1/* 2 * Copyright (c) 2011 Stefano Sabatini 3 * Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com> 4 * 5 * This file is part of FFmpeg. 6 * 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. 11 * 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. 16 * 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 20 */ 21 22/** 23 * @file 24 * audio volume filter 25 */ 26 27#include "libavutil/channel_layout.h" 28#include "libavutil/common.h" 29#include "libavutil/eval.h" 30#include "libavutil/float_dsp.h" 31#include "libavutil/intreadwrite.h" 32#include "libavutil/opt.h" 33#include "libavutil/replaygain.h" 34 35#include "audio.h" 36#include "avfilter.h" 37#include "formats.h" 38#include "internal.h" 39#include "af_volume.h" 40 41static const char *precision_str[] = { 42 "fixed", "float", "double" 43}; 44 45static const char *const var_names[] = { 46 "n", ///< frame number (starting at zero) 47 "nb_channels", ///< number of channels 48 "nb_consumed_samples", ///< number of samples consumed by the filter 49 "nb_samples", ///< number of samples in the current frame 50 "pos", ///< position in the file of the frame 51 "pts", ///< frame presentation timestamp 52 "sample_rate", ///< sample rate 53 "startpts", ///< PTS at start of stream 54 "startt", ///< time at start of stream 55 "t", ///< time in the file of the frame 56 "tb", ///< timebase 57 "volume", ///< last set value 58 NULL 59}; 60 61#define OFFSET(x) offsetof(VolumeContext, x) 62#define A AV_OPT_FLAG_AUDIO_PARAM 63#define F AV_OPT_FLAG_FILTERING_PARAM 64 65static const AVOption volume_options[] = { 66 { "volume", "set volume adjustment expression", 67 OFFSET(volume_expr), AV_OPT_TYPE_STRING, { .str = "1.0" }, .flags = A|F }, 68 { "precision", "select mathematical precision", 69 OFFSET(precision), AV_OPT_TYPE_INT, { .i64 = PRECISION_FLOAT }, PRECISION_FIXED, PRECISION_DOUBLE, A|F, "precision" }, 70 { "fixed", "select 8-bit fixed-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_FIXED }, INT_MIN, INT_MAX, A|F, "precision" }, 71 { "float", "select 32-bit floating-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_FLOAT }, INT_MIN, INT_MAX, A|F, "precision" }, 72 { "double", "select 64-bit floating-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_DOUBLE }, INT_MIN, INT_MAX, A|F, "precision" }, 73 { "eval", "specify when to evaluate expressions", OFFSET(eval_mode), AV_OPT_TYPE_INT, {.i64 = EVAL_MODE_ONCE}, 0, EVAL_MODE_NB-1, .flags = A|F, "eval" }, 74 { "once", "eval volume expression once", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_ONCE}, .flags = A|F, .unit = "eval" }, 75 { "frame", "eval volume expression per-frame", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_FRAME}, .flags = A|F, .unit = "eval" }, 76 { "replaygain", "Apply replaygain side data when present", 77 OFFSET(replaygain), AV_OPT_TYPE_INT, { .i64 = REPLAYGAIN_DROP }, REPLAYGAIN_DROP, REPLAYGAIN_ALBUM, A, "replaygain" }, 78 { "drop", "replaygain side data is dropped", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_DROP }, 0, 0, A, "replaygain" }, 79 { "ignore", "replaygain