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 pkt_frame_count; ///< frame count for the current packet
88 AudioFrameQueue afq; ///< frame queue
91 static av_cold void print_enc_params(AVCodecContext *avctx,
92 LibSpeexEncContext *s)
94 const char *mode_str = "unknown";
96 av_log(avctx, AV_LOG_DEBUG, "channels: %d\n", avctx->channels);
97 switch (s->header.mode) {
98 case SPEEX_MODEID_NB: mode_str = "narrowband"; break;
99 case SPEEX_MODEID_WB: mode_str = "wideband"; break;
100 case SPEEX_MODEID_UWB: mode_str = "ultra-wideband"; break;
102 av_log(avctx, AV_LOG_DEBUG, "mode: %s\n", mode_str);
104 av_log(avctx, AV_LOG_DEBUG, "rate control: VBR\n");
105 av_log(avctx, AV_LOG_DEBUG, " quality: %f\n", s->vbr_quality);
107 av_log(avctx, AV_LOG_DEBUG, "rate control: ABR\n");
108 av_log(avctx, AV_LOG_DEBUG, " bitrate: %d bps\n", avctx->bit_rate);
110 av_log(avctx, AV_LOG_DEBUG, "rate control: CBR\n");
111 av_log(avctx, AV_LOG_DEBUG, " bitrate: %d bps\n", avctx->bit_rate);
113 av_log(avctx, AV_LOG_DEBUG, "complexity: %d\n",
114 avctx->compression_level);
115 av_log(avctx, AV_LOG_DEBUG, "frame size: %d samples\n",
117 av_log(avctx, AV_LOG_DEBUG, "frames per packet: %d\n",
118 s->frames_per_packet);
119 av_log(avctx, AV_LOG_DEBUG, "packet size: %d\n",
120 avctx->frame_size * s->frames_per_packet);
123 static av_cold int encode_init(AVCodecContext *avctx)
125 LibSpeexEncContext *s = avctx->priv_data;
126 const SpeexMode *mode;
127 uint8_t *header_data;
132 if (avctx->channels < 1 || avctx->channels > 2) {
133 av_log(avctx, AV_LOG_ERROR, "Invalid channels (%d). Only stereo and "
134 "mono are supported\n", avctx->channels);
135 return AVERROR(EINVAL);
138 /* sample rate and encoding mode */
139 switch (avctx->sample_rate) {
140 case 8000: mode = &speex_nb_mode; break;
141 case 16000: mode = &speex_wb_mode; break;
142 case 32000: mode = &speex_uwb_mode; break;
144 av_log(avctx, AV_LOG_ERROR, "Sample rate of %d Hz is not supported. "
145 "Resample to 8, 16, or 32 kHz.\n", avctx->sample_rate);
146 return AVERROR(EINVAL);
149 /* initialize libspeex */
150 s->enc_state = speex_encoder_init(mode);
152 av_log(avctx, AV_LOG_ERROR, "Error initializing libspeex\n");
155 speex_init_header(&s->header, avctx->sample_rate, avctx->channels, mode);
157 /* rate control method and parameters */
158 if (avctx->flags & CODEC_FLAG_QSCALE) {
161 speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR, &s->header.vbr);
162 s->vbr_quality = av_clipf(avctx->global_quality / (float)FF_QP2LAMBDA,
164 speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR_QUALITY, &s->vbr_quality);
166 s->header.bitrate = avctx->bit_rate;
167 if (avctx->bit_rate > 0) {
168 /* CBR or ABR by bitrate */
170 speex_encoder_ctl(s->enc_state, SPEEX_SET_ABR,
172 speex_encoder_ctl(s->enc_state, SPEEX_GET_ABR,
175 speex_encoder_ctl(s->enc_state, SPEEX_SET_BITRATE,
177 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
182 speex_encoder_ctl(s->enc_state, SPEEX_SET_QUALITY,
184 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
187 /* stereo side information adds about 800 bps to the base bitrate */
188 /* TODO: this should be calculated exactly */
189 avctx->bit_rate = s->header.bitrate + (avctx->channels == 2 ? 800 : 0);
192 /* set encoding complexity */
193 if (avctx->compression_level > FF_COMPRESSION_DEFAULT) {
194 complexity = av_clip(avctx->compression_level, 0, 10);
195 speex_encoder_ctl(s->enc_state, SPEEX_SET_COMPLEXITY, &complexity);
197 speex_encoder_ctl(s->enc_state, SPEEX_GET_COMPLEXITY, &complexity);
198 avctx->compression_level = complexity;
200 /* set packet size */
201 avctx->frame_size = s->header.frame_size;
202 s->header.frames_per_packet = s->frames_per_packet;
204 /* set encoding delay */
205 speex_encoder_ctl(s->enc_state, SPEEX_GET_LOOKAHEAD, &avctx->delay);
206 ff_af_queue_init(avctx, &s->afq);
208 /* create header packet bytes from header struct */
209 /* note: libspeex allocates the memory for header_data, which is freed
