libavfilter/af_volume.c

   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/ffmath.h"
  31 #include "libavutil/float_dsp.h"
  32 #include "libavutil/mem.h"
  33 #include "libavutil/opt.h"
  34 #include "libavutil/replaygain.h"
  35
  36 #include "audio.h"
  37 #include "avfilter.h"
  38 #include "filters.h"
  39 #include "formats.h"
  40 #include "af_volume.h"
  41
  42 static const char * const precision_str[] = {
  43     "fixed", "float", "double"
  44 };
  45
  46 static const char *const var_names[] = {
  47     "n",                   ///< frame number (starting at zero)
  48     "nb_channels",         ///< number of channels
  49     "nb_consumed_samples", ///< number of samples consumed by the filter
  50     "nb_samples",          ///< number of samples in the current frame
  51 #if FF_API_FRAME_PKT
  52     "pos",                 ///< position in the file of the frame
  53 #endif
  54     "pts",                 ///< frame presentation timestamp
  55     "sample_rate",         ///< sample rate
  56     "startpts",            ///< PTS at start of stream
  57     "startt",              ///< time at start of stream
  58     "t",                   ///< time in the file of the frame
  59     "tb",                  ///< timebase
  60     "volume",              ///< last set value
  61     NULL
  62 };
  63
  64 #define OFFSET(x) offsetof(VolumeContext, x)
  65 #define A AV_OPT_FLAG_AUDIO_PARAM
  66 #define F AV_OPT_FLAG_FILTERING_PARAM
  67 #define T AV_OPT_FLAG_RUNTIME_PARAM
  68
  69 static const AVOption volume_options[] = {
  70     { "volume", "set volume adjustment expression",
  71             OFFSET(volume_expr), AV_OPT_TYPE_STRING, { .str = "1.0" }, .flags = A|F|T },
  72     { "precision", "select mathematical precision",
  73             OFFSET(precision), AV_OPT_TYPE_INT, { .i64 = PRECISION_FLOAT }, PRECISION_FIXED, PRECISION_DOUBLE, A|F, .unit = "precision" },
  74         { "fixed",  "select 8-bit fixed-point",     0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_FIXED  }, INT_MIN, INT_MAX, A|F, .unit = "precision" },
  75         { "float",  "select 32-bit floating-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_FLOAT  }, INT_MIN, INT_MAX, A|F, .unit = "precision" },
  76         { "double", "select 64-bit floating-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_DOUBLE }, INT_MIN, INT_MAX, A|F, .unit = "precision" },
  77     { "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, .unit = "eval" },
  78          { "once",  "eval volume expression once", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_ONCE},  .flags = A|F, .unit = "eval" },
  79          { "frame", "eval volume expression per-frame",                  0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_FRAME}, .flags = A|F, .unit = "eval" },
  80     { "replaygain", "Apply replaygain side data when present",
  81             OFFSET(replaygain), AV_OPT_TYPE_INT, { .i64 = REPLAYGAIN_DROP }, REPLAYGAIN_DROP, REPLAYGAIN_ALBUM, A|F, .unit = "replaygain" },
  82         { "drop",   "replaygain side data is dropped", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_DROP   }, 0, 0, A|F, .unit = "replaygain" },
  83         { "ignore", "replaygain side data is ignored", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_IGNORE }, 0, 0, A|F, .unit = "replaygain" },
  84         { "track",  "track gain is preferred",         0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_TRACK  }, 0, 0, A|F, .unit = "replaygain" },
