2 * Copyright (C) 2009 Justin Ruggles
3 * Copyright (c) 2009 Xuggle Incorporated
5 * This file is part of Libav.
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * libspeex Speex audio encoder
27 * This explains the values that need to be set prior to initialization in
28 * order to control various encoding parameters.
31 * Speex only supports mono or stereo, so avctx->channels must be set to
34 * Sample Rate / Encoding Mode
35 * Speex has 3 modes, each of which uses a specific sample rate.
38 * ultra-wideband : 32 kHz
39 * avctx->sample_rate must be set to one of these 3 values. This will be
40 * used to set the encoding mode.
43 * VBR mode is turned on by setting CODEC_FLAG_QSCALE in avctx->flags.
44 * avctx->global_quality is used to set the encoding quality.
45 * For CBR mode, avctx->bit_rate can be used to set the constant bitrate.
46 * Alternatively, the 'cbr_quality' option can be set from 0 to 10 to set
47 * a constant bitrate based on quality.
48 * For ABR mode, set avctx->bit_rate and set the 'abr' option to 1.
49 * Approx. Bitrate Range:
50 * narrowband : 2400 - 25600 bps
51 * wideband : 4000 - 43200 bps
52 * ultra-wideband : 4400 - 45200 bps
55 * Encoding complexity is controlled by setting avctx->compression_level.
56 * The valid range is 0 to 10. A higher setting gives generally better
57 * quality at the expense of encoding speed. This does not affect the
61 * The encoder defaults to using 1 frame-per-packet. However, it is
62 * sometimes desirable to use multiple frames-per-packet to reduce the
63 * amount of container overhead. This can be done by setting the
64 * 'frames_per_packet' option to a value 1 to 8.
67 #include <speex/speex.h>
68 #include <speex/speex_header.h>
69 #include <speex/speex_stereo.h>
71 #include "libavutil/audioconvert.h"
72 #include "libavutil/common.h"
73 #include "libavutil/opt.h"
76 #include "audio_frame_queue.h"
79 AVClass *class; ///< AVClass for private options
80 SpeexBits bits; ///< libspeex bitwriter context
81 SpeexHeader header; ///< libspeex header struct
82 void *enc_state; ///< libspeex encoder state
83 int frames_per_packet; ///< number of frames to encode in each packet
84 float vbr_quality; ///< VBR quality 0.0 to 10.0
85 int cbr_quality; ///< CBR quality 0 to 10
86 int abr; ///< flag to enable ABR
87 int vad; ///< flag to enable VAD
88 int pkt_frame_count; ///< frame count for the current packet
89 AudioFrameQueue afq; ///< frame queue
92 static av_cold void print_enc_params(AVCodecContext *avctx,
93 LibSpeexEncContext *s)
95 const char *mode_str = "unknown";
97 av_log(avctx, AV_LOG_DEBUG, "channels: %d\n", avctx->channels);
98 switch (s->header.mode) {
99 case SPEEX_MODEID_NB: mode_str = "narrowband"; break;
100 case SPEEX_MODEID_WB: mode_str = "wideband"; break;
101 case SPEEX_MODEID_UWB: mode_str = "ultra-wideband"; break;
103 av_log(avctx, AV_LOG_DEBUG, "mode: %s\n", mode_str);
105 av_log(avctx, AV_LOG_DEBUG, "rate control: VBR\n");
106 av_log(avctx, AV_LOG_DEBUG, " quality: %f\n", s->vbr_quality);
108 av_log(avctx, AV_LOG_DEBUG, "rate control: ABR\n");
109 av_log(avctx, AV_LOG_DEBUG, " bitrate: %d bps\n", avctx->bit_rate);
111 av_log(avctx, AV_LOG_DEBUG, "rate control: CBR\n");
112 av_log(avctx, AV_LOG_DEBUG, " bitrate: %d bps\n", avctx->bit_rate);
114 av_log(avctx, AV_LOG_DEBUG, "complexity: %d\n",
115 avctx->compression_level);
116 av_log(avctx, AV_LOG_DEBUG, "frame size: %d samples\n",
118 av_log(avctx, AV_LOG_DEBUG, "frames per packet: %d\n",
119 s->frames_per_packet);
