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
  25 #include <stdio.h>
  26 #include <stdlib.h>
  27 #include <string.h>
  28 #include <math.h>
  29
  30 // TODO: does not support multiple effects with SoX v14.1.0+
  31
  32 #ifdef SOX14
  33 #       include <sox.h>
  34 #       define ST_EOF SOX_EOF
  35 #       define ST_SUCCESS SOX_SUCCESS
  36 #       define st_sample_t sox_sample_t
  37 #       define eff_t sox_effect_t*
  38 #       define st_size_t sox_size_t
  39 #       define ST_LIB_VERSION_CODE SOX_LIB_VERSION_CODE
  40 #       define ST_LIB_VERSION SOX_LIB_VERSION
  41 #       define ST_SIGNED_WORD_TO_SAMPLE(d,clips) SOX_SIGNED_16BIT_TO_SAMPLE(d,clips)
  42 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
  43 #       define ST_SSIZE_MIN SOX_SAMPLE_MIN
  44 #else
  45 #       define ST_SSIZE_MIN SOX_SSIZE_MIN
  46 #endif
  47 #       define ST_SAMPLE_TO_SIGNED_WORD(d,clips) SOX_SAMPLE_TO_SIGNED_16BIT(d,clips)
  48 #else
  49 #       include <st.h>
  50 #endif
  51
  52 #define BUFFER_LEN 8192
  53 #define AMPLITUDE_NORM 0.2511886431509580 /* -12dBFS */
  54 #define AMPLITUDE_MIN 0.00001
  55
  56 /** Compute the mean of a set of doubles skipping unset values flagged as -1
  57 */
  58 static inline double mean( double *buf, int count )
  59 {
  60         double mean = 0;
  61         int i;
  62         int j = 0;
  63
  64         for ( i = 0; i < count; i++ )
  65         {
  66                 if ( buf[ i ] != -1.0 )
  67                 {
  68                         mean += buf[ i ];
  69                         j ++;
  70                 }
  71         }
  72         if ( j > 0 )
  73                 mean /= j;
  74
  75         return mean;
  76 }
  77
  78 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
  79 static void delete_effect( eff_t effp )
  80 {
  81         free( effp->priv );
  82         free( (void*)effp->in_encoding );
  83         free( effp );
  84 }
  85 #endif
  86
  87 /** Create an effect state instance for a channels
  88 */
  89 static int create_effect( mlt_filter this, char *value, int count, int channel, int frequency )
  90 {
  91         mlt_tokeniser tokeniser = mlt_tokeniser_init();
  92         char id[ 256 ];
  93         int error = 1;
  94
  95         // Tokenise the effect specification
  96         mlt_tokeniser_parse_new( tokeniser, value, " " );
  97         if ( tokeniser->count < 1 )
  98                 return error;
  99
 100         // Locate the effect
 101         mlt_destructor effect_destructor = mlt_pool_release;
 102 #ifdef SOX14
 103         //fprintf(stderr, "%s: effect %s count %d\n", __FUNCTION__, tokeniser->tokens[0], tokeniser->count );
 104 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 105         eff_t eff = sox_create_effect( sox_find_effect( tokeniser->tokens[0] ) );
 106         effect_destructor = ( mlt_destructor ) delete_effect;
 107         sox_encodinginfo_t *enc = calloc( 1, sizeof( sox_encodinginfo_t ) );
 108         enc->encoding = SOX_ENCODING_SIGN2;
 109         enc->bits_per_sample = 16;
 110         eff->in_encoding = eff->out_encoding = enc;
 111 #else
 112         eff_t eff = mlt_pool_alloc( sizeof( sox_effect_t ) );
 113         sox_create_effect( eff, sox_find_effect( tokeniser->tokens[0] ) );
 114 #endif
 115         int opt_count = tokeniser->count - 1;
 116 #else
 117         eff_t eff = mlt_pool_alloc( sizeof( struct st_effect ) );
 118         int opt_count = st_geteffect_opt( eff, tokeniser->count, tokeniser->tokens );
 119 #endif
 120
 121         // If valid effect
 122         if ( opt_count != ST_EOF )
 123         {
 124                 // Supply the effect parameters
 125 #ifdef SOX14
 126                 if ( ( * eff->handler.getopts )( eff, opt_count, &tokeniser->tokens[ tokeniser->count > 1 ? 1 : 0  ] ) == ST_SUCCESS )
 127 #else
 128                 if ( ( * eff->h->getopts )( eff, opt_count, &tokeniser->tokens[ tokeniser->count - opt_count ] ) == ST_SUCCESS )
 129 #endif
 130                 {
 131                         // Set the sox signal parameters
