git.sesse.net Git - mlt/blob - src/modules/sox/filter_sox.c

   1 /*
   2  * filter_sox.c -- apply any number of SOX effects using libst
   3  * Copyright (C) 2003-2004 Ushodaya Enterprises Limited
   4  * Author: Dan Dennedy <dan@dennedy.org>
   5  *
   6  * This library 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  * This library 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 this library; if not, write to the Free Software
  18  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  19  */
  20
  21 #include <framework/mlt_filter.h>
  22 #include <framework/mlt_frame.h>
  23 #include <framework/mlt_tokeniser.h>
  24 #include <framework/mlt_log.h>
  25
  26 #include <stdio.h>
  27 #include <stdlib.h>
  28 #include <string.h>
  29 #include <math.h>
  30
  31 // TODO: does not support multiple effects with SoX v14.1.0+
  32
  33 #ifdef SOX14
  34 #       include <sox.h>
  35 #       define ST_EOF SOX_EOF
  36 #       define ST_SUCCESS SOX_SUCCESS
  37 #       define st_sample_t sox_sample_t
  38 #       define eff_t sox_effect_t*
  39 #       define ST_LIB_VERSION_CODE SOX_LIB_VERSION_CODE
  40 #       define ST_LIB_VERSION SOX_LIB_VERSION
  41 #       if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,2,0))
  42 #               define st_size_t size_t
  43 #       else
  44 #               define st_size_t sox_size_t
  45 #       endif
  46 #       define ST_SIGNED_WORD_TO_SAMPLE(d,clips) SOX_SIGNED_16BIT_TO_SAMPLE(d,clips)
  47 #       if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
  48 #               define ST_SSIZE_MIN SOX_SAMPLE_MIN
  49 #       else
  50 #               define ST_SSIZE_MIN SOX_SSIZE_MIN
  51 #       endif
  52 #               define ST_SAMPLE_TO_SIGNED_WORD(d,clips) SOX_SAMPLE_TO_SIGNED_16BIT(d,clips)
  53 #else
  54 #       include <st.h>
  55 #endif
  56
  57 #define BUFFER_LEN 8192
  58 #define AMPLITUDE_NORM 0.2511886431509580 /* -12dBFS */
  59 #define AMPLITUDE_MIN 0.00001
  60 #define DBFSTOAMP(x) pow(10,(x)/20.0)
  61
  62 /** Compute the mean of a set of doubles skipping unset values flagged as -1
  63 */
  64 static inline double mean( double *buf, int count )
  65 {
  66         double mean = 0;
  67         int i;
  68         int j = 0;
  69
  70         for ( i = 0; i < count; i++ )
  71         {
  72                 if ( buf[ i ] != -1.0 )
  73                 {
  74                         mean += buf[ i ];
  75                         j ++;
  76                 }
  77         }
  78         if ( j > 0 )
  79                 mean /= j;
  80
  81         return mean;
  82 }
  83
  84 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
  85 static void delete_effect( eff_t effp )
  86 {
  87         free( effp->priv );
  88         free( (void*)effp->in_encoding );
  89         free( effp );
  90 }
  91 #endif
  92
  93 /** Create an effect state instance for a channels
  94 */
  95 static int create_effect( mlt_filter this, char *value, int count, int channel, int frequency )
  96 {
  97         mlt_tokeniser tokeniser = mlt_tokeniser_init();
  98         char id[ 256 ];
  99         int error = 1;
 100
 101         // Tokenise the effect specification
 102         mlt_tokeniser_parse_new( tokeniser, value, " " );
 103         if ( tokeniser->count < 1 )
 104         {
 105                 mlt_tokeniser_close( tokeniser );
 106                 return error;
 107         }
 108
 109         // Locate the effect
 110         mlt_destructor effect_destructor = mlt_pool_release;
 111 #ifdef SOX14
