| 1 | /* |
| 2 | * Copyright (c) 2002 Mark Hills <mark@pogo.org.uk> |
| 3 | * |
| 4 | * This file is part of FFmpeg. |
| 5 | * |
| 6 | * FFmpeg is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU Lesser General Public |
| 8 | * License as published by the Free Software Foundation; either |
| 9 | * version 2.1 of the License, or (at your option) any later version. |
| 10 | * |
| 11 | * FFmpeg is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 14 | * Lesser General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU Lesser General Public |
| 17 | * License along with FFmpeg; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 19 | */ |
| 20 | |
| 21 | #include <vorbis/vorbisenc.h> |
| 22 | |
| 23 | #include "libavutil/avassert.h" |
| 24 | #include "libavutil/fifo.h" |
| 25 | #include "libavutil/opt.h" |
| 26 | #include "avcodec.h" |
| 27 | #include "audio_frame_queue.h" |
| 28 | #include "internal.h" |
| 29 | #include "vorbis.h" |
| 30 | #include "vorbis_parser.h" |
| 31 | |
| 32 | |
| 33 | /* Number of samples the user should send in each call. |
| 34 | * This value is used because it is the LCD of all possible frame sizes, so |
| 35 | * an output packet will always start at the same point as one of the input |
| 36 | * packets. |
| 37 | */ |
| 38 | #define LIBVORBIS_FRAME_SIZE 64 |
| 39 | |
| 40 | #define BUFFER_SIZE (1024 * 64) |
| 41 | |
| 42 | typedef struct LibvorbisEncContext { |
| 43 | AVClass *av_class; /**< class for AVOptions */ |
| 44 | vorbis_info vi; /**< vorbis_info used during init */ |
| 45 | vorbis_dsp_state vd; /**< DSP state used for analysis */ |
| 46 | vorbis_block vb; /**< vorbis_block used for analysis */ |
| 47 | AVFifoBuffer *pkt_fifo; /**< output packet buffer */ |
| 48 | int eof; /**< end-of-file flag */ |
| 49 | int dsp_initialized; /**< vd has been initialized */ |
| 50 | vorbis_comment vc; /**< VorbisComment info */ |
| 51 | double iblock; /**< impulse block bias option */ |
| 52 | VorbisParseContext vp; /**< parse context to get durations */ |
| 53 | AudioFrameQueue afq; /**< frame queue for timestamps */ |
| 54 | } LibvorbisEncContext; |
| 55 | |
| 56 | static const AVOption options[] = { |
| 57 | { "iblock", "Sets the impulse block bias", offsetof(LibvorbisEncContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM }, |
| 58 | { NULL } |
| 59 | }; |
| 60 | |
| 61 | static const AVCodecDefault defaults[] = { |
| 62 | { "b", "0" }, |
| 63 | { NULL }, |
| 64 | }; |
| 65 | |
| 66 | static const AVClass vorbis_class = { |
| 67 | .class_name = "libvorbis", |
| 68 | .item_name = av_default_item_name, |
| 69 | .option = options, |
| 70 | .version = LIBAVUTIL_VERSION_INT, |
| 71 | }; |
| 72 | |
| 73 | static int vorbis_error_to_averror(int ov_err) |
| 74 | { |
| 75 | switch (ov_err) { |
| 76 | case OV_EFAULT: return AVERROR_BUG; |
| 77 | case OV_EINVAL: return AVERROR(EINVAL); |
| 78 | case OV_EIMPL: return AVERROR(EINVAL); |
| 79 | default: return AVERROR_UNKNOWN; |
| 80 | } |
| 81 | } |
| 82 | |
| 83 | static av_cold int libvorbis_setup(vorbis_info *vi, AVCodecContext *avctx) |
| 84 | { |
| 85 | LibvorbisEncContext *s = avctx->priv_data; |
| 86 | double cfreq; |
| 87 | int ret; |
| 88 | |
| 89 | if (avctx->flags & CODEC_FLAG_QSCALE || !avctx->bit_rate) { |
| 90 | /* variable bitrate |
| 91 | * NOTE: we use the oggenc range of -1 to 10 for global_quality for |
| 92 | * user convenience, but libvorbis uses -0.1 to 1.0. |
| 93 | */ |
| 94 | float q = avctx->global_quality / (float)FF_QP2LAMBDA; |
| 95 | /* default to 3 if the user did not set quality or bitrate */ |
