libavfilter/af_afade.c

   1 /*
   2  * Copyright (c) 2013-2015 Paul B Mahol
   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 /**
  22  * @file
  23  * fade audio filter
  24  */
  25
  26 #include "config_components.h"
  27
  28 #include "libavutil/opt.h"
  29 #include "audio.h"
  30 #include "avfilter.h"
  31 #include "filters.h"
  32 #include "internal.h"
  33
  34 typedef struct AudioFadeContext {
  35     const AVClass *class;
  36     int type;
  37     int curve, curve2;
  38     int64_t nb_samples;
  39     int64_t start_sample;
  40     int64_t duration;
  41     int64_t start_time;
  42     double silence;
  43     double unity;
  44     int overlap;
  45     int status[2];
  46     int passthrough;
  47     int64_t pts;
  48
  49     void (*fade_samples)(uint8_t **dst, uint8_t * const *src,
  50                          int nb_samples, int channels, int direction,
  51                          int64_t start, int64_t range, int curve,
  52                          double silence, double unity);
  53     void (*scale_samples)(uint8_t **dst, uint8_t * const *src,
  54                           int nb_samples, int channels, double unity);
  55     void (*crossfade_samples)(uint8_t **dst, uint8_t * const *cf0,
  56                               uint8_t * const *cf1,
  57                               int nb_samples, int channels,
  58                               int curve0, int curve1);
  59 } AudioFadeContext;
  60
  61 enum CurveType { NONE = -1, TRI, QSIN, ESIN, HSIN, LOG, IPAR, QUA, CUB, SQU, CBR, PAR, EXP, IQSIN, IHSIN, DESE, DESI, LOSI, SINC, ISINC, QUAT, QUATR, QSIN2, HSIN2, NB_CURVES };
  62
  63 #define OFFSET(x) offsetof(AudioFadeContext, x)
  64 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
  65 #define TFLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
  66
  67     static const enum AVSampleFormat sample_fmts[] = {
  68         AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P,
  69         AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_S32P,
  70         AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
  71         AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_DBLP,
  72         AV_SAMPLE_FMT_NONE
  73     };
  74
  75 static double fade_gain(int curve, int64_t index, int64_t range, double silence, double unity)
  76 {
  77 #define CUBE(a) ((a)*(a)*(a))
  78     double gain;
  79
  80     gain = av_clipd(1.0 * index / range, 0, 1.0);
  81
  82     switch (curve) {
  83     case QSIN:
  84         gain = sin(gain * M_PI / 2.0);
  85         break;
  86     case IQSIN:
  87         /* 0.6... = 2 / M_PI */
  88         gain = 0.6366197723675814 * asin(gain);
  89         break;
  90     case ESIN:
  91         gain = 1.0 - cos(M_PI / 4.0 * (CUBE(2.0*gain - 1) + 1));
  92         break;
  93     case HSIN:
  94         gain = (1.0 - cos(gain * M_PI)) / 2.0;
  95         break;
  96     case IHSIN:
  97         /* 0.3... = 1 / M_PI */
  98         gain = 0.3183098861837907 * acos(1 - 2 * gain);
  99         break;
 100     case EXP:
 101         /* -11.5... = 5*ln(0.1) */
 102         gain = exp(-11.512925464970227 * (1 - gain));
 103         break;
 104     case LOG:
 105         gain = av_clipd(1 + 0.2 * log10(gain), 0, 1.0);
 106         break;
 107     case PAR:
 108         gain = 1 - sqrt(1 - gain);
 109         break;
 110     case IPAR:
 111         gain = (1 - (1 - gain) * (1 - gain));
 112         break;
 113     case QUA:
 114         gain *= gain;
 115         break;
 116     case CUB:
 117         gain = CUBE(gain);
 118         break;
 119     case SQU:
 120         gain = sqrt(gain);
 121         break;
 122     case CBR:
 123         gain = cbrt(gain);
 124         break;
 125     case DESE:
 126         gain = gain <= 0.5 ? cbrt(2 * gain) / 2: 1 - cbrt(2 * (1 - gain)) / 2;
 127         break;
 128     case DESI:
 129         gain = gain <= 0.5 ? CUBE(2 * gain) / 2: 1 - CUBE(2 * (1 - gain)) / 2;
 130         break;
 131     case LOSI: {
 132                    const double a = 1. / (1. - 0.787) - 1;
 133                    double A = 1. / (1.0 + exp(0 -((gain-0.5) * a * 2.0)));
 134                    double B = 1. / (1.0 + exp(a));
 135                    double C = 1. / (1.0 + exp(0-a));
 136                    gain = (A - B) / (C - B);
 137                }
 138         break;
 139     case SINC:
 140         gain = gain >= 1.0 ? 1.0 : sin(M_PI * (1.0 - gain)) / (M_PI * (1.0 - gain));
 141         break;
 142     case ISINC:
 143         gain = gain <= 0.0 ? 0.0 : 1.0 - sin(M_PI * gain) / (M_PI * gain);
 144         break;
 145     case QUAT:
 146         gain = gain * gain * gain * gain;
 147         break;
 148     case QUATR:
 149         gain = pow(gain, 0.25);
 150         break;
 151     case QSIN2:
 152         gain = sin(gain * M_PI / 2.0) * sin(gain * M_PI / 2.0);
 153         break;
 154     case HSIN2:
 155         gain = pow((1.0 - cos(gain * M_PI)) / 2.0, 2.0);
 156         break;
 157     case NONE:
 158         gain = 1.0;
 159         break;
 160     }
 161