 112         //fprintf(stderr, "%s: effect %s count %d\n", __FUNCTION__, tokeniser->tokens[0], tokeniser->count );
 113 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 114         sox_effect_handler_t const *eff_handle = sox_find_effect( tokeniser->tokens[0] );
 115         if (eff_handle == NULL ) return error;
 116         eff_t eff = sox_create_effect( eff_handle );
 117         effect_destructor = ( mlt_destructor ) delete_effect;
 118         sox_encodinginfo_t *enc = calloc( 1, sizeof( sox_encodinginfo_t ) );
 119         enc->encoding = SOX_ENCODING_SIGN2;
 120         enc->bits_per_sample = 16;
 121         eff->in_encoding = eff->out_encoding = enc;
 122 #else
 123         eff_t eff = mlt_pool_alloc( sizeof( sox_effect_t ) );
 124         sox_create_effect( eff, sox_find_effect( tokeniser->tokens[0] ) );
 125 #endif
 126         int opt_count = tokeniser->count - 1;
 127 #else
 128         eff_t eff = mlt_pool_alloc( sizeof( struct st_effect ) );
 129         int opt_count = st_geteffect_opt( eff, tokeniser->count, tokeniser->tokens );
 130 #endif
 131
 132         // If valid effect
 133         if ( opt_count != ST_EOF )
 134         {
 135                 // Supply the effect parameters
 136 #ifdef SOX14
 137 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,2,0))
 138                 if ( sox_effect_options( eff, opt_count, &tokeniser->tokens[ tokeniser->count > 1 ? 1 : 0  ] ) == ST_SUCCESS )
 139 #else
 140                 if ( ( * eff->handler.getopts )( eff, opt_count, &tokeniser->tokens[ tokeniser->count > 1 ? 1 : 0  ] ) == ST_SUCCESS )
 141 #endif
 142 #else
 143                 if ( ( * eff->h->getopts )( eff, opt_count, &tokeniser->tokens[ tokeniser->count - opt_count ] ) == ST_SUCCESS )
 144 #endif
 145                 {
 146                         // Set the sox signal parameters
 147 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 148                         eff->in_signal.rate = frequency;
 149                         eff->out_signal.rate = frequency;
 150                         eff->in_signal.channels = 1;
 151                         eff->out_signal.channels = 1;
 152                         eff->in_signal.precision = 16;
 153                         eff->out_signal.precision = 16;
 154                         eff->in_signal.length = 0;
 155                         eff->out_signal.length = 0;
 156 #else
 157                         eff->ininfo.rate = frequency;
 158                         eff->outinfo.rate = frequency;
 159                         eff->ininfo.channels = 1;
 160                         eff->outinfo.channels = 1;
 161 #endif
 162
 163                         // Start the effect
 164 #ifdef SOX14
 165                         if ( ( * eff->handler.start )( eff ) == ST_SUCCESS )
 166 #else
 167                         if ( ( * eff->h->start )( eff ) == ST_SUCCESS )
 168 #endif
 169                         {
 170                                 // Construct id
 171                                 sprintf( id, "_effect_%d_%d", count, channel );
 172
 173                                 // Save the effect state
 174                                 mlt_properties_set_data( MLT_FILTER_PROPERTIES( this ), id, eff, 0, effect_destructor, NULL );
 175                                 error = 0;
 176                         }
 177                 }
 178         }
 179         // Some error occurred so delete the temp effect state
 180         if ( error == 1 )
 181                 effect_destructor( eff );
 182
 183         mlt_tokeniser_close( tokeniser );
 184
 185         return error;
 186 }
 187
 188 /** Get the audio.
