00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00027 #include <vorbis/vorbisenc.h>
00028
00029 #include "libavutil/opt.h"
00030 #include "avcodec.h"
00031 #include "bytestream.h"
00032 #include "vorbis.h"
00033 #include "libavutil/mathematics.h"
00034
00035 #undef NDEBUG
00036 #include <assert.h>
00037
00038 #define OGGVORBIS_FRAME_SIZE 64
00039
00040 #define BUFFER_SIZE (1024 * 64)
00041
00042 typedef struct OggVorbisContext {
00043 AVClass *av_class;
00044 vorbis_info vi;
00045 vorbis_dsp_state vd;
00046 vorbis_block vb;
00047 uint8_t buffer[BUFFER_SIZE];
00048 int buffer_index;
00049 int eof;
00050
00051
00052 vorbis_comment vc;
00053 ogg_packet op;
00054
00055 double iblock;
00056 } OggVorbisContext;
00057
00058 static const AVOption options[] = {
00059 { "iblock", "Sets the impulse block bias", offsetof(OggVorbisContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
00060 { NULL }
00061 };
00062 static const AVClass class = { "libvorbis", av_default_item_name, options, LIBAVUTIL_VERSION_INT };
00063
00064 static av_cold int oggvorbis_init_encoder(vorbis_info *vi, AVCodecContext *avccontext)
00065 {
00066 OggVorbisContext *context = avccontext->priv_data;
00067 double cfreq;
00068
00069 if (avccontext->flags & CODEC_FLAG_QSCALE) {
00070
00071 if (vorbis_encode_setup_vbr(vi, avccontext->channels,
00072 avccontext->sample_rate,
00073 avccontext->global_quality / (float)FF_QP2LAMBDA / 10.0))
00074 return -1;
00075 } else {
00076 int minrate = avccontext->rc_min_rate > 0 ? avccontext->rc_min_rate : -1;
00077 int maxrate = avccontext->rc_min_rate > 0 ? avccontext->rc_max_rate : -1;
00078
00079
00080 if (vorbis_encode_setup_managed(vi, avccontext->channels,
00081 avccontext->sample_rate, minrate,
00082 avccontext->bit_rate, maxrate))
00083 return -1;
00084
00085
00086 if (minrate == -1 && maxrate == -1)
00087 if (vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL))
00088 return -1;
00089 }
00090
00091
00092 if (avccontext->cutoff > 0) {
00093 cfreq = avccontext->cutoff / 1000.0;
00094 if (vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq))
00095 return -1;
00096 }
00097
00098 if (context->iblock) {
00099 vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &context->iblock);
00100 }
00101
00102 return vorbis_encode_setup_init(vi);
00103 }
00104
00105
00106 static int xiph_len(int l)
00107 {
00108 return 1 + l / 255 + l;
00109 }
00110
00111 static av_cold int oggvorbis_encode_init(AVCodecContext *avccontext)
00112 {
00113 OggVorbisContext *context = avccontext->priv_data;
00114 ogg_packet header, header_comm, header_code;
00115 uint8_t *p;
00116 unsigned int offset;
00117
00118 vorbis_info_init(&context->vi);
00119 if (oggvorbis_init_encoder(&context->vi, avccontext) < 0) {
00120 av_log(avccontext, AV_LOG_ERROR, "oggvorbis_encode_init: init_encoder failed\n");
00121 return -1;
00122 }
00123 vorbis_analysis_init(&context->vd, &context->vi);
00124 vorbis_block_init(&context->vd, &context->vb);
00125
00126 vorbis_comment_init(&context->vc);
00127 vorbis_comment_add_tag(&context->vc, "encoder", LIBAVCODEC_IDENT);
00128
00129 vorbis_analysis_headerout(&context->vd, &context->vc, &header,
00130 &header_comm, &header_code);
00131
00132 avccontext->extradata_size =
00133 1 + xiph_len(header.bytes) + xiph_len(header_comm.bytes) +
00134 header_code.bytes;
00135 p = avccontext->extradata =
00136 av_malloc(avccontext->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
00137 p[0] = 2;
00138 offset = 1;
00139 offset += av_xiphlacing(&p[offset], header.bytes);
00140 offset += av_xiphlacing(&p[offset], header_comm.bytes);
00141 memcpy(&p[offset], header.packet, header.bytes);
00142 offset += header.bytes;
00143 memcpy(&p[offset], header_comm.packet, header_comm.bytes);
00144 offset += header_comm.bytes;