 162     return silence + (unity - silence) * gain;
 163 }
 164
 165 #define FADE_PLANAR(name, type)                                             \
 166 static void fade_samples_## name ##p(uint8_t **dst, uint8_t * const *src,   \
 167                                      int nb_samples, int channels, int dir, \
 168                                      int64_t start, int64_t range,int curve,\
 169                                      double silence, double unity)          \
 170 {                                                                           \
 171     int i, c;                                                               \
 172                                                                             \
 173     for (i = 0; i < nb_samples; i++) {                                      \
 174         double gain = fade_gain(curve, start + i * dir,range,silence,unity);\
 175         for (c = 0; c < channels; c++) {                                    \
 176             type *d = (type *)dst[c];                                       \
 177             const type *s = (type *)src[c];                                 \
 178                                                                             \
 179             d[i] = s[i] * gain;                                             \
 180         }                                                                   \
 181     }                                                                       \
 182 }
 183
 184 #define FADE(name, type)                                                    \
 185 static void fade_samples_## name (uint8_t **dst, uint8_t * const *src,      \
 186                                   int nb_samples, int channels, int dir,    \
 187                                   int64_t start, int64_t range, int curve,  \
 188                                   double silence, double unity)             \
 189 {                                                                           \
 190     type *d = (type *)dst[0];                                               \
 191     const type *s = (type *)src[0];                                         \
 192     int i, c, k = 0;                                                        \
 193                                                                             \
 194     for (i = 0; i < nb_samples; i++) {                                      \
 195         double gain = fade_gain(curve, start + i * dir,range,silence,unity);\
 196         for (c = 0; c < channels; c++, k++)                                 \
 197             d[k] = s[k] * gain;                                             \
 198     }                                                                       \
 199 }
 200
 201 FADE_PLANAR(dbl, double)
 202 FADE_PLANAR(flt, float)
 203 FADE_PLANAR(s16, int16_t)
 204 FADE_PLANAR(s32, int32_t)
 205
 206 FADE(dbl, double)
 207 FADE(flt, float)
 208 FADE(s16, int16_t)
 209 FADE(s32, int32_t)
 210
 211 #define SCALE_PLANAR(name, type)                                            \
 212 static void scale_samples_## name ##p(uint8_t **dst, uint8_t * const *src,  \
 213                                      int nb_samples, int channels,          \
 214                                      double gain)                           \
 215 {                                                                           \
 216     int i, c;                                                               \
 217                                                                             \
 218     for (i = 0; i < nb_samples; i++) {                                      \
 219         for (c = 0; c < channels; c++) {                                    \
 220             type *d = (type *)dst[c];                                       \
 221             const type *s = (type *)src[c];                                 \
 222                                                                             \
 223             d[i] = s[i] * gain;                                             \
 224         }                                                                   \
 225     }                                                                       \
 226 }
 227
 228 #define SCALE(name, type)                                                   \
 229 static void scale_samples_## name (uint8_t **dst, uint8_t * const *src,     \
 230                                   int nb_samples, int channels, double gain)\
 231 {                                                                           \
 232     type *d = (type *)dst[0];                                               \
 233     const type *s = (type *)src[0];                                         \
 234     int i, c, k = 0;                                                        \
 235                                                                             \
 236     for (i = 0; i < nb_samples; i++) {                                      \
 237         for (c = 0; c < channels; c++, k++)                                 \
