2 * Copyright (C) 2009 Justin Ruggles
3 * Copyright (c) 2009 Xuggle Incorporated
5 * This file is part of FFmpeg.
7 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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"
78 /* TODO: Think about converting abr, vad, dtx (still to come) and such flags to a bit field */
80 AVClass *class; ///< AVClass for private options
81 SpeexBits bits; ///< libspeex bitwriter context
82 SpeexHeader header; ///< libspeex header struct
83 void *enc_state; ///< libspeex encoder state
84 int frames_per_packet; ///< number of frames to encode in each packet
85 float vbr_quality; ///< VBR quality 0.0 to 10.0
86 int cbr_quality; ///< CBR quality 0 to 10
87 int abr; ///< flag to enable ABR
88 int vad; ///< flag to enable VAD
89 int pkt_frame_count; ///< frame count for the current packet
90 AudioFrameQueue afq; ///< frame queue
93 static av_cold void print_enc_params(AVCodecContext *avctx,
94 LibSpeexEncContext *s)
96 const char *mode_str = "unknown";
98 av_log(avctx, AV_LOG_DEBUG, "channels: %d\n", avctx->channels);
99 switch (s->header.mode) {
100 case SPEEX_MODEID_NB: mode_str = "narrowband"; break;
101 case SPEEX_MODEID_WB: mode_str = "wideband"; break;
102 case SPEEX_MODEID_UWB: mode_str = "ultra-wideband"; break;
104 av_log(avctx, AV_LOG_DEBUG, "mode: %s\n", mode_str);
106 av_log(avctx, AV_LOG_DEBUG, "rate control: VBR\n");
107 av_log(avctx, AV_LOG_DEBUG, " quality: %f\n", s->vbr_quality);
109 av_log(avctx, AV_LOG_DEBUG, "rate control: ABR\n");
110 av_log(avctx, AV_LOG_DEBUG, " bitrate: %d bps\n", avctx->bit_rate);
112 av_log(avctx, AV_LOG_DEBUG, "rate control: CBR\n");
113 av_log(avctx, AV_LOG_DEBUG, " bitrate: %d bps\n", avctx->bit_rate);
115 av_log(avctx, AV_LOG_DEBUG, "complexity: %d\n",
116 avctx->compression_level);
117 av_log(avctx, AV_LOG_DEBUG, "frame size: %d samples\n",
119 av_log(avctx, AV_LOG_DEBUG, "frames per packet: %d\n",
120 s->frames_per_packet);
121 av_log(avctx, AV_LOG_DEBUG, "packet size: %d\n",
122 avctx->frame_size * s->frames_per_packet);
123 av_log(avctx, AV_LOG_DEBUG, "voice activity detection: %d\n", s->vad);
126 static av_cold int encode_init(AVCodecContext *avctx)
128 LibSpeexEncContext *s = avctx->priv_data;
129 const SpeexMode *mode;
130 uint8_t *header_data;
135 if (avctx->channels < 1 || avctx->channels > 2) {
136 av_log(avctx, AV_LOG_ERROR, "Invalid channels (%d). Only stereo and "
137 "mono are supported\n", avctx->channels);
138 return AVERROR(EINVAL);
141 /* sample rate and encoding mode */
142 switch (avctx->sample_rate) {
143 case 8000: mode = &speex_nb_mode; break;
144 case 16000: mode = &speex_wb_mode; break;
145 case 32000: mode = &speex_uwb_mode; break;
147 av_log(avctx, AV_LOG_ERROR, "Sample rate of %d Hz is not supported. "
148 "Resample to 8, 16, or 32 kHz.\n", avctx->sample_rate);
149 return AVERROR(EINVAL);
152 /* initialize libspeex */
153 s->enc_state = speex_encoder_init(mode);
155 av_log(avctx, AV_LOG_ERROR, "Error initializing libspeex\n");
158 speex_init_header(&s->header, avctx->sample_rate, avctx->channels, mode);
160 /* rate control method and parameters */
161 if (avctx->flags & CODEC_FLAG_QSCALE) {
164 s->vad = 1; /* VAD is always implicitly activated for VBR */
165 speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR, &s->header.vbr);
166 s->vbr_quality = av_clipf(avctx->global_quality / (float)FF_QP2LAMBDA,
168 speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR_QUALITY, &s->vbr_quality);
170 s->header.bitrate = avctx->bit_rate;
171 if (avctx->bit_rate > 0) {
172 /* CBR or ABR by bitrate */
174 speex_encoder_ctl(s->enc_state, SPEEX_SET_ABR,
176 speex_encoder_ctl(s->enc_state, SPEEX_GET_ABR,
179 speex_encoder_ctl(s->enc_state, SPEEX_SET_BITRATE,
181 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
186 speex_encoder_ctl(s->enc_state, SPEEX_SET_QUALITY,
188 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
191 /* stereo side information adds about 800 bps to the base bitrate */
192 /* TODO: this should be calculated exactly */
193 avctx->bit_rate = s->header.bitrate + (avctx->channels == 2 ? 800 : 0);
196 /* VAD is activated with VBR or can be turned on by itself */
198 speex_encoder_ctl(s->enc_state, SPEEX_SET_VAD, &s->vad);
200 /* set encoding complexity */
201 if (avctx->compression_level > FF_COMPRESSION_DEFAULT) {
202 complexity = av_clip(avctx->compression_level, 0, 10);
203 speex_encoder_ctl(s->enc_state, SPEEX_SET_COMPLEXITY, &complexity);