side data is ignored", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_IGNORE }, 0, 0, A, "replaygain" }, 80 { "track", "track gain is preferred", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_TRACK }, 0, 0, A, "replaygain" }, 81 { "album", "album gain is preferred", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_ALBUM }, 0, 0, A, "replaygain" }, 82 { "replaygain_preamp", "Apply replaygain pre-amplification", 83 OFFSET(replaygain_preamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.0 }, -15.0, 15.0, A }, 84 { "replaygain_noclip", "Apply replaygain clipping prevention", 85 OFFSET(replaygain_noclip), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, A }, 86 { NULL }, 87}; 88 89AVFILTER_DEFINE_CLASS(volume); 90 91static int set_expr(AVExpr **pexpr, const char *expr, void *log_ctx) 92{ 93 int ret; 94 AVExpr *old = NULL; 95 96 if (*pexpr) 97 old = *pexpr; 98 ret = av_expr_parse(pexpr, expr, var_names, 99 NULL, NULL, NULL, NULL, 0, log_ctx); 100 if (ret < 0) { 101 av_log(log_ctx, AV_LOG_ERROR, 102 "Error when evaluating the volume expression '%s'\n", expr); 103 *pexpr = old; 104 return ret; 105 } 106 107 av_expr_free(old); 108 return 0; 109} 110 111static av_cold int init(AVFilterContext *ctx) 112{ 113 VolumeContext *vol = ctx->priv; 114 return set_expr(&vol->volume_pexpr, vol->volume_expr, ctx); 115} 116 117static av_cold void uninit(AVFilterContext *ctx) 118{ 119 VolumeContext *vol = ctx->priv; 120 av_expr_free(vol->volume_pexpr); 121 av_opt_free(vol); 122} 123 124static int query_formats(AVFilterContext *ctx) 125{ 126 VolumeContext *vol = ctx->priv; 127 AVFilterFormats *formats = NULL; 128 AVFilterChannelLayouts *layouts; 129 static const enum AVSampleFormat sample_fmts[][7] = { 130 [PRECISION_FIXED] = { 131 AV_SAMPLE_FMT_U8, 132 AV_SAMPLE_FMT_U8P, 133 AV_SAMPLE_FMT_S16, 134 AV_SAMPLE_FMT_S16P, 135 AV_SAMPLE_FMT_S32, 136 AV_SAMPLE_FMT_S32P, 137 AV_SAMPLE_FMT_NONE 138 }, 139 [PRECISION_FLOAT] = { 140 AV_SAMPLE_FMT_FLT, 141 AV_SAMPLE_FMT_FLTP, 142 AV_SAMPLE_FMT_NONE 143 }, 144 [PRECISION_DOUBLE] = { 145 AV_SAMPLE_FMT_DBL, 146 AV_SAMPLE_FMT_DBLP, 147 AV_SAMPLE_FMT_NONE 148 } 149 }; 150 151 layouts = ff_all_channel_counts(); 152 if (!layouts) 153 return AVERROR(ENOMEM); 154 ff_set_common_channel_layouts(ctx, layouts); 155 156 formats = ff_make_format_list(sample_fmts[vol->precision]); 157 if (!formats) 158 return AVERROR(ENOMEM); 159 ff_set_common_formats(ctx, formats); 160 161 formats = ff_all_samplerates(); 162 if (!formats) 163 return AVERROR(ENOMEM); 164 ff_set_common_samplerates(ctx, formats); 165 166 return 0; 167} 168 169static inline void scale_samples_u8(uint8_t *dst, const uint8_t *src, 170 int nb_samples, int volume) 171{ 172 int i; 173 for (i = 0; i < nb_samples; i++) 174 dst[i] = av_clip_uint8(((((int64_t)src[i] - 128) * volume + 128) >> 8) + 128); 175} 176 177static inline void scale_samples_u8_small(uint8_t *dst, const uint8_t *src, 178 int nb_samples, int volume) 179{ 180 int i; 181 for (i = 0; i < nb_samples; i++) 182 dst[i] = av_clip_uint8((((src[i] - 128) * volume + 