 238             d[k] = s[k] * gain;                                             \
 239     }                                                                       \
 240 }
 241
 242 SCALE_PLANAR(dbl, double)
 243 SCALE_PLANAR(flt, float)
 244 SCALE_PLANAR(s16, int16_t)
 245 SCALE_PLANAR(s32, int32_t)
 246
 247 SCALE(dbl, double)
 248 SCALE(flt, float)
 249 SCALE(s16, int16_t)
 250 SCALE(s32, int32_t)
 251
 252 static int config_output(AVFilterLink *outlink)
 253 {
 254     AVFilterContext *ctx = outlink->src;
 255     AudioFadeContext *s  = ctx->priv;
 256
 257     switch (outlink->format) {
 258     case AV_SAMPLE_FMT_DBL:  s->fade_samples = fade_samples_dbl;
 259                              s->scale_samples = scale_samples_dbl;
 260                              break;
 261     case AV_SAMPLE_FMT_DBLP: s->fade_samples = fade_samples_dblp;
 262                              s->scale_samples = scale_samples_dblp;
 263                              break;
 264     case AV_SAMPLE_FMT_FLT:  s->fade_samples = fade_samples_flt;
 265                              s->scale_samples = scale_samples_flt;
 266                              break;
 267     case AV_SAMPLE_FMT_FLTP: s->fade_samples = fade_samples_fltp;
 268                              s->scale_samples = scale_samples_fltp;
 269                              break;
 270     case AV_SAMPLE_FMT_S16:  s->fade_samples = fade_samples_s16;
 271                              s->scale_samples = scale_samples_s16;
 272                              break;
 273     case AV_SAMPLE_FMT_S16P: s->fade_samples = fade_samples_s16p;
 274                              s->scale_samples = scale_samples_s16p;
 275                              break;
 276     case AV_SAMPLE_FMT_S32:  s->fade_samples = fade_samples_s32;
 277                              s->scale_samples = scale_samples_s32;
 278                              break;
 279     case AV_SAMPLE_FMT_S32P: s->fade_samples = fade_samples_s32p;
 280                              s->scale_samples = scale_samples_s32p;
 281                              break;
 282     }
 283
 284     if (s->duration)
 285         s->nb_samples = av_rescale(s->duration, outlink->sample_rate, AV_TIME_BASE);
 286     s->duration = 0;
 287     if (s->start_time)
 288         s->start_sample = av_rescale(s->start_time, outlink->sample_rate, AV_TIME_BASE);
 289     s->start_time = 0;
 290
 291     return 0;
 292 }
 293
 294 #if CONFIG_AFADE_FILTER
 295
 296 static const AVOption afade_options[] = {
 297     { "type",         "set the fade direction",                      OFFSET(type),         AV_OPT_TYPE_INT,    {.i64 = 0    }, 0, 1, TFLAGS, "type" },
 298     { "t",            "set the fade direction",                      OFFSET(type),         AV_OPT_TYPE_INT,    {.i64 = 0    }, 0, 1, TFLAGS, "type" },
 299     { "in",           "fade-in",                                     0,                    AV_OPT_TYPE_CONST,  {.i64 = 0    }, 0, 0, TFLAGS, "type" },
 300     { "out",          "fade-out",                                    0,                    AV_OPT_TYPE_CONST,  {.i64 = 1    }, 0, 0, TFLAGS, "type" },
 301     { "start_sample", "set number of first sample to start fading",  OFFSET(start_sample), AV_OPT_TYPE_INT64,  {.i64 = 0    }, 0, INT64_MAX, TFLAGS },
 302     { "ss",           "set number of first sample to start fading",  OFFSET(start_sample), AV_OPT_TYPE_INT64,  {.i64 = 0    }, 0, INT64_MAX, TFLAGS },
 303     { "nb_samples",   "set number of samples for fade duration",     OFFSET(nb_samples),   AV_OPT_TYPE_INT64,  {.i64 = 44100}, 1, INT64_MAX, TFLAGS },
 304     { "ns",           "set number of samples for fade duration",     OFFSET(nb_samples),   AV_OPT_TYPE_INT64,  {.i64 = 44100}, 1, INT64_MAX, TFLAGS },
 305     { "start_time",   "set time to start fading",                    OFFSET(start_time),   AV_OPT_TYPE_DURATION, {.i64 = 0 },  0, INT64_MAX, TFLAGS },
 306     { "st",           "set time to start fading",                    OFFSET(start_time),   AV_OPT_TYPE_DURATION, {.i64 = 0 },  0, INT64_MAX, TFLAGS },
 307     { "duration",     "set fade duration",                           OFFSET(duration),     AV_OPT_TYPE_DURATION, {.i64 = 0 },  0, INT64_MAX, TFLAGS },
 308     { "d",            "set fade duration",                           OFFSET(duration),     AV_OPT_TYPE_DURATION, {.i64 = 0 },  0, INT64_MAX, TFLAGS },
 309     { "curve",        "set fade curve type",                         OFFSET(curve),        AV_OPT_TYPE_INT,    {.i64 = TRI  }, NONE, NB_CURVES - 1, TFLAGS, "curve" },
 310     { "c",            "set fade curve type",                         OFFSET(curve),        AV_OPT_TYPE_INT,    {.i64 = TRI  }, NONE, NB_CURVES - 1, TFLAGS, "curve" },
 311     { "nofade",       "no fade; keep audio as-is",                   0,                    AV_OPT_TYPE_CONST,  {.i64 = NONE }, 0, 0, TFLAGS, "curve" },
 312     { "tri",          "linear slope",                                0,                    AV_OPT_TYPE_CONST,  {.i64 = TRI  }, 0, 0, TFLAGS, "curve" },
