1 /* 2 * Copyright (C) 2011-2013 Michael Niedermayer (michaelni@gmx.at) 3 * 4 * This file is part of libswresample 5 * 6 * libswresample 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 * libswresample 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 libswresample; if not, write to the Free Software 18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 19 */ 20 21 #ifndef SWRESAMPLE_SWRESAMPLE_H 22 #define SWRESAMPLE_SWRESAMPLE_H 23 24 /** 25 * @file 26 * @ingroup lswr 27 * libswresample public header 28 */ 29 30 /** 31 * @defgroup lswr Libswresample 32 * @{ 33 * 34 * Libswresample (lswr) is a library that handles audio resampling, sample 35 * format conversion and mixing. 36 * 37 * Interaction with lswr is done through SwrContext, which is 38 * allocated with swr_alloc() or swr_alloc_set_opts(). It is opaque, so all parameters 39 * must be set with the @ref avoptions API. 40 * 41 * The first thing you will need to do in order to use lswr is to allocate 42 * SwrContext. This can be done with swr_alloc() or swr_alloc_set_opts(). If you 43 * are using the former, you must set options through the @ref avoptions API. 44 * The latter function provides the same feature, but it allows you to set some 45 * common options in the same statement. 46 * 47 * For example the following code will setup conversion from planar float sample 48 * format to interleaved signed 16-bit integer, downsampling from 48kHz to 49 * 44.1kHz and downmixing from 5.1 channels to stereo (using the default mixing 50 * matrix). This is using the swr_alloc() function. 51 * @code 52 * SwrContext *swr = swr_alloc(); 53 * av_opt_set_channel_layout(swr, "in_channel_layout", AV_CH_LAYOUT_5POINT1, 0); 54 * av_opt_set_channel_layout(swr, "out_channel_layout", AV_CH_LAYOUT_STEREO, 0); 55 * av_opt_set_int(swr, "in_sample_rate", 48000, 0); 56 * av_opt_set_int(swr, "out_sample_rate", 44100, 0); 57 * av_opt_set_sample_fmt(swr, "in_sample_fmt", AV_SAMPLE_FMT_FLTP, 0); 58 * av_opt_set_sample_fmt(swr, "out_sample_fmt", AV_SAMPLE_FMT_S16, 0); 59 * @endcode 60 * 61 * The same job can be done using swr_alloc_set_opts() as well: 62 * @code 63 * SwrContext *swr = swr_alloc_set_opts(NULL, // we're allocating a new context 64 * AV_CH_LAYOUT_STEREO, // out_ch_layout 65 * AV_SAMPLE_FMT_S16, // out_sample_fmt 66 * 44100, // out_sample_rate 67 * AV_CH_LAYOUT_5POINT1, // in_ch_layout 68 * AV_SAMPLE_FMT_FLTP, // in_sample_fmt 69 * 48000, // in_sample_rate 70 * 0, // log_offset 71 * NULL); // log_ctx 72 * @endcode 73 * 74 * Once all values have been set, it must be initialized with swr_init(). If 75 * you need to change the conversion parameters, you can change the parameters 76 * using @ref AVOptions, as described above in the first example; or by using 77 * swr_alloc_set_opts(), but with the first argument the allocated context. 78 * You must then call swr_init() again. 79 * 80 * The conversion itself is done by repeatedly calling swr_convert(). 81 * Note that the samples may get buffered in swr if you provide insufficient 82 * output space or if sample rate conversion is done, which requires "future" 83 * samples. Samples that do not require future input can be retrieved at any 84 * time by using swr_convert() (in_count can be set to 0). 85 * At the end of conversion the resampling buffer can be flushed by calling 86 * swr_convert() with NULL in and 0 in_count. 87 * 88 * The samples used in the conversion process can be managed with the libavutil 89 * @ref lavu_sampmanip "samples manipulation" API, including av_samples_alloc() 90 * function used in the following example. 91 * 92 * The delay between input and output, can at any time be found by using 93 * swr_get_delay(). 94 * 95 * The following code demonstrates the conversion loop assuming the parameters 96 * from above and caller-defined functions get_input() and handle_output(): 97 * @code 98 * uint8_t **input; 99 * int in_samples; 100 * 101 * while (get_input(&input, &in_samples)) { 102 * uint8_t *output; 103 * int out_samples = av_rescale_rnd(swr_get_delay(swr, 48000) + 104 * in_samples, 44100, 48000, AV_ROUND_UP); 105 * av_samples_alloc(&output, NULL, 2, out_samples, 106 * AV_SAMPLE_FMT_S16, 0); 107 * out_samples = swr_convert(swr, &output, out_samples, 108 * input, in_samples); 109 * handle_output(output, out_samples); 110 * av_freep(&output); 111 * } 112 * @endcode 113 * 114 * When the conversion is finished, the conversion 115 * context and everything associated with it must be freed with swr_free(). 