| 96 | if (!(avctx->flags & CODEC_FLAG_QSCALE)) |
| 97 | q = 3.0; |
| 98 | if ((ret = vorbis_encode_setup_vbr(vi, avctx->channels, |
| 99 | avctx->sample_rate, |
| 100 | q / 10.0))) |
| 101 | goto error; |
| 102 | } else { |
| 103 | int minrate = avctx->rc_min_rate > 0 ? avctx->rc_min_rate : -1; |
| 104 | int maxrate = avctx->rc_max_rate > 0 ? avctx->rc_max_rate : -1; |
| 105 | |
| 106 | /* average bitrate */ |
| 107 | if ((ret = vorbis_encode_setup_managed(vi, avctx->channels, |
| 108 | avctx->sample_rate, maxrate, |
| 109 | avctx->bit_rate, minrate))) |
| 110 | goto error; |
| 111 | |
| 112 | /* variable bitrate by estimate, disable slow rate management */ |
| 113 | if (minrate == -1 && maxrate == -1) |
| 114 | if ((ret = vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL))) |
| 115 | goto error; /* should not happen */ |
| 116 | } |
| 117 | |
| 118 | /* cutoff frequency */ |
| 119 | if (avctx->cutoff > 0) { |
| 120 | cfreq = avctx->cutoff / 1000.0; |
| 121 | if ((ret = vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq))) |
| 122 | goto error; /* should not happen */ |
| 123 | } |
| 124 | |
| 125 | /* impulse block bias */ |
| 126 | if (s->iblock) { |
| 127 | if ((ret = vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &s->iblock))) |
| 128 | goto error; |
| 129 | } |
| 130 | |
| 131 | if (avctx->channels == 3 && |
| 132 | avctx->channel_layout != (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER) || |
| 133 | avctx->channels == 4 && |
| 134 | avctx->channel_layout != AV_CH_LAYOUT_2_2 && |
| 135 | avctx->channel_layout != AV_CH_LAYOUT_QUAD || |
| 136 | avctx->channels == 5 && |
| 137 | avctx->channel_layout != AV_CH_LAYOUT_5POINT0 && |
| 138 | avctx->channel_layout != AV_CH_LAYOUT_5POINT0_BACK || |
| 139 | avctx->channels == 6 && |
| 140 | avctx->channel_layout != AV_CH_LAYOUT_5POINT1 && |
| 141 | avctx->channel_layout != AV_CH_LAYOUT_5POINT1_BACK || |
| 142 | avctx->channels == 7 && |
| 143 | avctx->channel_layout != (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER) || |
| 144 | avctx->channels == 8 && |
| 145 | avctx->channel_layout != AV_CH_LAYOUT_7POINT1) { |
| 146 | if (avctx->channel_layout) { |
| 147 | char name[32]; |
| 148 | av_get_channel_layout_string(name, sizeof(name), avctx->channels, |
| 149 | avctx->channel_layout); |
| 150 | av_log(avctx, AV_LOG_ERROR, "%s not supported by Vorbis: " |
| 151 | "output stream will have incorrect " |
| 152 | "channel layout.\n", name); |
| 153 | } else { |
| 154 | av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The encoder " |
| 155 | "will use Vorbis channel layout for " |
| 156 | "%d channels.\n", avctx->channels); |
| 157 | } |
| 158 | } |
| 159 | |
| 160 | if ((ret = vorbis_encode_setup_init(vi))) |
| 161 | goto error; |
| 162 | |
| 163 | return 0; |
| 164 | error: |
| 165 | return vorbis_error_to_averror(ret); |
| 166 | } |
| 167 | |
| 168 | /* How many bytes are needed for a buffer of length 'l' */ |
| 169 | static int xiph_len(int l) |
| 170 | { |
| 171 | return 1 + l / 255 + l; |
| 172 | } |
| 173 | |
| 174 | static av_cold int libvorbis_encode_close(AVCodecContext *avctx) |
| 175 | { |
| 176 | LibvorbisEncContext *s = avctx->priv_data; |
| 177 | |
| 178 | /* notify vorbisenc this is EOF */ |
| 179 | if (s->dsp_initialized) |
| 180 | vorbis_analysis_wrote(&s->vd, 0); |
| 181 | |
| 182 | vorbis_block_clear(&s->vb); |
| 183 | vorbis_dsp_clear(&s->vd); |
| 184 | vorbis_info_clear(&s->vi); |
| 185 | |
| 186 | av_fifo_freep(&s->pkt_fifo); |
| 187 | ff_af_queue_close(&s->afq); |
| 188 | av_freep(&avctx->extradata); |