 313     { "qsin",         "quarter of sine wave",                        0,                    AV_OPT_TYPE_CONST,  {.i64 = QSIN }, 0, 0, TFLAGS, "curve" },
 314     { "esin",         "exponential sine wave",                       0,                    AV_OPT_TYPE_CONST,  {.i64 = ESIN }, 0, 0, TFLAGS, "curve" },
 315     { "hsin",         "half of sine wave",                           0,                    AV_OPT_TYPE_CONST,  {.i64 = HSIN }, 0, 0, TFLAGS, "curve" },
 316     { "log",          "logarithmic",                                 0,                    AV_OPT_TYPE_CONST,  {.i64 = LOG  }, 0, 0, TFLAGS, "curve" },
 317     { "ipar",         "inverted parabola",                           0,                    AV_OPT_TYPE_CONST,  {.i64 = IPAR }, 0, 0, TFLAGS, "curve" },
 318     { "qua",          "quadratic",                                   0,                    AV_OPT_TYPE_CONST,  {.i64 = QUA  }, 0, 0, TFLAGS, "curve" },
 319     { "cub",          "cubic",                                       0,                    AV_OPT_TYPE_CONST,  {.i64 = CUB  }, 0, 0, TFLAGS, "curve" },
 320     { "squ",          "square root",                                 0,                    AV_OPT_TYPE_CONST,  {.i64 = SQU  }, 0, 0, TFLAGS, "curve" },
 321     { "cbr",          "cubic root",                                  0,                    AV_OPT_TYPE_CONST,  {.i64 = CBR  }, 0, 0, TFLAGS, "curve" },
 322     { "par",          "parabola",                                    0,                    AV_OPT_TYPE_CONST,  {.i64 = PAR  }, 0, 0, TFLAGS, "curve" },
 323     { "exp",          "exponential",                                 0,                    AV_OPT_TYPE_CONST,  {.i64 = EXP  }, 0, 0, TFLAGS, "curve" },
 324     { "iqsin",        "inverted quarter of sine wave",               0,                    AV_OPT_TYPE_CONST,  {.i64 = IQSIN}, 0, 0, TFLAGS, "curve" },
 325     { "ihsin",        "inverted half of sine wave",                  0,                    AV_OPT_TYPE_CONST,  {.i64 = IHSIN}, 0, 0, TFLAGS, "curve" },
 326     { "dese",         "double-exponential seat",                     0,                    AV_OPT_TYPE_CONST,  {.i64 = DESE }, 0, 0, TFLAGS, "curve" },
 327     { "desi",         "double-exponential sigmoid",                  0,                    AV_OPT_TYPE_CONST,  {.i64 = DESI }, 0, 0, TFLAGS, "curve" },
 328     { "losi",         "logistic sigmoid",                            0,                    AV_OPT_TYPE_CONST,  {.i64 = LOSI }, 0, 0, TFLAGS, "curve" },
 329     { "sinc",         "sine cardinal function",                      0,                    AV_OPT_TYPE_CONST,  {.i64 = SINC }, 0, 0, TFLAGS, "curve" },
 330     { "isinc",        "inverted sine cardinal function",             0,                    AV_OPT_TYPE_CONST,  {.i64 = ISINC}, 0, 0, TFLAGS, "curve" },
 331     { "quat",         "quartic",                                     0,                    AV_OPT_TYPE_CONST,  {.i64 = QUAT }, 0, 0, TFLAGS, "curve" },
 332     { "quatr",        "quartic root",                                0,                    AV_OPT_TYPE_CONST,  {.i64 = QUATR}, 0, 0, TFLAGS, "curve" },
 333     { "qsin2",        "squared quarter of sine wave",                0,                    AV_OPT_TYPE_CONST,  {.i64 = QSIN2}, 0, 0, TFLAGS, "curve" },
 334     { "hsin2",        "squared half of sine wave",                   0,                    AV_OPT_TYPE_CONST,  {.i64 = HSIN2}, 0, 0, TFLAGS, "curve" },
 335     { "silence",      "set the silence gain",                        OFFSET(silence),      AV_OPT_TYPE_DOUBLE, {.dbl = 0 },    0, 1, TFLAGS },
 336     { "unity",        "set the unity gain",                          OFFSET(unity),        AV_OPT_TYPE_DOUBLE, {.dbl = 1 },    0, 1, TFLAGS },
 337     { NULL }
 338 };
 339
 340 AVFILTER_DEFINE_CLASS(afade);
 341
 342 static av_cold int init(AVFilterContext *ctx)
 343 {
 344     AudioFadeContext *s = ctx->priv;
 345
 346     if (INT64_MAX - s->nb_samples < s->start_sample)
 347         return AVERROR(EINVAL);
 348
 349     return 0;
 350 }
 351
 352 static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
 353 {
 354     AudioFadeContext *s     = inlink->dst->priv;
 355     AVFilterLink *outlink   = inlink->dst->outputs[0];
 356     int nb_samples          = buf->nb_samples;
 357     AVFrame *out_buf;
 358     int64_t cur_sample = av_rescale_q(buf->pts, inlink->time_base, (AVRational){1, inlink->sample_rate});
 359
 360     if (s->unity == 1.0 &&
 361         ((!s->type && (s->start_sample + s->nb_samples < cur_sample)) ||
 362          ( s->type && (cur_sample + nb_samples < s->start_sample))))
 363         return ff_filter_frame(outlink, buf);
 364
 365     if (av_frame_is_writable(buf)) {
 366         out_buf = buf;
 367     } else {
 368         out_buf = ff_get_audio_buffer(outlink, nb_samples);
 369         if (!out_buf)
 370             return AVERROR(ENOMEM);
 371         av_frame_copy_props(out_buf, buf);
 372     }
 373
 374     if ((!s->type && (cur_sample + nb_samples < s->start_sample)) ||
 375         ( s->type && (s->start_sample + s->nb_samples < cur_sample))) {