128) >> 8) + 128); 183} 184 185static inline void scale_samples_s16(uint8_t *dst, const uint8_t *src, 186 int nb_samples, int volume) 187{ 188 int i; 189 int16_t *smp_dst = (int16_t *)dst; 190 const int16_t *smp_src = (const int16_t *)src; 191 for (i = 0; i < nb_samples; i++) 192 smp_dst[i] = av_clip_int16(((int64_t)smp_src[i] * volume + 128) >> 8); 193} 194 195static inline void scale_samples_s16_small(uint8_t *dst, const uint8_t *src, 196 int nb_samples, int volume) 197{ 198 int i; 199 int16_t *smp_dst = (int16_t *)dst; 200 const int16_t *smp_src = (const int16_t *)src; 201 for (i = 0; i < nb_samples; i++) 202 smp_dst[i] = av_clip_int16((smp_src[i] * volume + 128) >> 8); 203} 204 205static inline void scale_samples_s32(uint8_t *dst, const uint8_t *src, 206 int nb_samples, int volume) 207{ 208 int i; 209 int32_t *smp_dst = (int32_t *)dst; 210 const int32_t *smp_src = (const int32_t *)src; 211 for (i = 0; i < nb_samples; i++) 212 smp_dst[i] = av_clipl_int32((((int64_t)smp_src[i] * volume + 128) >> 8)); 213} 214 215static av_cold void volume_init(VolumeContext *vol) 216{ 217 vol->samples_align = 1; 218 219 switch (av_get_packed_sample_fmt(vol->sample_fmt)) { 220 case AV_SAMPLE_FMT_U8: 221 if (vol->volume_i < 0x1000000) 222 vol->scale_samples = scale_samples_u8_small; 223 else 224 vol->scale_samples = scale_samples_u8; 225 break; 226 case AV_SAMPLE_FMT_S16: 227 if (vol->volume_i < 0x10000) 228 vol->scale_samples = scale_samples_s16_small; 229 else 230 vol->scale_samples = scale_samples_s16; 231 break; 232 case AV_SAMPLE_FMT_S32: 233 vol->scale_samples = scale_samples_s32; 234 break; 235 case AV_SAMPLE_FMT_FLT: 236 avpriv_float_dsp_init(&vol->fdsp, 0); 237 vol->samples_align = 4; 238 break; 239 case AV_SAMPLE_FMT_DBL: 240 avpriv_float_dsp_init(&vol->fdsp, 0); 241 vol->samples_align = 8; 242 break; 243 } 244 245 if (ARCH_X86) 246 ff_volume_init_x86(vol); 247} 248 249static int set_volume(AVFilterContext *ctx) 250{ 251 VolumeContext *vol = ctx->priv; 252 253 vol->volume = av_expr_eval(vol->volume_pexpr, vol->var_values, NULL); 254 if (isnan(vol->volume)) { 255 if (vol->eval_mode == EVAL_MODE_ONCE) { 256 av_log(ctx, AV_LOG_ERROR, "Invalid value NaN for volume\n"); 257 return AVERROR(EINVAL); 258 } else { 259 av_log(ctx, AV_LOG_WARNING, "Invalid value NaN for volume, setting to 0\n"); 260 vol->volume = 0; 261 } 262 } 263 vol->var_values[VAR_VOLUME] = vol->volume; 264 265 av_log(ctx, AV_LOG_VERBOSE, "n:%f t:%f pts:%f precision:%s ", 266 vol->var_values[VAR_N], vol->var_values[VAR_T], vol->var_values[VAR_PTS], 267 precision_str[vol->precision]); 268 269 if (vol->precision == PRECISION_FIXED) { 270 vol->volume_i = (int)(vol->volume * 256 + 0.5); 271 vol->volume = vol->volume_i / 256.0; 272 av_log(ctx, AV_LOG_VERBOSE, "volume_i:%d/255 ", vol->volume_i); 273 } 274 av_log(ctx, AV_LOG_VERBOSE, "volume:%f volume_dB:%f\n", 275 vol->volume, 20.0*log(vol->volume)/M_LN10); 276 277 volume_init(vol); 278 return 0; 279} 280 281static int config_output(AVFilterLink *outlink) 282{ 283 AVFilterContext *ctx = outlink->src; 284 VolumeContext *vol = ctx->priv; 285 AVFilterLink *inlink = ctx->inputs[0]; 286 287 vol->sample_fmt = inlink->format; 288 vol->channels = inlink->channels; 289 vol->planes = av_sample_fmt_is_planar(inlink->format) ? vol->channels : 1; 290 291 vol->var_values[VAR_N] = 292 vol->var_values[VAR_NB_CONSUMED_SAMPLES] = 293 vol->var_values[VAR_NB_SAMPLES] = 294 vol->var_values[VAR_POS] = 295 vol->var_values[VAR_PTS] = 296 vol->var_values[VAR_STARTPTS] = 297 vol->var_values[VAR_STARTT] = 298 vol->var_values[VAR_T] = 299 vol->var_values[VAR_VOLUME] = NAN; 300 301 vol->var_values[VAR_NB_CHANNELS] = inlink->channels; 302 vol->var_values[VAR_TB] = av_q2d(inlink->time_base); 303 vol->var_values[VAR_SAMPLE_RATE] = inlink->sample_rate; 304 305 av_log(inlink->src, AV_LOG_VERBOSE, "tb:%f sample_rate:%f nb_channels:%f\n", 306 vol->var_values[VAR_TB], 307 vol->var_values[VAR_SAMPLE_RATE], 308 vol->var_values[VAR_NB_CHANNELS]); 309 310 return set_volume(ctx); 311} 312 313static int process_command(AVFilterContext *ctx, const char *cmd, const char *args, 314 char *res, int res_len, int flags) 315{ 316 VolumeContext *vol = ctx->priv; 317 int ret = AVERROR(ENOSYS); 318 319 if (!strcmp(cmd, "volume")) { 320 if ((ret = set_expr(&vol->volume_pexpr, args, ctx)) < 0) 321 return ret; 322 if (vol->eval_mode == EVAL_MODE_ONCE) 323 set_volume(ctx); 324 } 325 326 return ret; 327} 328 329#define D2TS(d) (isnan(d) ? AV_NOPTS_VALUE : (int64_t)(d)) 330#define TS2D(ts) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts)) 331#define TS2T(ts, tb) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts)*av_q2d(tb)) 332 333static int filter_frame(AVFilterLink *inlink, AVFrame *buf) 334{ 335 AVFilterContext *ctx = inlink->dst; 336 VolumeContext *vol = inlink->dst->priv; 337 AVFilterLink *outlink = inlink->dst->outputs[0]; 338 int nb_samples = buf->nb_samples; 339 AVFrame *out_buf; 340 int64_t pos; 341 AVFrameSideData *sd = av_frame_get_side_data(buf, AV_FRAME_DATA_REPLAYGAIN); 342 int ret; 343 344 if (sd && vol->replaygain != REPLAYGAIN_IGNORE) { 345 if (vol->replaygain != REPLAYGAIN_DROP) { 346 AVReplayGain *replaygain = (AVReplayGain*)sd->data; 347 int32_t gain = 100000; 348 uint32_t peak = 100000; 349 float g, p; 350 351 if (vol->replaygain == REPLAYGAIN_TRACK && 352 replaygain->track_gain != INT32_MIN) { 353 gain = replaygain->track_gain; 354 355 if (replaygain->track_peak != 0) 356 peak = replaygain->track_peak; 357 } else if (replaygain->album_gain != INT32_MIN) { 358 gain = replaygain->album_gain; 359 360 if (replaygain->album_peak != 0) 361 peak = replaygain->album_peak; 362 } else { 363 av_log(inlink->dst, AV_LOG_WARNING, "Both ReplayGain gain " 364 "values are unknown.\n"); 365 } 366 g = gain / 100000.0f; 367 p = peak / 100000.0f; 368 369 av_log(inlink->dst, AV_LOG_VERBOSE, 370 "Using gain %f dB from replaygain side data.