120 av_log(avctx, AV_LOG_DEBUG, "packet size: %d\n",
121 avctx->frame_size * s->frames_per_packet);
122 av_log(avctx, AV_LOG_DEBUG, "voice activity detection: %d\n", s->vad);
125 static av_cold int encode_init(AVCodecContext *avctx)
127 LibSpeexEncContext *s = avctx->priv_data;
128 const SpeexMode *mode;
129 uint8_t *header_data;
134 if (avctx->channels < 1 || avctx->channels > 2) {
135 av_log(avctx, AV_LOG_ERROR, "Invalid channels (%d). Only stereo and "
136 "mono are supported\n", avctx->channels);
137 return AVERROR(EINVAL);
140 /* sample rate and encoding mode */
141 switch (avctx->sample_rate) {
142 case 8000: mode = &speex_nb_mode; break;
143 case 16000: mode = &speex_wb_mode; break;
144 case 32000: mode = &speex_uwb_mode; break;
146 av_log(avctx, AV_LOG_ERROR, "Sample rate of %d Hz is not supported. "
147 "Resample to 8, 16, or 32 kHz.\n", avctx->sample_rate);
148 return AVERROR(EINVAL);
151 /* initialize libspeex */
152 s->enc_state = speex_encoder_init(mode);
154 av_log(avctx, AV_LOG_ERROR, "Error initializing libspeex\n");
157 speex_init_header(&s->header, avctx->sample_rate, avctx->channels, mode);
159 /* rate control method and parameters */
160 if (avctx->flags & CODEC_FLAG_QSCALE) {
163 s->vad = 1; /* VAD is always implicitly activated for VBR */
164 speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR, &s->header.vbr);
165 s->vbr_quality = av_clipf(avctx->global_quality / (float)FF_QP2LAMBDA,
167 speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR_QUALITY, &s->vbr_quality);
169 s->header.bitrate = avctx->bit_rate;
170 if (avctx->bit_rate > 0) {
171 /* CBR or ABR by bitrate */
173 speex_encoder_ctl(s->enc_state, SPEEX_SET_ABR,
175 speex_encoder_ctl(s->enc_state, SPEEX_GET_ABR,
178 speex_encoder_ctl(s->enc_state, SPEEX_SET_BITRATE,
180 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
185 speex_encoder_ctl(s->enc_state, SPEEX_SET_QUALITY,
187 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
190 /* stereo side information adds about 800 bps to the base bitrate */
191 /* TODO: this should be calculated exactly */
192 avctx->bit_rate = s->header.bitrate + (avctx->channels == 2 ? 800 : 0);
195 /* VAD is activated with VBR or can be turned on by itself */
197 speex_encoder_ctl(s->enc_state, SPEEX_SET_VAD, &s->vad);
199 /* set encoding complexity */
200 if (avctx->compression_level > FF_COMPRESSION_DEFAULT) {
201 complexity = av_clip(avctx->compression_level, 0, 10);
202 speex_encoder_ctl(s->enc_state, SPEEX_SET_COMPLEXITY, &complexity);
204 speex_encoder_ctl(s->enc_state, SPEEX_GET_COMPLEXITY, &complexity);
205 avctx->compression_level = complexity;
207 /* set packet size */
208 avctx->frame_size = s->header.frame_size;
209 s->header.frames_per_packet = s->frames_per_packet;
211 /* set encoding delay */
212 speex_encoder_ctl(s->enc_state, SPEEX_GET_LOOKAHEAD, &avctx->delay);
213 ff_af_queue_init(avctx, &s->afq);
215 /* create header packet bytes from header struct */
216 /* note: libspeex allocates the memory for header_data, which is freed
217 below with speex_header_free() */
218 header_data = speex_header_to_packet(&s->header, &header_size);
220 /* allocate extradata and coded_frame */
221 avctx->extradata = av_malloc(header_size + FF_INPUT_BUFFER_PADDING_SIZE);
222 if (!avctx->extradata) {
223 speex_header_free(header_data);
224 speex_encoder_destroy(s->enc_state);
225 av_log(avctx, AV_LOG_ERROR, "memory allocation error\n");
226 return AVERROR(ENOMEM);
228 #if FF_API_OLD_ENCODE_AUDIO
229 avctx->coded_frame = avcodec_alloc_frame();