 376         if (s->silence == 0.) {
 377             av_samples_set_silence(out_buf->extended_data, 0, nb_samples,
 378                                    out_buf->ch_layout.nb_channels, out_buf->format);
 379         } else {
 380             s->scale_samples(out_buf->extended_data, buf->extended_data,
 381                              nb_samples, buf->ch_layout.nb_channels,
 382                              s->silence);
 383         }
 384     } else if (( s->type && (cur_sample + nb_samples < s->start_sample)) ||
 385                (!s->type && (s->start_sample + s->nb_samples < cur_sample))) {
 386         s->scale_samples(out_buf->extended_data, buf->extended_data,
 387                          nb_samples, buf->ch_layout.nb_channels,
 388                          s->unity);
 389     } else {
 390         int64_t start;
 391
 392         if (!s->type)
 393             start = cur_sample - s->start_sample;
 394         else
 395             start = s->start_sample + s->nb_samples - cur_sample;
 396
 397         s->fade_samples(out_buf->extended_data, buf->extended_data,
 398                         nb_samples, buf->ch_layout.nb_channels,
 399                         s->type ? -1 : 1, start,
 400                         s->nb_samples, s->curve, s->silence, s->unity);
 401     }
 402
 403     if (buf != out_buf)
 404         av_frame_free(&buf);
 405
 406     return ff_filter_frame(outlink, out_buf);
 407 }
 408
 409 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
 410                            char *res, int res_len, int flags)
 411 {
 412     int ret;
 413
 414     ret = ff_filter_process_command(ctx, cmd, args, res, res_len, flags);
 415     if (ret < 0)
 416         return ret;
 417
 418     return config_output(ctx->outputs[0]);
 419 }
 420
 421 static const AVFilterPad avfilter_af_afade_inputs[] = {
 422     {
 423         .name         = "default",
 424         .type         = AVMEDIA_TYPE_AUDIO,
 425         .filter_frame = filter_frame,
 426     },
 427 };
 428
 429 static const AVFilterPad avfilter_af_afade_outputs[] = {
 430     {
 431         .name         = "default",
 432         .type         = AVMEDIA_TYPE_AUDIO,
 433         .config_props = config_output,
 434     },
 435 };
 436
 437 const AVFilter ff_af_afade = {
 438     .name          = "afade",
 439     .description   = NULL_IF_CONFIG_SMALL("Fade in/out input audio."),
 440     .priv_size     = sizeof(AudioFadeContext),
 441     .init          = init,
 442     FILTER_INPUTS(avfilter_af_afade_inputs),
 443     FILTER_OUTPUTS(avfilter_af_afade_outputs),
 444     FILTER_SAMPLEFMTS_ARRAY(sample_fmts),
 445     .priv_class    = &afade_class,
 446     .process_command = process_command,
 447     .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
 448 };
 449
 450 #endif /* CONFIG_AFADE_FILTER */
 451
 452 #if CONFIG_ACROSSFADE_FILTER
 453
 454 static const AVOption acrossfade_options[] = {
 455     { "nb_samples",   "set number of samples for cross fade duration", OFFSET(nb_samples),   AV_OPT_TYPE_INT,    {.i64 = 44100}, 1, INT32_MAX/10, FLAGS },
 456     { "ns",           "set number of samples for cross fade duration", OFFSET(nb_samples),   AV_OPT_TYPE_INT,    {.i64 = 44100}, 1, INT32_MAX/10, FLAGS },
 457     { "duration",     "set cross fade duration",                       OFFSET(duration),     AV_OPT_TYPE_DURATION, {.i64 = 0 },  0, 60000000, FLAGS },
 458     { "d",            "set cross fade duration",                       OFFSET(duration),     AV_OPT_TYPE_DURATION, {.i64 = 0 },  0, 60000000, FLAGS },
 459     { "overlap",      "overlap 1st stream end with 2nd stream start",  OFFSET(overlap),      AV_OPT_TYPE_BOOL,   {.i64 = 1    }, 0,  1, FLAGS },
 460     { "o",            "overlap 1st stream end with 2nd stream start",  OFFSET(overlap),      AV_OPT_TYPE_BOOL,   {.i64 = 1    }, 0,  1, FLAGS },
 461     { "curve1",       "set fade curve type for 1st stream",            OFFSET(curve),        AV_OPT_TYPE_INT,    {.i64 = TRI  }, NONE, NB_CURVES - 1, FLAGS, "curve" },
 462     { "c1",           "set fade curve type for 1st stream",            OFFSET(curve),        AV_OPT_TYPE_INT,    {.i64 = TRI  }, NONE, NB_CURVES - 1, FLAGS, "curve" },
 463     {     "nofade",   "no fade; keep audio as-is",                     0,                    AV_OPT_TYPE_CONST,  {.i64 = NONE }, 0, 0, FLAGS, "curve" },
 464     {     "tri",      "linear slope",                                  0,                    AV_OPT_TYPE_CONST,  {.i64 = TRI  }, 0, 0, FLAGS, "curve" },
 465     {     "qsin",     "quarter of sine wave",                          0,                    AV_OPT_TYPE_CONST,  {.i64 = QSIN }, 0, 0, FLAGS, "curve" },
 466     {     "esin",     "exponential sine wave",                         0,                    AV_OPT_TYPE_CONST,  {.i64 = ESIN }, 0, 0, FLAGS, "curve" },
 467     {     "hsin",     "half of sine wave",                             0,                    AV_OPT_TYPE_CONST,  {.i64 = HSIN }, 0, 0, FLAGS, "curve" },