 189 */
 190
 191 static int filter_get_audio( mlt_frame frame, void **buffer, mlt_audio_format *format, int *frequency, int *channels, int *samples )
 192 {
 193 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,3,0))
 194         SOX_SAMPLE_LOCALS;
 195 #endif
 196         // Get the filter service
 197         mlt_filter filter = mlt_frame_pop_audio( frame );
 198
 199         // Get the filter properties
 200         mlt_properties filter_properties = MLT_FILTER_PROPERTIES( filter );
 201
 202         mlt_service_lock( MLT_FILTER_SERVICE( filter ) );
 203
 204         // Get the properties
 205         st_sample_t *input_buffer;// = mlt_properties_get_data( filter_properties, "input_buffer", NULL );
 206         st_sample_t *output_buffer = mlt_properties_get_data( filter_properties, "output_buffer", NULL );
 207         int i; // channel
 208         int count = mlt_properties_get_int( filter_properties, "_effect_count" );
 209         int analysis = mlt_properties_get( filter_properties, "effect" ) && !strcmp( mlt_properties_get( filter_properties, "effect" ), "analysis" );
 210
 211         // Get the producer's audio
 212         *format = mlt_audio_s32;
 213         mlt_frame_get_audio( frame, buffer, format, frequency, channels, samples );
 214
 215         // Even though some effects are multi-channel aware, it is not reliable
 216         // We must maintain a separate effect state for each channel
 217         for ( i = 0; i < *channels; i++ )
 218         {
 219                 char id[ 256 ];
 220                 sprintf( id, "_effect_0_%d", i );
 221
 222                 // Get an existing effect state
 223                 eff_t e = mlt_properties_get_data( filter_properties, id, NULL );
 224
 225                 // Validate the existing effect state
 226 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 227                 if ( e != NULL && ( e->in_signal.rate != *frequency ||
 228                                                         e->out_signal.rate != *frequency ) )
 229 #else
 230                 if ( e != NULL && ( e->ininfo.rate != *frequency ||
 231                                                         e->outinfo.rate != *frequency ) )
 232 #endif
 233                         e = NULL;
 234
 235                 // (Re)Create the effect state
 236                 if ( e == NULL )
 237                 {
 238                         int j = 0;
 239
 240                         // Reset the count
 241                         count = 0;
 242
 243                         // Loop over all properties
 244                         for ( j = 0; j < mlt_properties_count( filter_properties ); j ++ )
 245                         {
 246                                 // Get the name of this property
 247                                 char *name = mlt_properties_get_name( filter_properties, j );
 248
 249                                 // If the name does not contain a . and matches effect
 250                                 if ( !strncmp( name, "effect", 6 ) )
 251                                 {
 252                                         // Get the effect specification
 253                                         char *value = mlt_properties_get_value( filter_properties, j );
 254
 255                                         // Create an instance
 256                                         if ( create_effect( filter, value, count, i, *frequency ) == 0 )
 257                                                 count ++;
 258                                 }
 259                         }
 260
 261                         // Save the number of filters
 262                         mlt_properties_set_int( filter_properties, "_effect_count", count );
 263
 264                 }
 265                 if ( *samples > 0 && ( count > 0 || analysis ) )
 266                 {
 267                         input_buffer = (st_sample_t*) *buffer + i * *samples;
 268                         st_sample_t *p = input_buffer;
 269                         st_size_t isamp = *samples;
 270                         st_size_t osamp = *samples;
 271                         int j = *samples + 1;
 272                         char *normalise = mlt_properties_get( filter_properties, "normalise" );
 273                         double normalised_gain = 1.0;
 274
 275                         if ( analysis )
 276                         {
 277                                 // Run analysis to compute a gain level to normalize the audio across entire filter duration
 278                                 double max_power = mlt_properties_get_double( filter_properties, "_max_power" );
 279                                 double peak = mlt_properties_get_double( filter_properties, "_max_peak" );
 280                                 double use_peak = mlt_properties_get_int( filter_properties, "use_peak" );
 281                                 double power = 0;
 282                                 int n = *samples + 1;
 283
 284                                 // Compute power level of samples in this channel of this frame
 285                                 while ( --n )
 286                                 {
 287                                         double s = fabs( *p++ );
 288                                         // Track peak
 289                                         if ( s > peak )
 290                                         {
 291                                                 peak = s;
 292                                                 mlt_properties_set_double( filter_properties, "_max_peak", peak );
 293                                         }
 294                                         power += s * s;
 295                                 }
 296                                 power /= *samples;
 297                                 // Track maximum power
 298                                 if ( power > max_power )
 299                                 {
 300                                         max_power = power;
 301                                         mlt_properties_set_double( filter_properties, "_max_power", max_power );
 302                                 }
 303
 304                                 // Complete analysis the last channel of the last frame.
