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>
70 #include "libavutil/mathematics.h"
71 #include "libavutil/opt.h"
76 AVClass *class; ///< AVClass for private options
77 SpeexBits bits; ///< libspeex bitwriter context
78 SpeexHeader header; ///< libspeex header struct
79 void *enc_state; ///< libspeex encoder state
80 int frames_per_packet; ///< number of frames to encode in each packet
81 float vbr_quality; ///< VBR quality 0.0 to 10.0
82 int cbr_quality; ///< CBR quality 0 to 10
83 int abr; ///< flag to enable ABR
84 int pkt_frame_count; ///< frame count for the current packet
85 int lookahead; ///< encoder delay
86 int sample_count; ///< total sample count (used for pts)
89 static av_cold void print_enc_params(AVCodecContext *avctx,
90 LibSpeexEncContext *s)
92 const char *mode_str = "unknown";
94 av_log(avctx, AV_LOG_DEBUG, "channels: %d\n", avctx->channels);
95 switch (s->header.mode) {
96 case SPEEX_MODEID_NB: mode_str = "narrowband"; break;
97 case SPEEX_MODEID_WB: mode_str = "wideband"; break;
98 case SPEEX_MODEID_UWB: mode_str = "ultra-wideband"; break;
100 av_log(avctx, AV_LOG_DEBUG, "mode: %s\n", mode_str);
102 av_log(avctx, AV_LOG_DEBUG, "rate control: VBR\n");
103 av_log(avctx, AV_LOG_DEBUG, " quality: %f\n", s->vbr_quality);
105 av_log(avctx, AV_LOG_DEBUG, "rate control: ABR\n");
106 av_log(avctx, AV_LOG_DEBUG, " bitrate: %d bps\n", avctx->bit_rate);
108 av_log(avctx, AV_LOG_DEBUG, "rate control: CBR\n");
109 av_log(avctx, AV_LOG_DEBUG, " bitrate: %d bps\n", avctx->bit_rate);
111 av_log(avctx, AV_LOG_DEBUG, "complexity: %d\n",
112 avctx->compression_level);
113 av_log(avctx, AV_LOG_DEBUG, "frame size: %d samples\n",
115 av_log(avctx, AV_LOG_DEBUG, "frames per packet: %d\n",
116 s->frames_per_packet);
117 av_log(avctx, AV_LOG_DEBUG, "packet size: %d\n",
118 avctx->frame_size * s->frames_per_packet);
121 static av_cold int encode_init(AVCodecContext *avctx)
123 LibSpeexEncContext *s = avctx->priv_data;
124 const SpeexMode *mode;
125 uint8_t *header_data;
130 if (avctx->channels < 1 || avctx->channels > 2) {
131 av_log(avctx, AV_LOG_ERROR, "Invalid channels (%d). Only stereo and "
132 "mono are supported\n", avctx->channels);
133 return AVERROR(EINVAL);
136 /* sample rate and encoding mode */
137 switch (avctx->sample_rate) {
138 case 8000: mode = &speex_nb_mode; break;
139 case 16000: mode = &speex_wb_mode; break;
140 case 32000: mode = &speex_uwb_mode; break;
142 av_log(avctx, AV_LOG_ERROR, "Sample rate of %d Hz is not supported. "
143 "Resample to 8, 16, or 32 kHz.\n", avctx->sample_rate);
144 return AVERROR(EINVAL);
147 /* initialize libspeex */
148 s->enc_state = speex_encoder_init(mode);
150 av_log(avctx, AV_LOG_ERROR, "Error initializing libspeex\n");
153 speex_init_header(&s->header, avctx->sample_rate, avctx->channels, mode);
155 /* rate control method and parameters */
156 if (avctx->flags & CODEC_FLAG_QSCALE) {
159 speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR, &s->header.vbr);
160 s->vbr_quality = av_clipf(avctx->global_quality / (float)FF_QP2LAMBDA,
162 speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR_QUALITY, &s->vbr_quality);
164 s->header.bitrate = avctx->bit_rate;
165 if (avctx->bit_rate > 0) {
166 /* CBR or ABR by bitrate */
168 speex_encoder_ctl(s->enc_state, SPEEX_SET_ABR,
170 speex_encoder_ctl(s->enc_state, SPEEX_GET_ABR,
173 speex_encoder_ctl(s->enc_state, SPEEX_SET_BITRATE,
175 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
180 speex_encoder_ctl(s->enc_state, SPEEX_SET_QUALITY,
182 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
185 /* stereo side information adds about 800 bps to the base bitrate */
186 /* TODO: this should be calculated exactly */
187 avctx->bit_rate = s->header.bitrate + (avctx->channels == 2 ? 800 : 0);
190 /* set encoding complexity */
191 if (avctx->compression_level > FF_COMPRESSION_DEFAULT) {
192 complexity = av_clip(avctx->compression_level, 0, 10);