 468     {     "log",      "logarithmic",                                   0,                    AV_OPT_TYPE_CONST,  {.i64 = LOG  }, 0, 0, FLAGS, "curve" },
 469     {     "ipar",     "inverted parabola",                             0,                    AV_OPT_TYPE_CONST,  {.i64 = IPAR }, 0, 0, FLAGS, "curve" },
 470     {     "qua",      "quadratic",                                     0,                    AV_OPT_TYPE_CONST,  {.i64 = QUA  }, 0, 0, FLAGS, "curve" },
 471     {     "cub",      "cubic",                                         0,                    AV_OPT_TYPE_CONST,  {.i64 = CUB  }, 0, 0, FLAGS, "curve" },
 472     {     "squ",      "square root",                                   0,                    AV_OPT_TYPE_CONST,  {.i64 = SQU  }, 0, 0, FLAGS, "curve" },
 473     {     "cbr",      "cubic root",                                    0,                    AV_OPT_TYPE_CONST,  {.i64 = CBR  }, 0, 0, FLAGS, "curve" },
 474     {     "par",      "parabola",                                      0,                    AV_OPT_TYPE_CONST,  {.i64 = PAR  }, 0, 0, FLAGS, "curve" },
 475     {     "exp",      "exponential",                                   0,                    AV_OPT_TYPE_CONST,  {.i64 = EXP  }, 0, 0, FLAGS, "curve" },
 476     {     "iqsin",    "inverted quarter of sine wave",                 0,                    AV_OPT_TYPE_CONST,  {.i64 = IQSIN}, 0, 0, FLAGS, "curve" },
 477     {     "ihsin",    "inverted half of sine wave",                    0,                    AV_OPT_TYPE_CONST,  {.i64 = IHSIN}, 0, 0, FLAGS, "curve" },
 478     {     "dese",     "double-exponential seat",                       0,                    AV_OPT_TYPE_CONST,  {.i64 = DESE }, 0, 0, FLAGS, "curve" },
 479     {     "desi",     "double-exponential sigmoid",                    0,                    AV_OPT_TYPE_CONST,  {.i64 = DESI }, 0, 0, FLAGS, "curve" },
 480     {     "losi",     "logistic sigmoid",                              0,                    AV_OPT_TYPE_CONST,  {.i64 = LOSI }, 0, 0, FLAGS, "curve" },
 481     {     "sinc",     "sine cardinal function",                        0,                    AV_OPT_TYPE_CONST,  {.i64 = SINC }, 0, 0, FLAGS, "curve" },
 482     {     "isinc",    "inverted sine cardinal function",               0,                    AV_OPT_TYPE_CONST,  {.i64 = ISINC}, 0, 0, FLAGS, "curve" },
 483     {     "quat",     "quartic",                                       0,                    AV_OPT_TYPE_CONST,  {.i64 = QUAT }, 0, 0, FLAGS, "curve" },
 484     {     "quatr",    "quartic root",                                  0,                    AV_OPT_TYPE_CONST,  {.i64 = QUATR}, 0, 0, FLAGS, "curve" },
 485     {     "qsin2",    "squared quarter of sine wave",                  0,                    AV_OPT_TYPE_CONST,  {.i64 = QSIN2}, 0, 0, FLAGS, "curve" },
 486     {     "hsin2",    "squared half of sine wave",                     0,                    AV_OPT_TYPE_CONST,  {.i64 = HSIN2}, 0, 0, FLAGS, "curve" },
 487     { "curve2",       "set fade curve type for 2nd stream",            OFFSET(curve2),       AV_OPT_TYPE_INT,    {.i64 = TRI  }, NONE, NB_CURVES - 1, FLAGS, "curve" },
 488     { "c2",           "set fade curve type for 2nd stream",            OFFSET(curve2),       AV_OPT_TYPE_INT,    {.i64 = TRI  }, NONE, NB_CURVES - 1, FLAGS, "curve" },
 489     { NULL }
 490 };
 491
 492 AVFILTER_DEFINE_CLASS(acrossfade);
 493
 494 #define CROSSFADE_PLANAR(name, type)                                           \
 495 static void crossfade_samples_## name ##p(uint8_t **dst, uint8_t * const *cf0, \
 496                                           uint8_t * const *cf1,                \
 497                                           int nb_samples, int channels,        \
 498                                           int curve0, int curve1)              \
 499 {                                                                              \
 500     int i, c;                                                                  \
 501                                                                                \
 502     for (i = 0; i < nb_samples; i++) {                                         \
 503         double gain0 = fade_gain(curve0, nb_samples - 1 - i, nb_samples,0.,1.);\
 504         double gain1 = fade_gain(curve1, i, nb_samples, 0., 1.);               \
 505         for (c = 0; c < channels; c++) {                                       \
 506             type *d = (type *)dst[c];                                          \
 507             const type *s0 = (type *)cf0[c];                                   \
 508             const type *s1 = (type *)cf1[c];                                   \
 509                                                                                \
 510             d[i] = s0[i] * gain0 + s1[i] * gain1;                              \
 511         }                                                                      \
 512     }                                                                          \
 513 }
 514