  85         { "album",  "album gain is preferred",         0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_ALBUM  }, 0, 0, A|F, .unit = "replaygain" },
  86     { "replaygain_preamp", "Apply replaygain pre-amplification",
  87             OFFSET(replaygain_preamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.0 }, -15.0, 15.0, A|F },
  88     { "replaygain_noclip", "Apply replaygain clipping prevention",
  89             OFFSET(replaygain_noclip), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, A|F },
  90     { NULL }
  91 };
  92
  93 AVFILTER_DEFINE_CLASS(volume);
  94
  95 static int set_expr(AVExpr **pexpr, const char *expr, void *log_ctx)
  96 {
  97     int ret;
  98     AVExpr *old = NULL;
  99
 100     if (*pexpr)
 101         old = *pexpr;
 102     ret = av_expr_parse(pexpr, expr, var_names,
 103                         NULL, NULL, NULL, NULL, 0, log_ctx);
 104     if (ret < 0) {
 105         av_log(log_ctx, AV_LOG_ERROR,
 106                "Error when evaluating the volume expression '%s'\n", expr);
 107         *pexpr = old;
 108         return ret;
 109     }
 110
 111     av_expr_free(old);
 112     return 0;
 113 }
 114
 115 static av_cold int init(AVFilterContext *ctx)
 116 {
 117     VolumeContext *vol = ctx->priv;
 118
 119     vol->fdsp = avpriv_float_dsp_alloc(0);
 120     if (!vol->fdsp)
 121         return AVERROR(ENOMEM);
 122
 123     return set_expr(&vol->volume_pexpr, vol->volume_expr, ctx);
 124 }
 125
 126 static av_cold void uninit(AVFilterContext *ctx)
 127 {
 128     VolumeContext *vol = ctx->priv;
 129     av_expr_free(vol->volume_pexpr);
 130     av_freep(&vol->fdsp);
 131 }
 132
 133 static int query_formats(const AVFilterContext *ctx,
 134                          AVFilterFormatsConfig **cfg_in,
 135                          AVFilterFormatsConfig **cfg_out)
 136 {
 137     const VolumeContext *vol = ctx->priv;
 138     static const enum AVSampleFormat sample_fmts[][7] = {
 139         [PRECISION_FIXED] = {
 140             AV_SAMPLE_FMT_U8,
 141             AV_SAMPLE_FMT_U8P,
 142             AV_SAMPLE_FMT_S16,
 143             AV_SAMPLE_FMT_S16P,
 144             AV_SAMPLE_FMT_S32,
 145             AV_SAMPLE_FMT_S32P,
 146             AV_SAMPLE_FMT_NONE
 147         },
 148         [PRECISION_FLOAT] = {
 149             AV_SAMPLE_FMT_FLT,
 150             AV_SAMPLE_FMT_FLTP,
 151             AV_SAMPLE_FMT_NONE
 152         },
 153         [PRECISION_DOUBLE] = {
 154             AV_SAMPLE_FMT_DBL,
 155             AV_SAMPLE_FMT_DBLP,
 156             AV_SAMPLE_FMT_NONE
 157         }
 158     };
 159     int ret;
 160
 161     ret = ff_set_common_formats_from_list2(ctx, cfg_in, cfg_out, sample_fmts[vol->precision]);
 162     if (ret < 0)
 163         return ret;
 164
 165     return 0;
 166 }
 167
 168 static inline void scale_samples_u8(uint8_t *dst, const uint8_t *src,
 169                                     int nb_samples, int volume)
 170 {
 171     int i;
 172     for (i = 0; i < nb_samples; i++)
 173         dst[i] = av_clip_uint8(((((int64_t)src[i] - 128) * volume + 128) >> 8) + 128);
 174 }
 175
 176 static inline void scale_samples_u8_small(uint8_t *dst, const uint8_t *src,
 177                                           int nb_samples, int volume)
 178 {
 179     int i;
 180     for (i = 0; i < nb_samples; i++)
 181         dst[i] = av_clip_uint8((((src[i] - 128) * volume + 128) >> 8) + 128);
 182 }
 183
 184 static inline void scale_samples_s16(uint8_t *dst, const uint8_t *src,