| 189 | |
| 190 | return 0; |
| 191 | } |
| 192 | |
| 193 | static av_cold int libvorbis_encode_init(AVCodecContext *avctx) |
| 194 | { |
| 195 | LibvorbisEncContext *s = avctx->priv_data; |
| 196 | ogg_packet header, header_comm, header_code; |
| 197 | uint8_t *p; |
| 198 | unsigned int offset; |
| 199 | int ret; |
| 200 | |
| 201 | vorbis_info_init(&s->vi); |
| 202 | if ((ret = libvorbis_setup(&s->vi, avctx))) { |
| 203 | av_log(avctx, AV_LOG_ERROR, "encoder setup failed\n"); |
| 204 | goto error; |
| 205 | } |
| 206 | if ((ret = vorbis_analysis_init(&s->vd, &s->vi))) { |
| 207 | av_log(avctx, AV_LOG_ERROR, "analysis init failed\n"); |
| 208 | ret = vorbis_error_to_averror(ret); |
| 209 | goto error; |
| 210 | } |
| 211 | s->dsp_initialized = 1; |
| 212 | if ((ret = vorbis_block_init(&s->vd, &s->vb))) { |
| 213 | av_log(avctx, AV_LOG_ERROR, "dsp init failed\n"); |
| 214 | ret = vorbis_error_to_averror(ret); |
| 215 | goto error; |
| 216 | } |
| 217 | |
| 218 | vorbis_comment_init(&s->vc); |
| 219 | if (!(avctx->flags & CODEC_FLAG_BITEXACT)) |
| 220 | vorbis_comment_add_tag(&s->vc, "encoder", LIBAVCODEC_IDENT); |
| 221 | |
| 222 | if ((ret = vorbis_analysis_headerout(&s->vd, &s->vc, &header, &header_comm, |
| 223 | &header_code))) { |
| 224 | ret = vorbis_error_to_averror(ret); |
| 225 | goto error; |
| 226 | } |
| 227 | |
| 228 | avctx->extradata_size = 1 + xiph_len(header.bytes) + |
| 229 | xiph_len(header_comm.bytes) + |
| 230 | header_code.bytes; |
| 231 | p = avctx->extradata = av_malloc(avctx->extradata_size + |
| 232 | FF_INPUT_BUFFER_PADDING_SIZE); |
| 233 | if (!p) { |
| 234 | ret = AVERROR(ENOMEM); |
| 235 | goto error; |
| 236 | } |
| 237 | p[0] = 2; |
| 238 | offset = 1; |
| 239 | offset += av_xiphlacing(&p[offset], header.bytes); |
| 240 | offset += av_xiphlacing(&p[offset], header_comm.bytes); |
| 241 | memcpy(&p[offset], header.packet, header.bytes); |
| 242 | offset += header.bytes; |
| 243 | memcpy(&p[offset], header_comm.packet, header_comm.bytes); |
| 244 | offset += header_comm.bytes; |
| 245 | memcpy(&p[offset], header_code.packet, header_code.bytes); |
| 246 | offset += header_code.bytes; |
| 247 | av_assert0(offset == avctx->extradata_size); |
| 248 | |
| 249 | if ((ret = avpriv_vorbis_parse_extradata(avctx, &s->vp)) < 0) { |
| 250 | av_log(avctx, AV_LOG_ERROR, "invalid extradata\n"); |
| 251 | return ret; |
| 252 | } |
| 253 | |
| 254 | vorbis_comment_clear(&s->vc); |
| 255 | |
| 256 | avctx->frame_size = LIBVORBIS_FRAME_SIZE; |
| 257 | ff_af_queue_init(avctx, &s->afq); |
| 258 | |
| 259 | s->pkt_fifo = av_fifo_alloc(BUFFER_SIZE); |
| 260 | if (!s->pkt_fifo) { |
| 261 | ret = AVERROR(ENOMEM); |
| 262 | goto error; |
| 263 | } |
| 264 | |
| 265 | return 0; |
| 266 | error: |
| 267 | libvorbis_encode_close(avctx); |
| 268 | return ret; |
| 269 | } |
| 270 | |
| 271 | static int libvorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, |
| 272 | const AVFrame *frame, int *got_packet_ptr) |
| 273 | { |
| 274 | LibvorbisEncContext *s = avctx->priv_data; |
| 275 | ogg_packet op; |
| 276 | int ret, duration; |
| 277 | |
| 278 | /* send samples to libvorbis */ |
| 279 | if (frame) { |
| 280 | const int samples = frame->nb_samples; |
| 281 | float **buffer; |
| 282 | int c, channels = s->vi.channels; |
| 283 | |
| 284 | buffer = vorbis_analysis_buffer(&s->vd, samples); |
| 285 | for (c = 0; c < channels; c++) { |
| 286 | int co = (channels > 8) ? c : |
| 287 | ff_vorbis_encoding_channel_layout_offsets[channels - 1][c]; |
| 288 | memcpy(buffer[c], frame->extended_data[co], |