 515 #define CROSSFADE(name, type)                                               \
 516 static void crossfade_samples_## name (uint8_t **dst, uint8_t * const *cf0, \
 517                                        uint8_t * const *cf1,                \
 518                                        int nb_samples, int channels,        \
 519                                        int curve0, int curve1)              \
 520 {                                                                           \
 521     type *d = (type *)dst[0];                                               \
 522     const type *s0 = (type *)cf0[0];                                        \
 523     const type *s1 = (type *)cf1[0];                                        \
 524     int i, c, k = 0;                                                        \
 525                                                                             \
 526     for (i = 0; i < nb_samples; i++) {                                      \
 527         double gain0 = fade_gain(curve0, nb_samples - 1-i,nb_samples,0.,1.);\
 528         double gain1 = fade_gain(curve1, i, nb_samples, 0., 1.);            \
 529         for (c = 0; c < channels; c++, k++)                                 \
 530             d[k] = s0[k] * gain0 + s1[k] * gain1;                           \
 531     }                                                                       \
 532 }
 533
 534 CROSSFADE_PLANAR(dbl, double)
 535 CROSSFADE_PLANAR(flt, float)
 536 CROSSFADE_PLANAR(s16, int16_t)
 537 CROSSFADE_PLANAR(s32, int32_t)
 538
 539 CROSSFADE(dbl, double)
 540 CROSSFADE(flt, float)
 541 CROSSFADE(s16, int16_t)
 542 CROSSFADE(s32, int32_t)
 543
 544 static int check_input(AVFilterLink *inlink)
 545 {
 546     const int queued_samples = ff_inlink_queued_samples(inlink);
 547
 548     return ff_inlink_check_available_samples(inlink, queued_samples + 1) == 1;
 549 }
 550
 551 static int activate(AVFilterContext *ctx)
 552 {
 553     AudioFadeContext *s   = ctx->priv;
 554     AVFilterLink *outlink = ctx->outputs[0];
 555     AVFrame *in = NULL, *out, *cf[2] = { NULL };
 556     int ret = 0, nb_samples, status;
 557     int64_t pts;
 558
 559     FF_FILTER_FORWARD_STATUS_BACK_ALL(outlink, ctx);
 560
 561     if (s->passthrough && s->status[0]) {
 562         ret = ff_inlink_consume_frame(ctx->inputs[1], &in);
 563         if (ret > 0) {
 564             in->pts = s->pts;
 565             s->pts += av_rescale_q(in->nb_samples,
 566                       (AVRational){ 1, outlink->sample_rate }, outlink->time_base);
 567             return ff_filter_frame(outlink, in);
 568         } else if (ret < 0) {
 569             return ret;
 570         } else if (ff_inlink_acknowledge_status(ctx->inputs[1], &status, &pts)) {
 571             ff_outlink_set_status(outlink, status, pts);
 572             return 0;
 573         } else if (!ret) {
 574             if (ff_outlink_frame_wanted(outlink)) {
 575                 ff_inlink_request_frame(ctx->inputs[1]);
 576                 return 0;
 577             }
 578         }
 579     }
 580
 581     nb_samples = ff_inlink_queued_samples(ctx->inputs[0]);
 582     if (nb_samples  > s->nb_samples) {
 583         nb_samples -= s->nb_samples;
 584         s->passthrough = 1;
 585         ret = ff_inlink_consume_samples(ctx->inputs[0], nb_samples, nb_samples, &in);
 586         if (ret < 0)
 587             return ret;
 588         in->pts = s->pts;
 589         s->pts += av_rescale_q(in->nb_samples,
 590             (AVRational){ 1, outlink->sample_rate }, outlink->time_base);
 591         return ff_filter_frame(outlink, in);
 592     } else if (s->status[0] && nb_samples >= s->nb_samples &&
 593                ff_inlink_queued_samples(ctx->inputs[1]) >= s->nb_samples) {
 594         if (s->overlap) {
 595             out = ff_get_audio_buffer(outlink, s->nb_samples);
 596             if (!out)
 597                 return AVERROR(ENOMEM);
 598
 599             ret = ff_inlink_consume_samples(ctx->inputs[0], s->nb_samples, s->nb_samples, &cf[0]);
 600             if (ret < 0) {
 601                 av_frame_free(&out);
 602                 return ret;
 603             }
 604
 605             ret = ff_inlink_consume_samples(ctx->inputs[1], s->nb_samples, s->nb_samples, &cf[1]);
 606             if (ret < 0) {
 607                 av_frame_free(&out);
 608                 return ret;
 609             }
 610
 611             s->crossfade_samples(out->extended_data, cf[0]->extended_data,
 612                                  cf[1]->extended_data,
 613                                  s->nb_samples, out->ch_layout.nb_channels,
 614                                  s->curve, s->curve2);
 615             out->pts = s->pts;
 616             s->pts += av_rescale_q(s->nb_samples,
 617                 (AVRational){ 1, outlink->sample_rate }, outlink->time_base);
 618             s->passthrough = 1;
 619             av_frame_free(&cf[0]);
 620             av_frame_free(&cf[1]);
 621             return ff_filter_frame(outlink, out);
 622         } else {
