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