FFmpeg: libavcodec/libspeexenc.c Source File

00001 /*
00002  * Copyright (C) 2009 Justin Ruggles
00003  * Copyright (c) 2009 Xuggle Incorporated
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00067 #include <speex/speex.h>
00068 #include <speex/speex_header.h>
00069 #include <speex/speex_stereo.h>
00070 #include "libavutil/mathematics.h"
00071 #include "libavutil/opt.h"
00072 #include "avcodec.h"
00073 #include "internal.h"
00074 
00075 typedef struct {
00076     AVClass *class;             
00077     SpeexBits bits;             
00078     SpeexHeader header;         
00079     void *enc_state;            
00080     int frames_per_packet;      
00081     float vbr_quality;          
00082     int cbr_quality;            
00083     int abr;                    
00084     int pkt_frame_count;        
00085     int lookahead;              
00086     int64_t next_pts;           
00087     int pkt_sample_count;       
00088 } LibSpeexEncContext;
00089 
00090 static av_cold void print_enc_params(AVCodecContext *avctx,
00091                                      LibSpeexEncContext *s)
00092 {
00093     const char *mode_str = "unknown";
00094 
00095     av_log(avctx, AV_LOG_DEBUG, "channels: %d\n", avctx->channels);
00096     switch (s->header.mode) {
00097     case SPEEX_MODEID_NB:  mode_str = "narrowband";     break;
00098     case SPEEX_MODEID_WB:  mode_str = "wideband";       break;
00099     case SPEEX_MODEID_UWB: mode_str = "ultra-wideband"; break;
00100     }
00101     av_log(avctx, AV_LOG_DEBUG, "mode: %s\n", mode_str);
00102     if (s->header.vbr) {
00103         av_log(avctx, AV_LOG_DEBUG, "rate control: VBR\n");
00104         av_log(avctx, AV_LOG_DEBUG, "  quality: %f\n", s->vbr_quality);
00105     } else if (s->abr) {
00106         av_log(avctx, AV_LOG_DEBUG, "rate control: ABR\n");
00107         av_log(avctx, AV_LOG_DEBUG, "  bitrate: %d bps\n", avctx->bit_rate);
00108     } else {
00109         av_log(avctx, AV_LOG_DEBUG, "rate control: CBR\n");
00110         av_log(avctx, AV_LOG_DEBUG, "  bitrate: %d bps\n", avctx->bit_rate);
00111     }
00112     av_log(avctx, AV_LOG_DEBUG, "complexity: %d\n",
00113            avctx->compression_level);
00114     av_log(avctx, AV_LOG_DEBUG, "frame size: %d samples\n",
00115            avctx->frame_size);
00116     av_log(avctx, AV_LOG_DEBUG, "frames per packet: %d\n",
00117            s->frames_per_packet);
00118     av_log(avctx, AV_LOG_DEBUG, "packet size: %d\n",
00119            avctx->frame_size * s->frames_per_packet);
00120 }
00121 
00122 static av_cold int encode_init(AVCodecContext *avctx)
00123 {
00124     LibSpeexEncContext *s = avctx->priv_data;
00125     const SpeexMode *mode;
00126     uint8_t *header_data;
00127     int header_size;
00128     int32_t complexity;
00129 
00130     /* channels */
00131     if (avctx->channels < 1 || avctx->channels > 2) {
00132         av_log(avctx, AV_LOG_ERROR, "Invalid channels (%d). Only stereo and "
00133                "mono are supported\n", avctx->channels);
00134         return AVERROR(EINVAL);
00135     }
00136 
00137     /* sample rate and encoding mode */
00138     switch (avctx->sample_rate) {
00139     case  8000: mode = &speex_nb_mode;  break;
00140     case 16000: mode = &speex_wb_mode;  break;
00141     case 32000: mode = &speex_uwb_mode; break;
00142     default:
00143         av_log(avctx, AV_LOG_ERROR, "Sample rate of %d Hz is not supported. "
00144                "Resample to 8, 16, or 32 kHz.\n", avctx->sample_rate);
00145         return AVERROR(EINVAL);
00146     }
00147 
00148     /* initialize libspeex */
00149     s->enc_state = speex_encoder_init(mode);
00150     if (!s->enc_state) {
00151         av_log(avctx, AV_LOG_ERROR, "Error initializing libspeex\n");
00152         return -1;
00153     }