\n", g); 371 372 vol->volume = pow(10, (g + vol->replaygain_preamp) / 20); 373 if (vol->replaygain_noclip) 374 vol->volume = FFMIN(vol->volume, 1.0 / p); 375 vol->volume_i = (int)(vol->volume * 256 + 0.5); 376 377 volume_init(vol); 378 } 379 av_frame_remove_side_data(buf, AV_FRAME_DATA_REPLAYGAIN); 380 } 381 382 if (isnan(vol->var_values[VAR_STARTPTS])) { 383 vol->var_values[VAR_STARTPTS] = TS2D(buf->pts); 384 vol->var_values[VAR_STARTT ] = TS2T(buf->pts, inlink->time_base); 385 } 386 vol->var_values[VAR_PTS] = TS2D(buf->pts); 387 vol->var_values[VAR_T ] = TS2T(buf->pts, inlink->time_base); 388 vol->var_values[VAR_N ] = inlink->frame_count; 389 390 pos = av_frame_get_pkt_pos(buf); 391 vol->var_values[VAR_POS] = pos == -1 ? NAN : pos; 392 if (vol->eval_mode == EVAL_MODE_FRAME) 393 set_volume(ctx); 394 395 if (vol->volume == 1.0 || vol->volume_i == 256) { 396 out_buf = buf; 397 goto end; 398 } 399 400 /* do volume scaling in-place if input buffer is writable */ 401 if (av_frame_is_writable(buf)) { 402 out_buf = buf; 403 } else { 404 out_buf = ff_get_audio_buffer(inlink, nb_samples); 405 if (!out_buf) 406 return AVERROR(ENOMEM); 407 ret = av_frame_copy_props(out_buf, buf); 408 if (ret < 0) { 409 av_frame_free(&out_buf); 410 av_frame_free(&buf); 411 return ret; 412 } 413 } 414 415 if (vol->precision != PRECISION_FIXED || vol->volume_i > 0) { 416 int p, plane_samples; 417 418 if (av_sample_fmt_is_planar(buf->format)) 419 plane_samples = FFALIGN(nb_samples, vol->samples_align); 420 else 421 plane_samples = FFALIGN(nb_samples * vol->channels, vol->samples_align); 422 423 if (vol->precision == PRECISION_FIXED) { 424 for (p = 0; p < vol->planes; p++) { 425 vol->scale_samples(out_buf->extended_data[p], 426 buf->extended_data[p], plane_samples, 427 vol->volume_i); 428 } 429 } else if (av_get_packed_sample_fmt(vol->sample_fmt) == AV_SAMPLE_FMT_FLT) { 430 for (p = 0; p < vol->planes; p++) { 431 vol->fdsp.vector_fmul_scalar((float *)out_buf->extended_data[p], 432 (const float *)buf->extended_data[p], 433 vol->volume, plane_samples); 434 } 435 } else { 436 for (p = 0; p < vol->planes; p++) { 437 vol->fdsp.vector_dmul_scalar((double *)out_buf->extended_data[p], 438 (const double *)buf->extended_data[p], 439 vol->volume, plane_samples); 440 } 441 } 442 } 443 444 emms_c(); 445 446 if (buf != out_buf) 447 av_frame_free(&buf); 448 449end: 450 vol->var_values[VAR_NB_CONSUMED_SAMPLES] += out_buf->nb_samples; 451 return ff_filter_frame(outlink, out_buf); 452} 453 454static const AVFilterPad avfilter_af_volume_inputs[] = { 455 { 456 .name = "default", 457 .type = AVMEDIA_TYPE_AUDIO, 458 .filter_frame = filter_frame, 459 }, 460 { NULL } 461}; 462 463static const AVFilterPad avfilter_af_volume_outputs[] = { 464 { 465 .name = "default", 466 .type = AVMEDIA_TYPE_AUDIO, 467 .config_props = config_output, 468 }, 469 { NULL } 470}; 471 472AVFilter ff_af_volume = { 473 .name = "volume", 474 .description = NULL_IF_CONFIG_SMALL("Change input volume."), 475 .query_formats = query_formats, 476 .priv_size = sizeof(VolumeContext), 477 .priv_class = &volume_class, 478 .init = init, 479 .uninit = uninit, 480 .inputs = avfilter_af_volume_inputs, 481 .outputs = avfilter_af_volume_outputs, 482 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC, 483 .process_command = process_command, 484}; 485