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