 132 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 133                         eff->in_signal.rate = frequency;
 134                         eff->out_signal.rate = frequency;
 135                         eff->in_signal.channels = 1;
 136                         eff->out_signal.channels = 1;
 137                         eff->in_signal.precision = 16;
 138                         eff->out_signal.precision = 16;
 139                         eff->in_signal.length = 0;
 140                         eff->out_signal.length = 0;
 141 #else
 142                         eff->ininfo.rate = frequency;
 143                         eff->outinfo.rate = frequency;
 144                         eff->ininfo.channels = 1;
 145                         eff->outinfo.channels = 1;
 146 #endif
 147
 148                         // Start the effect
 149 #ifdef SOX14
 150                         if ( ( * eff->handler.start )( eff ) == ST_SUCCESS )
 151 #else
 152                         if ( ( * eff->h->start )( eff ) == ST_SUCCESS )
 153 #endif
 154                         {
 155                                 // Construct id
 156                                 sprintf( id, "_effect_%d_%d", count, channel );
 157
 158                                 // Save the effect state
 159                                 mlt_properties_set_data( MLT_FILTER_PROPERTIES( this ), id, eff, 0, effect_destructor, NULL );
 160                                 error = 0;
 161                         }
 162                 }
 163         }
 164         // Some error occurred so delete the temp effect state
 165         if ( error == 1 )
 166                 effect_destructor( eff );
 167
 168         mlt_tokeniser_close( tokeniser );
 169
 170         return error;
 171 }
 172
 173 /** Get the audio.
 174 */
 175
 176 static int filter_get_audio( mlt_frame frame, int16_t **buffer, mlt_audio_format *format, int *frequency, int *channels, int *samples )
 177 {
 178         // Get the properties of the frame
 179         mlt_properties properties = MLT_FRAME_PROPERTIES( frame );
 180
 181         // Get the filter service
 182         mlt_filter filter = mlt_frame_pop_audio( frame );
 183
 184         // Get the filter properties
 185         mlt_properties filter_properties = MLT_FILTER_PROPERTIES( filter );
 186
 187         // Get the properties
 188         st_sample_t *input_buffer = mlt_properties_get_data( filter_properties, "input_buffer", NULL );
 189         st_sample_t *output_buffer = mlt_properties_get_data( filter_properties, "output_buffer", NULL );
 190         int channels_avail = *channels;
 191         int i; // channel
 192         int count = mlt_properties_get_int( filter_properties, "_effect_count" );
 193
 194         // Get the producer's audio
 195         mlt_frame_get_audio( frame, buffer, format, frequency, &channels_avail, samples );
 196
 197         // Duplicate channels as necessary
 198         if ( channels_avail < *channels )
 199         {
 200                 int size = *channels * *samples * sizeof( int16_t );
 201                 int16_t *new_buffer = mlt_pool_alloc( size );
 202                 int j, k = 0;
 203
 204                 // Duplicate the existing channels
 205                 for ( i = 0; i < *samples; i++ )
 206                 {
 207                         for ( j = 0; j < *channels; j++ )
 208                         {
 209                                 new_buffer[ ( i * *channels ) + j ] = (*buffer)[ ( i * channels_avail ) + k ];
 210                                 k = ( k + 1 ) % channels_avail;
 211                         }
 212                 }
 213
 214                 // Update the audio buffer now - destroys the old
 215                 mlt_properties_set_data( properties, "audio", new_buffer, size, ( mlt_destructor )mlt_pool_release, NULL );
 216
 217                 *buffer = new_buffer;
 218         }
 219         else if ( channels_avail == 6 && *channels == 2 )
 220         {
 221                 // Nasty hack for ac3 5.1 audio - may be a cause of failure?