00145 memcpy(&p[offset], header_code.packet, header_code.bytes);
00146 offset += header_code.bytes;
00147 assert(offset == avccontext->extradata_size);
00148
00149 #if 0
00150 vorbis_block_clear(&context->vb);
00151 vorbis_dsp_clear(&context->vd);
00152 vorbis_info_clear(&context->vi);
00153 #endif
00154 vorbis_comment_clear(&context->vc);
00155
00156 avccontext->frame_size = OGGVORBIS_FRAME_SIZE;
00157
00158 avccontext->coded_frame = avcodec_alloc_frame();
00159 avccontext->coded_frame->key_frame = 1;
00160
00161 return 0;
00162 }
00163
00164 static int oggvorbis_encode_frame(AVCodecContext *avccontext,
00165 unsigned char *packets,
00166 int buf_size, void *data)
00167 {
00168 OggVorbisContext *context = avccontext->priv_data;
00169 ogg_packet op;
00170 signed short *audio = data;
00171 int l;
00172
00173 if (data) {
00174 const int samples = avccontext->frame_size;
00175 float **buffer;
00176 int c, channels = context->vi.channels;
00177
00178 buffer = vorbis_analysis_buffer(&context->vd, samples);
00179 for (c = 0; c < channels; c++) {
00180 int co = (channels > 8) ? c :
00181 ff_vorbis_encoding_channel_layout_offsets[channels - 1][c];
00182 for (l = 0; l < samples; l++)
00183 buffer[c][l] = audio[l * channels + co] / 32768.f;
00184 }
00185 vorbis_analysis_wrote(&context->vd, samples);
00186 } else {
00187 if (!context->eof)
00188 vorbis_analysis_wrote(&context->vd, 0);
00189 context->eof = 1;
00190 }
00191
00192 while (vorbis_analysis_blockout(&context->vd, &context->vb) == 1) {
00193 vorbis_analysis(&context->vb, NULL);
00194 vorbis_bitrate_addblock(&context->vb);
00195
00196 while (vorbis_bitrate_flushpacket(&context->vd, &op)) {
00197
00198
00199 if (op.bytes == 1 && op.e_o_s)
00200 continue;
00201 if (context->buffer_index + sizeof(ogg_packet) + op.bytes > BUFFER_SIZE) {
00202 av_log(avccontext, AV_LOG_ERROR, "libvorbis: buffer overflow.");
00203 return -1;
00204 }
00205 memcpy(context->buffer + context->buffer_index, &op, sizeof(ogg_packet));
00206 context->buffer_index += sizeof(ogg_packet);
00207 memcpy(context->buffer + context->buffer_index, op.packet, op.bytes);
00208 context->buffer_index += op.bytes;
00209
00210 }
00211 }
00212
00213 l = 0;
00214 if (context->buffer_index) {
00215 ogg_packet *op2 = (ogg_packet *)context->buffer;
00216 op2->packet = context->buffer + sizeof(ogg_packet);
00217
00218 l = op2->bytes;
00219 avccontext->coded_frame->pts = av_rescale_q(op2->granulepos, (AVRational) { 1, avccontext->sample_rate }, avccontext->time_base);
00220
00221
00222 if (l > buf_size) {
00223 av_log(avccontext, AV_LOG_ERROR, "libvorbis: buffer overflow.");
00224 return -1;
00225 }
00226
00227 memcpy(packets, op2->packet, l);
00228 context->buffer_index -= l + sizeof(ogg_packet);
00229 memmove(context->buffer, context->buffer + l + sizeof(ogg_packet), context->buffer_index);
00230
00231 }
00232
00233 return l;
00234 }
00235
00236 static av_cold int oggvorbis_encode_close(AVCodecContext *avccontext)
00237 {
00238 OggVorbisContext *context = avccontext->priv_data;
00239
00240
00241 vorbis_analysis_wrote(&context->vd, 0);
00242
00243 vorbis_block_clear(&context->vb);
00244 vorbis_dsp_clear(&context->vd);
00245 vorbis_info_clear(&context->vi);
00246
00247 av_freep(&avccontext->coded_frame);
00248 av_freep(&avccontext->extradata);
00249
00250 return 0;
00251 }
00252
00253 AVCodec ff_libvorbis_encoder = {
00254 .name = "libvorbis",
00255 .type = AVMEDIA_TYPE_AUDIO,
00256 .id = CODEC_ID_VORBIS,
00257 .priv_data_size = sizeof(OggVorbisContext),
00258 .init = oggvorbis_encode_init,
00259 .encode = oggvorbis_encode_frame,
00260 .close = oggvorbis_encode_close,
00261 .capabilities = CODEC_CAP_DELAY,
00262 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE },
00263 .long_name = NULL_IF_CONFIG_SMALL("libvorbis Vorbis"),
00264 .priv_class = &class,
00265 };