00154     speex_init_header(&s->header, avctx->sample_rate, avctx->channels, mode);
00155 
00156     /* rate control method and parameters */
00157     if (avctx->flags & CODEC_FLAG_QSCALE) {
00158         /* VBR */
00159         s->header.vbr = 1;
00160         speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR, &s->header.vbr);
00161         s->vbr_quality = av_clipf(avctx->global_quality / (float)FF_QP2LAMBDA,
00162                                   0.0f, 10.0f);
00163         speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR_QUALITY, &s->vbr_quality);
00164     } else {
00165         s->header.bitrate = avctx->bit_rate;
00166         if (avctx->bit_rate > 0) {
00167             /* CBR or ABR by bitrate */
00168             if (s->abr) {
00169                 speex_encoder_ctl(s->enc_state, SPEEX_SET_ABR,
00170                                   &s->header.bitrate);
00171                 speex_encoder_ctl(s->enc_state, SPEEX_GET_ABR,
00172                                   &s->header.bitrate);
00173             } else {
00174                 speex_encoder_ctl(s->enc_state, SPEEX_SET_BITRATE,
00175                                   &s->header.bitrate);
00176                 speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
00177                                   &s->header.bitrate);
00178             }
00179         } else {
00180             /* CBR by quality */
00181             speex_encoder_ctl(s->enc_state, SPEEX_SET_QUALITY,
00182                               &s->cbr_quality);
00183             speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE,
00184                               &s->header.bitrate);
00185         }
00186         /* stereo side information adds about 800 bps to the base bitrate */
00187         /* TODO: this should be calculated exactly */
00188         avctx->bit_rate = s->header.bitrate + (avctx->channels == 2 ? 800 : 0);
00189     }
00190 
00191     /* set encoding complexity */
00192     if (avctx->compression_level > FF_COMPRESSION_DEFAULT) {
00193         complexity = av_clip(avctx->compression_level, 0, 10);
00194         speex_encoder_ctl(s->enc_state, SPEEX_SET_COMPLEXITY, &complexity);
00195     }
00196     speex_encoder_ctl(s->enc_state, SPEEX_GET_COMPLEXITY, &complexity);
00197     avctx->compression_level = complexity;
00198 
00199     /* set packet size */
00200     avctx->frame_size = s->header.frame_size;
00201     s->header.frames_per_packet = s->frames_per_packet;
00202 
00203     /* set encoding delay */
00204     speex_encoder_ctl(s->enc_state, SPEEX_GET_LOOKAHEAD, &s->lookahead);
00205     s->next_pts = -s->lookahead;
00206 
00207     /* create header packet bytes from header struct */
00208     /* note: libspeex allocates the memory for header_data, which is freed
00209              below with speex_header_free() */
00210     header_data = speex_header_to_packet(&s->header, &header_size);
00211 
00212     /* allocate extradata and coded_frame */
00213     avctx->extradata   = av_malloc(header_size + FF_INPUT_BUFFER_PADDING_SIZE);
00214     avctx->coded_frame = avcodec_alloc_frame();
00215     if (!avctx->extradata || !avctx->coded_frame) {
00216         speex_header_free(header_data);
00217         speex_encoder_destroy(s->enc_state);
00218         av_log(avctx, AV_LOG_ERROR, "memory allocation error\n");
00219         return AVERROR(ENOMEM);
00220     }
00221 
00222     /* copy header packet to extradata */
00223     memcpy(avctx->extradata, header_data, header_size);
00224     avctx->extradata_size = header_size;
00225     speex_header_free(header_data);
00226 
00227     /* init libspeex bitwriter */
00228     speex_bits_init(&s->bits);
00229 
00230     print_enc_params(avctx, s);
00231     return 0;
00232 }
00233 