 623             out = ff_get_audio_buffer(outlink, s->nb_samples);
 624             if (!out)
 625                 return AVERROR(ENOMEM);
 626
 627             ret = ff_inlink_consume_samples(ctx->inputs[0], s->nb_samples, s->nb_samples, &cf[0]);
 628             if (ret < 0) {
 629                 av_frame_free(&out);
 630                 return ret;
 631             }
 632
 633             s->fade_samples(out->extended_data, cf[0]->extended_data, s->nb_samples,
 634                             outlink->ch_layout.nb_channels, -1, s->nb_samples - 1, s->nb_samples, s->curve, 0., 1.);
 635             out->pts = s->pts;
 636             s->pts += av_rescale_q(s->nb_samples,
 637                 (AVRational){ 1, outlink->sample_rate }, outlink->time_base);
 638             av_frame_free(&cf[0]);
 639             ret = ff_filter_frame(outlink, out);
 640             if (ret < 0)
 641                 return ret;
 642
 643             out = ff_get_audio_buffer(outlink, s->nb_samples);
 644             if (!out)
 645                 return AVERROR(ENOMEM);
 646
 647             ret = ff_inlink_consume_samples(ctx->inputs[1], s->nb_samples, s->nb_samples, &cf[1]);
 648             if (ret < 0) {
 649                 av_frame_free(&out);
 650                 return ret;
 651             }
 652
 653             s->fade_samples(out->extended_data, cf[1]->extended_data, s->nb_samples,
 654                             outlink->ch_layout.nb_channels, 1, 0, s->nb_samples, s->curve2, 0., 1.);
 655             out->pts = s->pts;
 656             s->pts += av_rescale_q(s->nb_samples,
 657                 (AVRational){ 1, outlink->sample_rate }, outlink->time_base);
 658             s->passthrough = 1;
 659             av_frame_free(&cf[1]);
 660             return ff_filter_frame(outlink, out);
 661         }
 662     } else if (ff_outlink_frame_wanted(outlink)) {
 663         if (!s->status[0] && check_input(ctx->inputs[0]))
 664             s->status[0] = AVERROR_EOF;
 665         s->passthrough = !s->status[0];
 666         if (check_input(ctx->inputs[1])) {
 667             s->status[1] = AVERROR_EOF;
 668             ff_outlink_set_status(outlink, AVERROR_EOF, AV_NOPTS_VALUE);
 669             return 0;
 670         }
 671         if (!s->status[0])
 672             ff_inlink_request_frame(ctx->inputs[0]);
 673         else
 674             ff_inlink_request_frame(ctx->inputs[1]);
 675         return 0;
 676     }
 677
 678     return ret;
 679 }
 680
 681 static int acrossfade_config_output(AVFilterLink *outlink)
 682 {
 683     AVFilterContext *ctx = outlink->src;
 684     AudioFadeContext *s  = ctx->priv;
 685
 686     outlink->time_base   = ctx->inputs[0]->time_base;
 687
 688     switch (outlink->format) {
 689     case AV_SAMPLE_FMT_DBL:  s->crossfade_samples = crossfade_samples_dbl;  break;
 690     case AV_SAMPLE_FMT_DBLP: s->crossfade_samples = crossfade_samples_dblp; break;
 691     case AV_SAMPLE_FMT_FLT:  s->crossfade_samples = crossfade_samples_flt;  break;
 692     case AV_SAMPLE_FMT_FLTP: s->crossfade_samples = crossfade_samples_fltp; break;
 693     case AV_SAMPLE_FMT_S16:  s->crossfade_samples = crossfade_samples_s16;  break;
 694     case AV_SAMPLE_FMT_S16P: s->crossfade_samples = crossfade_samples_s16p; break;
 695     case AV_SAMPLE_FMT_S32:  s->crossfade_samples = crossfade_samples_s32;  break;
 696     case AV_SAMPLE_FMT_S32P: s->crossfade_samples = crossfade_samples_s32p; break;
 697     }
 698
 699     config_output(outlink);
 700
 701     return 0;
 702 }
 703
 704 static AVFrame *get_audio_buffer(AVFilterLink *inlink, int nb_samples)
 705 {
 706     AVFilterContext *ctx = inlink->dst;
 707     AudioFadeContext *s = ctx->priv;
 708
 709     return s->passthrough ?
 710         ff_null_get_audio_buffer   (inlink, nb_samples) :
 711         ff_default_get_audio_buffer(inlink, nb_samples);
 712 }
 713
 714 static const AVFilterPad avfilter_af_acrossfade_inputs[] = {
 715     {
 716         .name         = "crossfade0",
 717         .type         = AVMEDIA_TYPE_AUDIO,
 718         .get_buffer.audio = get_audio_buffer,
 719     },
 720     {
 721         .name         = "crossfade1",
 722         .type         = AVMEDIA_TYPE_AUDIO,
 723         .get_buffer.audio = get_audio_buffer,
 724     },
 725 };
 726
 727 static const AVFilterPad avfilter_af_acrossfade_outputs[] = {
 728     {
 729         .name          = "default",
 730         .type          = AVMEDIA_TYPE_AUDIO,
 731         .config_props  = acrossfade_config_output,
 732     },
 733 };
 734
 735 const AVFilter ff_af_acrossfade = {
 736     .name          = "acrossfade",
 737     .description   = NULL_IF_CONFIG_SMALL("Cross fade two input audio streams."),
 738     .priv_size     = sizeof(AudioFadeContext),
 739     .activate      = activate,
 740     .priv_class    = &acrossfade_class,
 741     FILTER_INPUTS(avfilter_af_acrossfade_inputs),
 742     FILTER_OUTPUTS(avfilter_af_acrossfade_outputs),
 743     FILTER_SAMPLEFMTS_ARRAY(sample_fmts),
 744 };
 745
 746 #endif /* CONFIG_ACROSSFADE_FILTER */