 185                                      int nb_samples, int volume)
 186 {
 187     int i;
 188     int16_t *smp_dst       = (int16_t *)dst;
 189     const int16_t *smp_src = (const int16_t *)src;
 190     for (i = 0; i < nb_samples; i++)
 191         smp_dst[i] = av_clip_int16(((int64_t)smp_src[i] * volume + 128) >> 8);
 192 }
 193
 194 static inline void scale_samples_s16_small(uint8_t *dst, const uint8_t *src,
 195                                            int nb_samples, int volume)
 196 {
 197     int i;
 198     int16_t *smp_dst       = (int16_t *)dst;
 199     const int16_t *smp_src = (const int16_t *)src;
 200     for (i = 0; i < nb_samples; i++)
 201         smp_dst[i] = av_clip_int16((smp_src[i] * volume + 128) >> 8);
 202 }
 203
 204 static inline void scale_samples_s32(uint8_t *dst, const uint8_t *src,
 205                                      int nb_samples, int volume)
 206 {
 207     int i;
 208     int32_t *smp_dst       = (int32_t *)dst;
 209     const int32_t *smp_src = (const int32_t *)src;
 210     for (i = 0; i < nb_samples; i++)
 211         smp_dst[i] = av_clipl_int32((((int64_t)smp_src[i] * volume + 128) >> 8));
 212 }
 213
 214 static av_cold void volume_init(VolumeContext *vol)
 215 {
 216     vol->samples_align = 1;
 217
 218     switch (av_get_packed_sample_fmt(vol->sample_fmt)) {
 219     case AV_SAMPLE_FMT_U8:
 220         if (vol->volume_i < 0x1000000)
 221             vol->scale_samples = scale_samples_u8_small;
 222         else
 223             vol->scale_samples = scale_samples_u8;
 224         break;
 225     case AV_SAMPLE_FMT_S16:
 226         if (vol->volume_i < 0x10000)
 227             vol->scale_samples = scale_samples_s16_small;
 228         else
 229             vol->scale_samples = scale_samples_s16;
 230         break;
 231     case AV_SAMPLE_FMT_S32:
 232         vol->scale_samples = scale_samples_s32;
 233         break;
 234     case AV_SAMPLE_FMT_FLT:
 235         vol->samples_align = 4;
 236         break;
 237     case AV_SAMPLE_FMT_DBL:
 238         vol->samples_align = 8;
 239         break;
 240     }
 241
 242 #if ARCH_X86
 243     ff_volume_init_x86(vol);
 244 #endif
 245 }
 246
 247 static int set_volume(AVFilterContext *ctx)
 248 {
 249     VolumeContext *vol = ctx->priv;
 250
 251     vol->volume = av_expr_eval(vol->volume_pexpr, vol->var_values, NULL);
 252     if (isnan(vol->volume)) {
 253         if (vol->eval_mode == EVAL_MODE_ONCE) {
 254             av_log(ctx, AV_LOG_ERROR, "Invalid value NaN for volume\n");
 255             return AVERROR(EINVAL);
 256         } else {
 257             av_log(ctx, AV_LOG_WARNING, "Invalid value NaN for volume, setting to 0\n");
 258             vol->volume = 0;
 259         }
 260     }
 261     vol->var_values[VAR_VOLUME] = vol->volume;
 262
 263     av_log(ctx, AV_LOG_VERBOSE, "n:%f t:%f pts:%f precision:%s ",
 264            vol->var_values[VAR_N], vol->var_values[VAR_T], vol->var_values[VAR_PTS],
 265            precision_str[vol->precision]);
 266
 267     if (vol->precision == PRECISION_FIXED) {
 268         vol->volume_i = (int)(vol->volume * 256 + 0.5);
 269         vol->volume   = vol->volume_i / 256.0;
 270         av_log(ctx, AV_LOG_VERBOSE, "volume_i:%d/255 ", vol->volume_i);
 271     }
 272     av_log(ctx, AV_LOG_VERBOSE, "volume:%f volume_dB:%f\n",
 273            vol->volume, 20.0*log10(vol->volume));
 274
 275     volume_init(vol);
 276     return 0;
 277 }
 278
 279 static int config_output(AVFilterLink *outlink)
 280 {
 281     AVFilterContext *ctx = outlink->src;