 222                 int size = *channels * *samples * sizeof( int16_t );
 223                 int16_t *new_buffer = mlt_pool_alloc( size );
 224
 225                 // Drop all but the first *channels
 226                 for ( i = 0; i < *samples; i++ )
 227                 {
 228                         new_buffer[ ( i * *channels ) + 0 ] = (*buffer)[ ( i * channels_avail ) + 2 ];
 229                         new_buffer[ ( i * *channels ) + 1 ] = (*buffer)[ ( i * channels_avail ) + 3 ];
 230                 }
 231
 232                 // Update the audio buffer now - destroys the old
 233                 mlt_properties_set_data( properties, "audio", new_buffer, size, ( mlt_destructor )mlt_pool_release, NULL );
 234
 235                 *buffer = new_buffer;
 236         }
 237
 238         // Even though some effects are multi-channel aware, it is not reliable
 239         // We must maintain a separate effect state for each channel
 240         for ( i = 0; i < *channels; i++ )
 241         {
 242                 char id[ 256 ];
 243                 sprintf( id, "_effect_0_%d", i );
 244
 245                 // Get an existing effect state
 246                 eff_t e = mlt_properties_get_data( filter_properties, id, NULL );
 247
 248                 // Validate the existing effect state
 249 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 250                 if ( e != NULL && ( e->in_signal.rate != *frequency ||
 251                                                         e->out_signal.rate != *frequency ) )
 252 #else
 253                 if ( e != NULL && ( e->ininfo.rate != *frequency ||
 254                                                         e->outinfo.rate != *frequency ) )
 255 #endif
 256                         e = NULL;
 257
 258                 // (Re)Create the effect state
 259                 if ( e == NULL )
 260                 {
 261                         int j = 0;
 262
 263                         // Reset the count
 264                         count = 0;
 265
 266                         // Loop over all properties
 267                         for ( j = 0; j < mlt_properties_count( filter_properties ); j ++ )
 268                         {
 269                                 // Get the name of this property
 270                                 char *name = mlt_properties_get_name( filter_properties, j );
 271
 272                                 // If the name does not contain a . and matches effect
 273                                 if ( !strncmp( name, "effect", 6 ) )
 274                                 {
 275                                         // Get the effect specification
 276                                         char *value = mlt_properties_get( filter_properties, name );
 277
 278                                         // Create an instance
 279                                         if ( create_effect( filter, value, count, i, *frequency ) == 0 )
 280                                                 count ++;
 281                                 }
 282                         }
 283
 284                         // Save the number of filters
 285                         mlt_properties_set_int( filter_properties, "_effect_count", count );
 286
 287                 }
 288                 if ( *samples > 0 && count > 0 )
 289                 {
 290                         st_sample_t *p = input_buffer;
 291                         st_sample_t *end = p + *samples;
 292                         int16_t *q = *buffer + i;
 293                         st_size_t isamp = *samples;
 294                         st_size_t osamp = *samples;
 295                         double rms = 0;
 296                         int j;
 297                         char *normalise = mlt_properties_get( filter_properties, "normalise" );
 298                         double normalised_gain = 1.0;
 299 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(13,0,0))
 300                         st_sample_t dummy_clipped_count = 0;
 301 #endif
 302
 303                         // Convert to sox encoding
 304                         while( p != end )
 305                         {
 306 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(13,0,0))
 307                                 *p = ST_SIGNED_WORD_TO_SAMPLE( *q, dummy_clipped_count );
 308 #else
 309                                 *p = ST_SIGNED_WORD_TO_SAMPLE( *q );
 310 #endif
 311                                 // Compute rms amplitude while we are accessing each sample
 312                                 rms += ( double )*p * ( double )*p;
 313
 314                                 p ++;
 315                                 q += *channels;
 316                         }
 317
 318                         // Compute final rms amplitude
 319                         rms = sqrt( rms / *samples / ST_SSIZE_MIN / ST_SSIZE_MIN );
 320
 321                         if ( normalise )
 322                         {
 323                                 int window = mlt_properties_get_int( filter_properties, "window" );
 324                                 double *smooth_buffer = mlt_properties_get_data( filter_properties, "smooth_buffer", NULL );
 325                                 double max_gain = mlt_properties_get_double( filter_properties, "max_gain" );
 326