193 speex_encoder_ctl(s->enc_state, SPEEX_SET_COMPLEXITY, &complexity);
195 speex_encoder_ctl(s->enc_state, SPEEX_GET_COMPLEXITY, &complexity);
196 avctx->compression_level = complexity;
198 /* set packet size */
199 avctx->frame_size = s->header.frame_size;
200 s->header.frames_per_packet = s->frames_per_packet;
202 /* set encoding delay */
203 speex_encoder_ctl(s->enc_state, SPEEX_GET_LOOKAHEAD, &s->lookahead);
204 s->sample_count = -s->lookahead;
206 /* create header packet bytes from header struct */
207 /* note: libspeex allocates the memory for header_data, which is freed
208 below with speex_header_free() */
209 header_data = speex_header_to_packet(&s->header, &header_size);
211 /* allocate extradata and coded_frame */
212 avctx->extradata = av_malloc(header_size + FF_INPUT_BUFFER_PADDING_SIZE);
213 avctx->coded_frame = avcodec_alloc_frame();
214 if (!avctx->extradata || !avctx->coded_frame) {
215 speex_header_free(header_data);
216 speex_encoder_destroy(s->enc_state);
217 av_log(avctx, AV_LOG_ERROR, "memory allocation error\n");
218 return AVERROR(ENOMEM);
221 /* copy header packet to extradata */
222 memcpy(avctx->extradata, header_data, header_size);
223 avctx->extradata_size = header_size;
224 speex_header_free(header_data);
226 /* init libspeex bitwriter */
227 speex_bits_init(&s->bits);
229 print_enc_params(avctx, s);
233 static int encode_frame(AVCodecContext *avctx, uint8_t *frame, int buf_size,
236 LibSpeexEncContext *s = avctx->priv_data;
237 int16_t *samples = data;
238 int sample_count = s->sample_count;
241 /* encode Speex frame */
242 if (avctx->channels == 2)
243 speex_encode_stereo_int(samples, s->header.frame_size, &s->bits);
244 speex_encode_int(s->enc_state, samples, &s->bits);
245 s->pkt_frame_count++;
246 s->sample_count += avctx->frame_size;
248 /* handle end-of-stream */
249 if (!s->pkt_frame_count)
251 /* add extra terminator codes for unused frames in last packet */
252 while (s->pkt_frame_count < s->frames_per_packet) {
253 speex_bits_pack(&s->bits, 15, 5);
254 s->pkt_frame_count++;
258 /* write output if all frames for the packet have been encoded */
259 if (s->pkt_frame_count == s->frames_per_packet) {
260 s->pkt_frame_count = 0;
261 avctx->coded_frame->pts =
262 av_rescale_q(sample_count, (AVRational){ 1, avctx->sample_rate },
264 if (buf_size > speex_bits_nbytes(&s->bits)) {
265 int ret = speex_bits_write(&s->bits, frame, buf_size);
266 speex_bits_reset(&s->bits);
269 av_log(avctx, AV_LOG_ERROR, "output buffer too small");
270 return AVERROR(EINVAL);
276 static av_cold int encode_close(AVCodecContext *avctx)
278 LibSpeexEncContext *s = avctx->priv_data;
280 speex_bits_destroy(&s->bits);
281 speex_encoder_destroy(s->enc_state);
283 av_freep(&avctx->coded_frame);
284 av_freep(&avctx->extradata);
289 #define OFFSET(x) offsetof(LibSpeexEncContext, x)
290 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
291 static const AVOption options[] = {
292 { "abr", "Use average bit rate", OFFSET(abr), AV_OPT_TYPE_INT, { 0 }, 0, 1, AE },
293 { "cbr_quality", "Set quality value (0 to 10) for CBR", OFFSET(cbr_quality), AV_OPT_TYPE_INT, { 8 }, 0, 10, AE },
294 { "frames_per_packet", "Number of frames to encode in each packet", OFFSET(frames_per_packet), AV_OPT_TYPE_INT, { 1 }, 1, 8, AE },
298 static const AVClass class = {
299 .class_name = "libspeex",
300 .item_name = av_default_item_name,
302 .version = LIBAVUTIL_VERSION_INT,
305 static const AVCodecDefault defaults[] = {
307 { "compression_level", "3" },
311 AVCodec ff_libspeex_encoder = {
313 .type = AVMEDIA_TYPE_AUDIO,
314 .id = CODEC_ID_SPEEX,
315 .priv_data_size = sizeof(LibSpeexEncContext),
317 .encode = encode_frame,
318 .close = encode_close,
319 .capabilities = CODEC_CAP_DELAY,
320 .sample_fmts = (const enum SampleFormat[]){ AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE },
321 .long_name = NULL_IF_CONFIG_SMALL("libspeex Speex"),
322 .priv_class = &class,
323 .defaults = defaults,