205 speex_encoder_ctl(s->enc_state, SPEEX_GET_COMPLEXITY, &complexity);
206 avctx->compression_level = complexity;
208 /* set packet size */
209 avctx->frame_size = s->header.frame_size;
210 s->header.frames_per_packet = s->frames_per_packet;
212 /* set encoding delay */
213 speex_encoder_ctl(s->enc_state, SPEEX_GET_LOOKAHEAD, &avctx->delay);
214 ff_af_queue_init(avctx, &s->afq);
216 /* create header packet bytes from header struct */
217 /* note: libspeex allocates the memory for header_data, which is freed
218 below with speex_header_free() */
219 header_data = speex_header_to_packet(&s->header, &header_size);
221 /* allocate extradata and coded_frame */
222 avctx->extradata = av_malloc(header_size + FF_INPUT_BUFFER_PADDING_SIZE);
223 if (!avctx->extradata) {
224 speex_header_free(header_data);
225 speex_encoder_destroy(s->enc_state);
226 av_log(avctx, AV_LOG_ERROR, "memory allocation error\n");
227 return AVERROR(ENOMEM);
229 #if FF_API_OLD_ENCODE_AUDIO
230 avctx->coded_frame = avcodec_alloc_frame();
231 if (!avctx->coded_frame) {
232 av_freep(&avctx->extradata);
233 speex_header_free(header_data);
234 speex_encoder_destroy(s->enc_state);
235 av_log(avctx, AV_LOG_ERROR, "memory allocation error\n");
236 return AVERROR(ENOMEM);
240 /* copy header packet to extradata */
241 memcpy(avctx->extradata, header_data, header_size);
242 avctx->extradata_size = header_size;
243 speex_header_free(header_data);
245 /* init libspeex bitwriter */
246 speex_bits_init(&s->bits);
248 print_enc_params(avctx, s);
252 static int encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
253 const AVFrame *frame, int *got_packet_ptr)
255 LibSpeexEncContext *s = avctx->priv_data;
256 int16_t *samples = frame ? (int16_t *)frame->data[0] : NULL;
260 /* encode Speex frame */
261 if (avctx->channels == 2)
262 speex_encode_stereo_int(samples, s->header.frame_size, &s->bits);
263 speex_encode_int(s->enc_state, samples, &s->bits);
264 s->pkt_frame_count++;
265 if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
268 /* handle end-of-stream */
269 if (!s->pkt_frame_count)
271 /* add extra terminator codes for unused frames in last packet */
272 while (s->pkt_frame_count < s->frames_per_packet) {
273 speex_bits_pack(&s->bits, 15, 5);
274 s->pkt_frame_count++;
278 /* write output if all frames for the packet have been encoded */
279 if (s->pkt_frame_count == s->frames_per_packet) {
280 s->pkt_frame_count = 0;
281 if ((ret = ff_alloc_packet2(avctx, avpkt, speex_bits_nbytes(&s->bits))))
283 ret = speex_bits_write(&s->bits, avpkt->data, avpkt->size);
284 speex_bits_reset(&s->bits);
286 /* Get the next frame pts/duration */
287 ff_af_queue_remove(&s->afq, s->frames_per_packet * avctx->frame_size,
288 &avpkt->pts, &avpkt->duration);
297 static av_cold int encode_close(AVCodecContext *avctx)
299 LibSpeexEncContext *s = avctx->priv_data;
301 speex_bits_destroy(&s->bits);
302 speex_encoder_destroy(s->enc_state);
304 ff_af_queue_close(&s->afq);
305 #if FF_API_OLD_ENCODE_AUDIO
306 av_freep(&avctx->coded_frame);
308 av_freep(&avctx->extradata);
313 #define OFFSET(x) offsetof(LibSpeexEncContext, x)
314 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
315 static const AVOption options[] = {
316 { "abr", "Use average bit rate", OFFSET(abr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE },
317 { "cbr_quality", "Set quality value (0 to 10) for CBR", OFFSET(cbr_quality), AV_OPT_TYPE_INT, { .i64 = 8 }, 0, 10, AE },
318 { "frames_per_packet", "Number of frames to encode in each packet", OFFSET(frames_per_packet), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 8, AE },
319 { "vad", "Voice Activity Detection", OFFSET(vad), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE },
323 static const AVClass class = {
324 .class_name = "libspeex",
325 .item_name = av_default_item_name,
327 .version = LIBAVUTIL_VERSION_INT,
330 static const AVCodecDefault defaults[] = {
332 { "compression_level", "3" },
336 AVCodec ff_libspeex_encoder = {
338 .type = AVMEDIA_TYPE_AUDIO,
339 .id = AV_CODEC_ID_SPEEX,
340 .priv_data_size = sizeof(LibSpeexEncContext),
342 .encode2 = encode_frame,
343 .close = encode_close,
344 .capabilities = CODEC_CAP_DELAY,
345 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
346 AV_SAMPLE_FMT_NONE },
347 .channel_layouts = (const uint64_t[]){ AV_CH_LAYOUT_MONO,
350 .supported_samplerates = (const int[]){ 8000, 16000, 32000, 0 },
351 .long_name = NULL_IF_CONFIG_SMALL("libspeex Speex"),
352 .priv_class = &class,
353 .defaults = defaults,