 305                                 if ( i + 1 == *channels && mlt_filter_get_position( filter, frame ) + 1
 306                                          == mlt_filter_get_length2( filter, frame ) )
 307                                 {
 308                                         double rms = sqrt( max_power / ST_SSIZE_MIN / ST_SSIZE_MIN );
 309                                         char effect[32];
 310
 311                                         // Convert RMS or peak to gain
 312                                         if ( use_peak )
 313                                                 normalised_gain = ST_SSIZE_MIN / -peak;
 314                                         else
 315                                         {
 316                                                 double gain = DBFSTOAMP(-12); // default -12 dBFS
 317                                                 char *p = mlt_properties_get( filter_properties, "analysis_level" );
 318                                                 if (p)
 319                                                 {
 320                                                         gain = mlt_properties_get_double( filter_properties, "analysis_level" );
 321                                                         if ( strstr( p, "dB" ) )
 322                                                                 gain = DBFSTOAMP( gain );
 323                                                 }
 324                                                 normalised_gain = gain / rms;
 325                                         }
 326
 327                                         // Set properties for serialization
 328                                         snprintf( effect, sizeof(effect), "vol %f", normalised_gain );
 329                                         effect[31] = 0;
 330                                         mlt_properties_set( filter_properties, "effect", effect );
 331                                         mlt_properties_set( filter_properties, "analyze", NULL );
 332
 333                                         // Show output comparable to normalize --no-adjust --fractions
 334                                         mlt_properties_set_double( filter_properties, "level", rms );
 335                                         mlt_properties_set_double( filter_properties, "gain", normalised_gain );
 336                                         mlt_properties_set_double( filter_properties, "peak", -peak / ST_SSIZE_MIN );
 337                                 }
 338
 339                                 // restore some variables
 340                                 p = input_buffer;
 341                         }
 342
 343                         if ( normalise )
 344                         {
 345                                 int window = mlt_properties_get_int( filter_properties, "window" );
 346                                 double *smooth_buffer = mlt_properties_get_data( filter_properties, "smooth_buffer", NULL );
 347                                 double max_gain = mlt_properties_get_double( filter_properties, "max_gain" );
 348                                 double rms = 0;
 349
 350                                 // Default the maximum gain factor to 20dBFS
 351                                 if ( max_gain == 0 )
 352                                         max_gain = 10.0;
 353
 354                                 // Compute rms amplitude
 355                                 while( --j )
 356                                 {
 357                                         rms += ( double )*p * ( double )*p;
 358                                         p ++;
 359                                 }
 360                                 rms = sqrt( rms / *samples / ST_SSIZE_MIN / ST_SSIZE_MIN );
 361
 362                                 // The smoothing buffer prevents radical shifts in the gain level
 363                                 if ( window > 0 && smooth_buffer != NULL )
 364                                 {
 365                                         int smooth_index = mlt_properties_get_int( filter_properties, "_smooth_index" );
 366                                         smooth_buffer[ smooth_index ] = rms;
 367
 368                                         // Ignore very small values that adversely affect the mean
 369                                         if ( rms > AMPLITUDE_MIN )
 370                                                 mlt_properties_set_int( filter_properties, "_smooth_index", ( smooth_index + 1 ) % window );
 371
 372                                         // Smoothing is really just a mean over the past N values
 373                                         normalised_gain = AMPLITUDE_NORM / mean( smooth_buffer, window );
 374                                 }
 375                                 else if ( rms > 0 )
 376                                 {
 377                                         // Determine gain to apply as current amplitude
 378                                         normalised_gain = AMPLITUDE_NORM / rms;
 379                                 }
 380
 381                                 //printf("filter_sox: rms %.3f gain %.3f\n", rms, normalised_gain );
 382
 383                                 // Govern the maximum gain
 384                                 if ( normalised_gain > max_gain )
 385                                         normalised_gain = max_gain;
 386                         }
 387
 388                         // For each effect
 389                         for ( j = 0; j < count; j++ )
 390                         {
 391                                 sprintf( id, "_effect_%d_%d", j, i );