00234 static int encode_frame(AVCodecContext *avctx, uint8_t *frame, int buf_size,
00235                         void *data)
00236 {
00237     LibSpeexEncContext *s = avctx->priv_data;
00238     int16_t *samples      = data;
00239 
00240     if (data) {
00241         /* encode Speex frame */
00242         if (avctx->channels == 2)
00243             speex_encode_stereo_int(samples, s->header.frame_size, &s->bits);
00244         speex_encode_int(s->enc_state, samples, &s->bits);
00245         s->pkt_frame_count++;
00246         s->pkt_sample_count += avctx->frame_size;
00247     } else {
00248         /* handle end-of-stream */
00249         if (!s->pkt_frame_count)
00250             return 0;
00251         /* add extra terminator codes for unused frames in last packet */
00252         while (s->pkt_frame_count < s->frames_per_packet) {
00253             speex_bits_pack(&s->bits, 15, 5);
00254             s->pkt_frame_count++;
00255         }
00256     }
00257 
00258     /* write output if all frames for the packet have been encoded */
00259     if (s->pkt_frame_count == s->frames_per_packet) {
00260         s->pkt_frame_count = 0;
00261         avctx->coded_frame->pts =
00262             av_rescale_q(s->next_pts, (AVRational){ 1, avctx->sample_rate },
00263                          avctx->time_base);
00264         s->next_pts += s->pkt_sample_count;
00265         s->pkt_sample_count = 0;
00266         if (buf_size > speex_bits_nbytes(&s->bits)) {
00267             int ret = speex_bits_write(&s->bits, frame, buf_size);
00268             speex_bits_reset(&s->bits);
00269             return ret;
00270         } else {
00271             av_log(avctx, AV_LOG_ERROR, "output buffer too small");
00272             return AVERROR(EINVAL);
00273         }
00274     }
00275     return 0;
00276 }
00277 
00278 static av_cold int encode_close(AVCodecContext *avctx)
00279 {
00280     LibSpeexEncContext *s = avctx->priv_data;
00281 
00282     speex_bits_destroy(&s->bits);
00283     speex_encoder_destroy(s->enc_state);
00284 
00285     av_freep(&avctx->coded_frame);
00286     av_freep(&avctx->extradata);
00287 
00288     return 0;
00289 }
00290 
00291 #define OFFSET(x) offsetof(LibSpeexEncContext, x)
00292 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
00293 static const AVOption options[] = {
00294     { "abr",               "Use average bit rate",                      OFFSET(abr),               AV_OPT_TYPE_INT, { 0 }, 0,   1, AE },
00295     { "cbr_quality",       "Set quality value (0 to 10) for CBR",       OFFSET(cbr_quality),       AV_OPT_TYPE_INT, { 8 }, 0,  10, AE },
00296     { "frames_per_packet", "Number of frames to encode in each packet", OFFSET(frames_per_packet), AV_OPT_TYPE_INT, { 1 }, 1,   8, AE },
00297     { NULL },
00298 };
00299 
00300 static const AVClass class = {
00301     .class_name = "libspeex",
00302     .item_name  = av_default_item_name,
00303     .option     = options,
00304     .version    = LIBAVUTIL_VERSION_INT,
00305 };
00306 
00307 static const AVCodecDefault defaults[] = {
00308     { "b",                 "0" },
00309     { "compression_level", "3" },
00310     { NULL },
00311 };
00312 
00313 AVCodec ff_libspeex_encoder = {
00314     .name           = "libspeex",
00315     .type           = AVMEDIA_TYPE_AUDIO,
00316     .id             = CODEC_ID_SPEEX,
00317     .priv_data_size = sizeof(LibSpeexEncContext),
00318     .init           = encode_init,
00319     .encode         = encode_frame,
00320     .close          = encode_close,
00321     .capabilities   = CODEC_CAP_DELAY,
00322     .sample_fmts    = (const enum SampleFormat[]){ AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE },
00323     .long_name      = NULL_IF_CONFIG_SMALL("libspeex Speex"),
00324     .priv_class     = &class,
00325     .defaults       = defaults,
00326 };