 327                                 // Default the maximum gain factor to 20dBFS
 328                                 if ( max_gain == 0 )
 329                                         max_gain = 10.0;
 330
 331                                 // The smoothing buffer prevents radical shifts in the gain level
 332                                 if ( window > 0 && smooth_buffer != NULL )
 333                                 {
 334                                         int smooth_index = mlt_properties_get_int( filter_properties, "_smooth_index" );
 335                                         smooth_buffer[ smooth_index ] = rms;
 336
 337                                         // Ignore very small values that adversely affect the mean
 338                                         if ( rms > AMPLITUDE_MIN )
 339                                                 mlt_properties_set_int( filter_properties, "_smooth_index", ( smooth_index + 1 ) % window );
 340
 341                                         // Smoothing is really just a mean over the past N values
 342                                         normalised_gain = AMPLITUDE_NORM / mean( smooth_buffer, window );
 343                                 }
 344                                 else if ( rms > 0 )
 345                                 {
 346                                         // Determine gain to apply as current amplitude
 347                                         normalised_gain = AMPLITUDE_NORM / rms;
 348                                 }
 349
 350                                 //printf("filter_sox: rms %.3f gain %.3f\n", rms, normalised_gain );
 351
 352                                 // Govern the maximum gain
 353                                 if ( normalised_gain > max_gain )
 354                                         normalised_gain = max_gain;
 355                         }
 356
 357                         // For each effect
 358                         for ( j = 0; j < count; j++ )
 359                         {
 360                                 sprintf( id, "_effect_%d_%d", j, i );
 361                                 e = mlt_properties_get_data( filter_properties, id, NULL );
 362
 363                                 // We better have this guy
 364                                 if ( e != NULL )
 365                                 {
 366                                         float saved_gain = 1.0;
 367
 368                                         // XXX: hack to apply the normalised gain level to the vol effect
 369 #ifdef SOX14
 370                                         if ( normalise && strcmp( e->handler.name, "vol" ) == 0 )
 371 #else
 372                                         if ( normalise && strcmp( e->name, "vol" ) == 0 )
 373 #endif
 374                                         {
 375                                                 float *f = ( float * )( e->priv );
 376                                                 saved_gain = *f;
 377                                                 *f = saved_gain * normalised_gain;
 378                                         }
 379
 380                                         // Apply the effect
 381 #ifdef SOX14
 382                                         if ( ( * e->handler.flow )( e, input_buffer, output_buffer, &isamp, &osamp ) == ST_SUCCESS )
 383 #else
 384                                         if ( ( * e->h->flow )( e, input_buffer, output_buffer, &isamp, &osamp ) == ST_SUCCESS )
 385 #endif
 386                                         {
 387                                                 // Swap input and output buffer pointers for subsequent effects
 388                                                 p = input_buffer;
 389                                                 input_buffer = output_buffer;
 390                                                 output_buffer = p;
 391                                         }
 392
 393                                         // XXX: hack to restore the original vol gain to prevent accumulation
 394 #ifdef SOX14
 395                                         if ( normalise && strcmp( e->handler.name, "vol" ) == 0 )
 396 #else
 397                                         if ( normalise && strcmp( e->name, "vol" ) == 0 )
 398 #endif
 399                                         {
 400                                                 float *f = ( float * )( e->priv );
 401                                                 *f = saved_gain;
 402                                         }
 403                                 }
 404                         }
 405
 406                         // Convert back to signed 16bit
 407                         p = input_buffer;
 408                         q = *buffer + i;
 409                         end = p + *samples;
 410                         while ( p != end )
 411                         {
 412 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(13,0,0))
 413                                 *q = ST_SAMPLE_TO_SIGNED_WORD( *p ++, dummy_clipped_count );
 414 #else
 415                                 *q = ST_SAMPLE_TO_SIGNED_WORD( *p ++ );
 416 #endif
 417                                 q += *channels;
 418                         }
 419                 }
 420         }
 421
 422         return 0;
 423 }
 424
 425 /** Filter processing.