 392                                 e = mlt_properties_get_data( filter_properties, id, NULL );
 393
 394                                 // We better have this guy
 395                                 if ( e != NULL )
 396                                 {
 397                                         float saved_gain = 1.0;
 398
 399                                         // XXX: hack to apply the normalised gain level to the vol effect
 400 #ifdef SOX14
 401                                         if ( normalise && strcmp( e->handler.name, "vol" ) == 0 )
 402 #else
 403                                         if ( normalise && strcmp( e->name, "vol" ) == 0 )
 404 #endif
 405                                         {
 406                                                 float *f = ( float * )( e->priv );
 407                                                 saved_gain = *f;
 408                                                 *f = saved_gain * normalised_gain;
 409                                         }
 410
 411                                         // Apply the effect
 412 #ifdef SOX14
 413                                         if ( ( * e->handler.flow )( e, input_buffer, output_buffer, &isamp, &osamp ) != ST_SUCCESS )
 414 #else
 415                                         if ( ( * e->h->flow )( e, input_buffer, output_buffer, &isamp, &osamp ) != ST_SUCCESS )
 416 #endif
 417                                         {
 418                                                 mlt_log_warning( MLT_FILTER_SERVICE(filter), "effect processing failed\n" );
 419                                         }
 420
 421                                         // XXX: hack to restore the original vol gain to prevent accumulation
 422 #ifdef SOX14
 423                                         if ( normalise && strcmp( e->handler.name, "vol" ) == 0 )
 424 #else
 425                                         if ( normalise && strcmp( e->name, "vol" ) == 0 )
 426 #endif
 427                                         {
 428                                                 float *f = ( float * )( e->priv );
 429                                                 *f = saved_gain;
 430                                         }
 431                                 }
 432                         }
 433
 434                         // Write back
 435                         memcpy( input_buffer, output_buffer, *samples * sizeof(st_sample_t) );
 436                 }
 437         }
 438
 439         mlt_service_unlock( MLT_FILTER_SERVICE( filter ) );
 440
 441         return 0;
 442 }
 443
 444 /** Filter processing.
 445 */
 446
 447 static mlt_frame filter_process( mlt_filter this, mlt_frame frame )
 448 {
 449         if ( mlt_frame_is_test_audio( frame ) == 0 )
 450         {
 451                 // Add the filter to the frame
 452                 mlt_frame_push_audio( frame, this );
 453                 mlt_frame_push_audio( frame, filter_get_audio );
 454
 455                 // Parse the window property and allocate smoothing buffer if needed
 456                 mlt_properties properties = MLT_FILTER_PROPERTIES( this );
 457                 int window = mlt_properties_get_int( properties, "window" );
 458                 if ( mlt_properties_get( properties, "smooth_buffer" ) == NULL && window > 1 )
 459                 {
 460                         // Create a smoothing buffer for the calculated "max power" of frame of audio used in normalisation
 461                         double *smooth_buffer = (double*) calloc( window, sizeof( double ) );
 462                         int i;
 463                         for ( i = 0; i < window; i++ )
 464                                 smooth_buffer[ i ] = -1.0;
 465                         mlt_properties_set_data( properties, "smooth_buffer", smooth_buffer, 0, free, NULL );
 466                 }
 467         }
 468
 469         return frame;
 470 }
 471
 472 /** Constructor for the filter.
 473 */
 474
 475 mlt_filter filter_sox_init( mlt_profile profile, mlt_service_type type, const char *id, char *arg )
 476 {
 477         mlt_filter this = mlt_filter_new( );
 478         if ( this != NULL )
 479         {
 480                 void *input_buffer = mlt_pool_alloc( BUFFER_LEN );
 481                 void *output_buffer = mlt_pool_alloc( BUFFER_LEN );
 482                 mlt_properties properties = MLT_FILTER_PROPERTIES( this );
 483
 484                 this->process = filter_process;
 485
 486                 if ( !strncmp( id, "sox.", 4 ) )
 487                 {
 488                         char *s = malloc( strlen( id ) + ( arg? strlen( arg ) + 2 : 1 ) );
 489                         strcpy( s, id + 4 );
 490                         if ( arg )
 491                         {
 492                                 strcat( s, " " );
 493                                 strcat( s, arg );
 494                         }
 495                         mlt_properties_set( properties, "effect", s );
 496                         free( s );
 497                 }
 498                 else if ( arg )
 499                         mlt_properties_set( properties, "effect", arg );
 500                 mlt_properties_set_data( properties, "input_buffer", input_buffer, BUFFER_LEN, mlt_pool_release, NULL );
 501                 mlt_properties_set_data( properties, "output_buffer", output_buffer, BUFFER_LEN, mlt_pool_release, NULL );
 502                 mlt_properties_set_int( properties, "window", 75 );
 503                 mlt_properties_set( properties, "version", sox_version() );
 504         }
 505         return this;
 506 }
 507
 508 // What to do when a libst internal failure occurs
 509 void cleanup(void){}