 282     VolumeContext *vol   = ctx->priv;
 283     AVFilterLink *inlink = ctx->inputs[0];
 284
 285     vol->sample_fmt = inlink->format;
 286     vol->channels   = inlink->ch_layout.nb_channels;
 287     vol->planes     = av_sample_fmt_is_planar(inlink->format) ? vol->channels : 1;
 288
 289     vol->var_values[VAR_N] =
 290     vol->var_values[VAR_NB_CONSUMED_SAMPLES] =
 291     vol->var_values[VAR_NB_SAMPLES] =
 292 #if FF_API_FRAME_PKT
 293     vol->var_values[VAR_POS] =
 294 #endif
 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->ch_layout.nb_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
 313 static 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 static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
 330 {
 331     FilterLink      *inl = ff_filter_link(inlink);
 332     AVFilterContext *ctx = inlink->dst;
 333     VolumeContext *vol    = inlink->dst->priv;
 334     AVFilterLink *outlink = inlink->dst->outputs[0];
 335     int nb_samples        = buf->nb_samples;
 336     AVFrame *out_buf;
 337     AVFrameSideData *sd = av_frame_get_side_data(buf, AV_FRAME_DATA_REPLAYGAIN);
 338     int ret;
 339
 340     if (sd && vol->replaygain != REPLAYGAIN_IGNORE) {
 341         if (vol->replaygain != REPLAYGAIN_DROP) {
 342             AVReplayGain *replaygain = (AVReplayGain*)sd->data;
 343             int32_t gain  = 100000;
 344             uint32_t peak = 100000;
 345             float g, p;
 346
 347             if (vol->replaygain == REPLAYGAIN_TRACK &&
 348                 replaygain->track_gain != INT32_MIN) {
 349                 gain = replaygain->track_gain;
 350
 351                 if (replaygain->track_peak != 0)
 352                     peak = replaygain->track_peak;
 353             } else if (replaygain->album_gain != INT32_MIN) {
 354                 gain = replaygain->album_gain;
 355
 356                 if (replaygain->album_peak != 0)
 357                     peak = replaygain->album_peak;
 358             } else {
 359                 av_log(inlink->dst, AV_LOG_WARNING, "Both ReplayGain gain "
 360                        "values are unknown.\n");
 361             }
 362             g = gain / 100000.0f;
 363             p = peak / 100000.0f;
 364
 365             av_log(inlink->dst, AV_LOG_VERBOSE,
 366                    "Using gain %f dB from replaygain side data.\n", g);
 367
 368             vol->volume   = ff_exp10((g + vol->replaygain_preamp) / 20);
 369             if (vol->replaygain_noclip)
 370                 vol->volume = FFMIN(vol->volume, 1.0 / p);
 371             vol->volume_i = (int)(vol->volume * 256 + 0.5);
 372
 373             volume_init(vol);
 374         }
 375         av_frame_remove_side_data(buf, AV_FRAME_DATA_REPLAYGAIN);
 376     }
 377
 378     if (isnan(vol->var_values[VAR_STARTPTS])) {
 379         vol->var_values[VAR_STARTPTS] = TS2D(buf->pts);
 380         vol->var_values[VAR_STARTT  ] = TS2T(buf->pts, inlink->time_base);
 381     }
 382     vol->var_values[VAR_PTS] = TS2D(buf->pts);
 383     vol->var_values[VAR_T  ] = TS2T(buf->pts, inlink->time_base);
 384     vol->var_values[VAR_N  ] = inl->frame_count_out;
 385
 386 #if FF_API_FRAME_PKT
 387 FF_DISABLE_DEPRECATION_WARNINGS
 388     {
 389         int64_t pos;
 390         pos = buf->pkt_pos;
 391         vol->var_values[VAR_POS] = pos == -1 ? NAN : pos;
 392     }
 393 FF_ENABLE_DEPRECATION_WARNINGS
 394 #endif
 395     if (vol->eval_mode == EVAL_MODE_FRAME)
 396         set_volume(ctx);
 397