| 289 | samples * sizeof(*buffer[c])); |
| 290 | } |
| 291 | if ((ret = vorbis_analysis_wrote(&s->vd, samples)) < 0) { |
| 292 | av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n"); |
| 293 | return vorbis_error_to_averror(ret); |
| 294 | } |
| 295 | if ((ret = ff_af_queue_add(&s->afq, frame)) < 0) |
| 296 | return ret; |
| 297 | } else { |
| 298 | if (!s->eof && s->afq.frame_alloc) |
| 299 | if ((ret = vorbis_analysis_wrote(&s->vd, 0)) < 0) { |
| 300 | av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n"); |
| 301 | return vorbis_error_to_averror(ret); |
| 302 | } |
| 303 | s->eof = 1; |
| 304 | } |
| 305 | |
| 306 | /* retrieve available packets from libvorbis */ |
| 307 | while ((ret = vorbis_analysis_blockout(&s->vd, &s->vb)) == 1) { |
| 308 | if ((ret = vorbis_analysis(&s->vb, NULL)) < 0) |
| 309 | break; |
| 310 | if ((ret = vorbis_bitrate_addblock(&s->vb)) < 0) |
| 311 | break; |
| 312 | |
| 313 | /* add any available packets to the output packet buffer */ |
| 314 | while ((ret = vorbis_bitrate_flushpacket(&s->vd, &op)) == 1) { |
| 315 | if (av_fifo_space(s->pkt_fifo) < sizeof(ogg_packet) + op.bytes) { |
| 316 | av_log(avctx, AV_LOG_ERROR, "packet buffer is too small\n"); |
| 317 | return AVERROR_BUG; |
| 318 | } |
| 319 | av_fifo_generic_write(s->pkt_fifo, &op, sizeof(ogg_packet), NULL); |
| 320 | av_fifo_generic_write(s->pkt_fifo, op.packet, op.bytes, NULL); |
| 321 | } |
| 322 | if (ret < 0) { |
| 323 | av_log(avctx, AV_LOG_ERROR, "error getting available packets\n"); |
| 324 | break; |
| 325 | } |
| 326 | } |
| 327 | if (ret < 0) { |
| 328 | av_log(avctx, AV_LOG_ERROR, "error getting available packets\n"); |
| 329 | return vorbis_error_to_averror(ret); |
| 330 | } |
| 331 | |
| 332 | /* check for available packets */ |
| 333 | if (av_fifo_size(s->pkt_fifo) < sizeof(ogg_packet)) |
| 334 | return 0; |
| 335 | |
| 336 | av_fifo_generic_read(s->pkt_fifo, &op, sizeof(ogg_packet), NULL); |
| 337 | |
| 338 | if ((ret = ff_alloc_packet2(avctx, avpkt, op.bytes)) < 0) |
| 339 | return ret; |
| 340 | av_fifo_generic_read(s->pkt_fifo, avpkt->data, op.bytes, NULL); |
| 341 | |
| 342 | avpkt->pts = ff_samples_to_time_base(avctx, op.granulepos); |
| 343 | |
| 344 | duration = avpriv_vorbis_parse_frame(&s->vp, avpkt->data, avpkt->size); |
| 345 | if (duration > 0) { |
| 346 | /* we do not know encoder delay until we get the first packet from |
| 347 | * libvorbis, so we have to update the AudioFrameQueue counts */ |
| 348 | if (!avctx->delay && s->afq.frames) { |
| 349 | avctx->delay = duration; |
| 350 | av_assert0(!s->afq.remaining_delay); |
| 351 | s->afq.frames->duration += duration; |
| 352 | if (s->afq.frames->pts != AV_NOPTS_VALUE) |
| 353 | s->afq.frames->pts -= duration; |
| 354 | s->afq.remaining_samples += duration; |
| 355 | } |
| 356 | ff_af_queue_remove(&s->afq, duration, &avpkt->pts, &avpkt->duration); |
| 357 | } |
| 358 | |
| 359 | *got_packet_ptr = 1; |
| 360 | return 0; |
| 361 | } |
| 362 | |
| 363 | AVCodec ff_libvorbis_encoder = { |
| 364 | .name = "libvorbis", |
| 365 | .long_name = NULL_IF_CONFIG_SMALL("libvorbis"), |
| 366 | .type = AVMEDIA_TYPE_AUDIO, |
| 367 | .id = AV_CODEC_ID_VORBIS, |
| 368 | .priv_data_size = sizeof(LibvorbisEncContext), |
| 369 | .init = libvorbis_encode_init, |
| 370 | .encode2 = libvorbis_encode_frame, |
| 371 | .close = libvorbis_encode_close, |
| 372 | .capabilities = CODEC_CAP_DELAY, |
| 373 | .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP, |
| 374 | AV_SAMPLE_FMT_NONE }, |
| 375 | .priv_class = &vorbis_class, |
| 376 | .defaults = defaults, |
| 377 | }; |