH:
1181 GST_OPUS_UPDATE_PROPERTY (bandwidth, enum, BANDWIDTH);
1183 case PROP_FRAME_SIZE:
1184 g_mutex_lock (&enc->property_lock);
1185 enc->frame_size = g_value_get_enum (value);
1186 enc->frame_samples = gst_opus_enc_get_frame_samples (enc);
1187 gst_opus_enc_setup_base_class (enc, GST_AUDIO_ENCODER (enc));
1188 g_mutex_unlock (&enc->property_lock);
1190 case PROP_BITRATE_TYPE:
1191 /* this one has an opposite meaning to the opus ctl... */
1192 g_mutex_lock (&enc->property_lock);
1193 enc->bitrate_type = g_value_get_enum (value);
1195 opus_multistream_encoder_ctl (enc->state,
1196 OPUS_SET_VBR (enc->bitrate_type != BITRATE_TYPE_CBR));
1197 opus_multistream_encoder_ctl (enc->state,
1198 OPUS_SET_VBR_CONSTRAINT (enc->bitrate_type ==
1199 BITRATE_TYPE_CONSTRAINED_VBR), 0);
1201 g_mutex_unlock (&enc->property_lock);
1203 case PROP_COMPLEXITY:
1204 GST_OPUS_UPDATE_PROPERTY (complexity, int, COMPLEXITY);
1206 case PROP_INBAND_FEC:
1207 GST_OPUS_UPDATE_PROPERTY (inband_fec, boolean, INBAND_FEC);
1210 GST_OPUS_UPDATE_PROPERTY (dtx, boolean, DTX);
1212 case PROP_PACKET_LOSS_PERCENT:
1213 GST_OPUS_UPDATE_PROPERTY (packet_loss_percentage, int, PACKET_LOSS_PERC);
1215 case PROP_MAX_PAYLOAD_SIZE:
1216 g_mutex_lock (&enc->property_lock);
1217 enc->max_payload_size = g_value_get_uint (value);
1218 g_mutex_unlock (&enc->property_lock);
1221 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
1225 #undef GST_OPUS_UPDATE_PROPERTY