210 below with speex_header_free() */
211 header_data = speex_header_to_packet(&s->header, &header_size);
213 /* allocate extradata and coded_frame */
214 avctx->extradata = av_malloc(header_size + FF_INPUT_BUFFER_PADDING_SIZE);
215 if (!avctx->extradata) {
216 speex_header_free(header_data);
217 speex_encoder_destroy(s->enc_state);
218 av_log(avctx, AV_LOG_ERROR, "memory allocation error\n");
219 return AVERROR(ENOMEM);
221 #if FF_API_OLD_ENCODE_AUDIO
222 avctx->coded_frame = avcodec_alloc_frame();
223 if (!avctx->coded_frame) {
224 av_freep(&avctx->extradata);
225 speex_header_free(header_data);
226 speex_encoder_destroy(s->enc_state);
227 av_log(avctx, AV_LOG_ERROR, "memory allocation error\n");
228 return AVERROR(ENOMEM);
232 /* copy header packet to extradata */
233 memcpy(avctx->extradata, header_data, header_size);
234 avctx->extradata_size = header_size;
235 speex_header_free(header_data);
237 /* init libspeex bitwriter */
238 speex_bits_init(&s->bits);
240 print_enc_params(avctx, s);
244 static int encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
245 const AVFrame *frame, int *got_packet_ptr)
247 LibSpeexEncContext *s = avctx->priv_data;
248 int16_t *samples = frame ? (int16_t *)frame->data[0] : NULL;
252 /* encode Speex frame */
253 if (avctx->channels == 2)
254 speex_encode_stereo_int(samples, s->header.frame_size, &s->bits);
255 speex_encode_int(s->enc_state, samples, &s->bits);
256 s->pkt_frame_count++;
257 if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
260 /* handle end-of-stream */
261 if (!s->pkt_frame_count)
263 /* add extra terminator codes for unused frames in last packet */
264 while (s->pkt_frame_count < s->frames_per_packet) {
265 speex_bits_pack(&s->bits, 15, 5);
266 s->pkt_frame_count++;
270 /* write output if all frames for the packet have been encoded */
271 if (s->pkt_frame_count == s->frames_per_packet) {
272 s->pkt_frame_count = 0;
273 if ((ret = ff_alloc_packet(avpkt, speex_bits_nbytes(&s->bits)))) {
274 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
277 ret = speex_bits_write(&s->bits, avpkt->data, avpkt->size);
278 speex_bits_reset(&s->bits);
280 /* Get the next frame pts/duration */
281 ff_af_queue_remove(&s->afq, s->frames_per_packet * avctx->frame_size,
282 &avpkt->pts, &avpkt->duration);
291 static av_cold int encode_close(AVCodecContext *avctx)
293 LibSpeexEncContext *s = avctx->priv_data;
295 speex_bits_destroy(&s->bits);
296 speex_encoder_destroy(s->enc_state);
298 ff_af_queue_close(&s->afq);
299 #if FF_API_OLD_ENCODE_AUDIO
300 av_freep(&avctx->coded_frame);
302 av_freep(&avctx->extradata);
307 #define OFFSET(x) offsetof(LibSpeexEncContext, x)
308 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
309 static const AVOption options[] = {
310 { "abr", "Use average bit rate", OFFSET(abr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE },
311 { "cbr_quality", "Set quality value (0 to 10) for CBR", OFFSET(cbr_quality), AV_OPT_TYPE_INT, { .i64 = 8 }, 0, 10, AE },
312 { "frames_per_packet", "Number of frames to encode in each packet", OFFSET(frames_per_packet), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 8, AE },
316 static const AVClass class = {
317 .class_name = "libspeex",
318 .item_name = av_default_item_name,
320 .version = LIBAVUTIL_VERSION_INT,
323 static const AVCodecDefault defaults[] = {
325 { "compression_level", "3" },
329 AVCodec ff_libspeex_encoder = {
331 .type = AVMEDIA_TYPE_AUDIO,
332 .id = AV_CODEC_ID_SPEEX,
333 .priv_data_size = sizeof(LibSpeexEncContext),
335 .encode2 = encode_frame,
336 .close = encode_close,
337 .capabilities = CODEC_CAP_DELAY,
338 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
339 AV_SAMPLE_FMT_NONE },
340 .channel_layouts = (const uint64_t[]){ AV_CH_LAYOUT_MONO,
343 .supported_samplerates = (const int[]){ 8000, 16000, 32000, 0 },
344 .long_name = NULL_IF_CONFIG_SMALL("libspeex Speex"),
345 .priv_class = &class,
346 .defaults = defaults,