230 if (!avctx->coded_frame) {
231 av_freep(&avctx->extradata);
232 speex_header_free(header_data);
233 speex_encoder_destroy(s->enc_state);
234 av_log(avctx, AV_LOG_ERROR, "memory allocation error\n");
235 return AVERROR(ENOMEM);
239 /* copy header packet to extradata */
240 memcpy(avctx->extradata, header_data, header_size);
241 avctx->extradata_size = header_size;
242 speex_header_free(header_data);
244 /* init libspeex bitwriter */
245 speex_bits_init(&s->bits);
247 print_enc_params(avctx, s);
251 static int encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
252 const AVFrame *frame, int *got_packet_ptr)
254 LibSpeexEncContext *s = avctx->priv_data;
255 int16_t *samples = frame ? (int16_t *)frame->data[0] : NULL;
259 /* encode Speex frame */
260 if (avctx->channels == 2)
261 speex_encode_stereo_int(samples, s->header.frame_size, &s->bits);
262 speex_encode_int(s->enc_state, samples, &s->bits);
263 s->pkt_frame_count++;
264 if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
267 /* handle end-of-stream */
268 if (!s->pkt_frame_count)
270 /* add extra terminator codes for unused frames in last packet */
271 while (s->pkt_frame_count < s->frames_per_packet) {
272 speex_bits_pack(&s->bits, 15, 5);
273 s->pkt_frame_count++;
277 /* write output if all frames for the packet have been encoded */
278 if (s->pkt_frame_count == s->frames_per_packet) {
279 s->pkt_frame_count = 0;
280 if ((ret = ff_alloc_packet(avpkt, speex_bits_nbytes(&s->bits)))) {
281 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
284 ret = speex_bits_write(&s->bits, avpkt->data, avpkt->size);
285 speex_bits_reset(&s->bits);
287 /* Get the next frame pts/duration */
288 ff_af_queue_remove(&s->afq, s->frames_per_packet * avctx->frame_size,
289 &avpkt->pts, &avpkt->duration);
298 static av_cold int encode_close(AVCodecContext *avctx)
300 LibSpeexEncContext *s = avctx->priv_data;
302 speex_bits_destroy(&s->bits);
303 speex_encoder_destroy(s->enc_state);
305 ff_af_queue_close(&s->afq);
306 #if FF_API_OLD_ENCODE_AUDIO
307 av_freep(&avctx->coded_frame);
309 av_freep(&avctx->extradata);
314 #define OFFSET(x) offsetof(LibSpeexEncContext, x)
315 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
316 static const AVOption options[] = {
317 { "abr", "Use average bit rate", OFFSET(abr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE },
318 { "cbr_quality", "Set quality value (0 to 10) for CBR", OFFSET(cbr_quality), AV_OPT_TYPE_INT, { .i64 = 8 }, 0, 10, AE },
319 { "frames_per_packet", "Number of frames to encode in each packet", OFFSET(frames_per_packet), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 8, AE },
320 { "vad", "Voice Activity Detection", OFFSET(vad), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE },
324 static const AVClass class = {
325 .class_name = "libspeex",
326 .item_name = av_default_item_name,
328 .version = LIBAVUTIL_VERSION_INT,
331 static const AVCodecDefault defaults[] = {
333 { "compression_level", "3" },
337 AVCodec ff_libspeex_encoder = {
339 .type = AVMEDIA_TYPE_AUDIO,
340 .id = AV_CODEC_ID_SPEEX,
341 .priv_data_size = sizeof(LibSpeexEncContext),
343 .encode2 = encode_frame,
344 .close = encode_close,
345 .capabilities = CODEC_CAP_DELAY,
346 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
347 AV_SAMPLE_FMT_NONE },
348 .channel_layouts = (const uint64_t[]){ AV_CH_LAYOUT_MONO,
351 .supported_samplerates = (const int[]){ 8000, 16000, 32000, 0 },
352 .long_name = NULL_IF_CONFIG_SMALL("libspeex Speex"),
353 .priv_class = &class,
354 .defaults = defaults,