 426 */
 427
 428 static mlt_frame filter_process( mlt_filter this, mlt_frame frame )
 429 {
 430         if ( mlt_frame_is_test_audio( frame ) == 0 )
 431         {
 432                 // Add the filter to the frame
 433                 mlt_frame_push_audio( frame, this );
 434                 mlt_frame_push_audio( frame, filter_get_audio );
 435
 436                 // Parse the window property and allocate smoothing buffer if needed
 437                 mlt_properties properties = MLT_FILTER_PROPERTIES( this );
 438                 int window = mlt_properties_get_int( properties, "window" );
 439                 if ( mlt_properties_get( properties, "smooth_buffer" ) == NULL && window > 1 )
 440                 {
 441                         // Create a smoothing buffer for the calculated "max power" of frame of audio used in normalisation
 442                         double *smooth_buffer = (double*) calloc( window, sizeof( double ) );
 443                         int i;
 444                         for ( i = 0; i < window; i++ )
 445                                 smooth_buffer[ i ] = -1.0;
 446                         mlt_properties_set_data( properties, "smooth_buffer", smooth_buffer, 0, free, NULL );
 447                 }
 448         }
 449
 450         return frame;
 451 }
 452
 453 /** Constructor for the filter.
 454 */
 455
 456 mlt_filter filter_sox_init( mlt_profile profile, mlt_service_type type, const char *id, char *arg )
 457 {
 458         mlt_filter this = mlt_filter_new( );
 459         if ( this != NULL )
 460         {
 461                 void *input_buffer = mlt_pool_alloc( BUFFER_LEN );
 462                 void *output_buffer = mlt_pool_alloc( BUFFER_LEN );
 463                 mlt_properties properties = MLT_FILTER_PROPERTIES( this );
 464
 465                 this->process = filter_process;
 466
 467                 if ( arg != NULL )
 468                         mlt_properties_set( properties, "effect", arg );
 469                 mlt_properties_set_data( properties, "input_buffer", input_buffer, BUFFER_LEN, mlt_pool_release, NULL );
 470                 mlt_properties_set_data( properties, "output_buffer", output_buffer, BUFFER_LEN, mlt_pool_release, NULL );
 471                 mlt_properties_set_int( properties, "window", 75 );
 472         }
 473         return this;
 474 }
 475
 476 // What to do when a libst internal failure occurs
 477 void cleanup(void){}
 478
 479 // Is there a build problem with my sox-devel package?
 480 #ifndef gsm_create
 481 void gsm_create(void){}
 482 #endif
 483 #ifndef gsm_decode
 484 void gsm_decode(void){}
 485 #endif
 486 #ifndef gdm_encode
 487 void gsm_encode(void){}
 488 #endif
 489 #ifndef gsm_destroy
 490 void gsm_destroy(void){}
 491 #endif
 492 #ifndef gsm_option
 493 void gsm_option(void){}
 494 #endif