 398     if (vol->volume == 1.0 || vol->volume_i == 256) {
 399         out_buf = buf;
 400         goto end;
 401     }
 402
 403     /* do volume scaling in-place if input buffer is writable */
 404     if (av_frame_is_writable(buf)
 405             && (vol->precision != PRECISION_FIXED || vol->volume_i > 0)) {
 406         out_buf = buf;
 407     } else {
 408         out_buf = ff_get_audio_buffer(outlink, nb_samples);
 409         if (!out_buf) {
 410             av_frame_free(&buf);
 411             return AVERROR(ENOMEM);
 412         }
 413         ret = av_frame_copy_props(out_buf, buf);
 414         if (ret < 0) {
 415             av_frame_free(&out_buf);
 416             av_frame_free(&buf);
 417             return ret;
 418         }
 419     }
 420
 421     if (vol->precision != PRECISION_FIXED || vol->volume_i > 0) {
 422         int p, plane_samples;
 423
 424         if (av_sample_fmt_is_planar(buf->format))
 425             plane_samples = FFALIGN(nb_samples, vol->samples_align);
 426         else
 427             plane_samples = FFALIGN(nb_samples * vol->channels, vol->samples_align);
 428
 429         if (vol->precision == PRECISION_FIXED) {
 430             for (p = 0; p < vol->planes; p++) {
 431                 vol->scale_samples(out_buf->extended_data[p],
 432                                    buf->extended_data[p], plane_samples,
 433                                    vol->volume_i);
 434             }
 435         } else if (av_get_packed_sample_fmt(vol->sample_fmt) == AV_SAMPLE_FMT_FLT) {
 436             for (p = 0; p < vol->planes; p++) {
 437                 vol->fdsp->vector_fmul_scalar((float *)out_buf->extended_data[p],
 438                                              (const float *)buf->extended_data[p],
 439                                              vol->volume, plane_samples);
 440             }
 441         } else {
 442             for (p = 0; p < vol->planes; p++) {
 443                 vol->fdsp->vector_dmul_scalar((double *)out_buf->extended_data[p],
 444                                              (const double *)buf->extended_data[p],
 445                                              vol->volume, plane_samples);
 446             }
 447         }
 448     }
 449
 450     if (buf != out_buf)
 451         av_frame_free(&buf);
 452
 453 end:
 454     vol->var_values[VAR_NB_CONSUMED_SAMPLES] += out_buf->nb_samples;
 455     return ff_filter_frame(outlink, out_buf);
 456 }
 457
 458 static const AVFilterPad avfilter_af_volume_inputs[] = {
 459     {
 460         .name           = "default",
 461         .type           = AVMEDIA_TYPE_AUDIO,
 462         .filter_frame   = filter_frame,
 463     },
 464 };
 465
 466 static const AVFilterPad avfilter_af_volume_outputs[] = {
 467     {
 468         .name         = "default",
 469         .type         = AVMEDIA_TYPE_AUDIO,
 470         .config_props = config_output,
 471     },
 472 };
 473
 474 const FFFilter ff_af_volume = {
 475     .p.name         = "volume",
 476     .p.description  = NULL_IF_CONFIG_SMALL("Change input volume."),
 477     .p.priv_class   = &volume_class,
 478     .p.flags        = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
 479     .priv_size      = sizeof(VolumeContext),
 480     .init           = init,
 481     .uninit         = uninit,
 482     FILTER_INPUTS(avfilter_af_volume_inputs),
 483     FILTER_OUTPUTS(avfilter_af_volume_outputs),
 484     FILTER_QUERY_FUNC2(query_formats),
 485     .process_command = process_command,
 486 };