116 * A swr_close() function is also available, but it exists mainly for 117 * compatibility with libavresample, and is not required to be called. 118 * 119 * There will be no memory leak if the data is not completely flushed before 120 * swr_free(). 121 */ 122 123 #include <stdint.h> 124 #include "libavutil/frame.h" 125 #include "libavutil/samplefmt.h" 126 127 #include "libswresample/version.h" 128 129 #if LIBSWRESAMPLE_VERSION_MAJOR < 1 130 #define SWR_CH_MAX 32 ///< Maximum number of channels 131 #endif 132 133 /** 134 * @name Option constants 135 * These constants are used for the @ref avoptions interface for lswr. 136 * @{ 137 * 138 */ 139 140 #define SWR_FLAG_RESAMPLE 1 ///< Force resampling even if equal sample rate 141 //TODO use int resample ? 142 //long term TODO can we enable this dynamically? 143 144 /** Dithering algorithms */ 145 enum SwrDitherType { 146 SWR_DITHER_NONE = 0, 147 SWR_DITHER_RECTANGULAR, 148 SWR_DITHER_TRIANGULAR, 149 SWR_DITHER_TRIANGULAR_HIGHPASS, 150 151 SWR_DITHER_NS = 64, ///< not part of API/ABI 152 SWR_DITHER_NS_LIPSHITZ, 153 SWR_DITHER_NS_F_WEIGHTED, 154 SWR_DITHER_NS_MODIFIED_E_WEIGHTED, 155 SWR_DITHER_NS_IMPROVED_E_WEIGHTED, 156 SWR_DITHER_NS_SHIBATA, 157 SWR_DITHER_NS_LOW_SHIBATA, 158 SWR_DITHER_NS_HIGH_SHIBATA, 159 SWR_DITHER_NB, ///< not part of API/ABI 160 }; 161 162 /** Resampling Engines */ 163 enum SwrEngine { 164 SWR_ENGINE_SWR, /**< SW Resampler */ 165 SWR_ENGINE_SOXR, /**< SoX Resampler */ 166 SWR_ENGINE_NB, ///< not part of API/ABI 167 }; 168 169 /** Resampling Filter Types */ 170 enum SwrFilterType { 171 SWR_FILTER_TYPE_CUBIC, /**< Cubic */ 172 SWR_FILTER_TYPE_BLACKMAN_NUTTALL, /**< Blackman Nuttall Windowed Sinc */ 173 SWR_FILTER_TYPE_KAISER, /**< Kaiser Windowed Sinc */ 174 }; 175 176 /** 177 * @} 178 */ 179 180 /** 181 * The libswresample context. Unlike libavcodec and libavformat, this structure 182 * is opaque. This means that if you would like to set options, you must use 183 * the @ref avoptions API and cannot directly set values to members of the 184 * structure. 185 */ 186 typedef struct SwrContext SwrContext; 187 188 /** 189 * Get the AVClass for SwrContext. It can be used in combination with 190 * AV_OPT_SEARCH_FAKE_OBJ for examining options. 191 * 192 * @see av_opt_find(). 193 * @return the AVClass of SwrContext 194 */ 195 const AVClass *swr_get_class(void); 196 197 /** 198 * @name SwrContext constructor functions 199 * @{ 200 */ 201 202 /** 203 * Allocate SwrContext. 204 * 205 * If you use this function you will need to set the parameters (manually or 206 * with swr_alloc_set_opts()) before calling swr_init(). 207 * 208 * @see swr_alloc_set_opts(), swr_init(), swr_free() 209 * @return NULL on error, allocated context otherwise 210 */ 211 struct SwrContext *swr_alloc(void); 212 213 /** 214 * Initialize context after user parameters have been set. 215 * @note The context must be configured using the AVOption API. 216 * 217 * @see av_opt_set_int() 218 * @see av_opt_set_dict() 219 * 220 * @param[in,out] s Swr context to initialize 221 * @return AVERROR error code in case of failure. 222 */ 223 int swr_init(struct SwrContext *s); 224 225 /** 226 * Check whether an swr context has been initialized or not. 227 * 228 * @param[in] s Swr context to check 229 * @see swr_init() 230 * @return positive if it has been initialized, 0 if not initialized 231 */ 232 int swr_is_initialized(struct SwrContext *s); 233 234 /** 235 * Allocate SwrContext if needed and set/reset common parameters. 236 * 237 * This function does not require s to be allocated with swr_alloc(). On the 238 * other hand, swr_alloc() can use swr_alloc_set_opts() to set the parameters 239 * on the allocated context. 240 * 241 * @param s existing Swr context if available, or NULL if not 242 * @param out_ch_layout output channel layout (AV_CH_LAYOUT_*) 243 * @param out_sample_fmt output sample format (AV_SAMPLE_FMT_*). 244 * @param out_sample_rate output sample rate (frequency in Hz) 245 * @param in_ch_layout input channel layout (AV_CH_LAYOUT_*) 246 * @param in_sample_fmt input sample format (AV_SAMPLE_FMT_*). 247 * @param in_sample_rate input sample rate (frequency in Hz) 248 * @param log_offset logging level offset 249 * @param log_ctx parent logging context, can be NULL 250 * 251 * @see swr_init(), swr_free() 252 * @return NULL on error, allocated context otherwise 253 */ 254 struct SwrContext *swr_alloc_set_opts(struct SwrContext *s, 255 int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate, 256 int64_t in_ch_layout, enum AVSampleFormat in_sample_fmt, int in_sample_rate, 257 int log_offset, void *log_ctx); 258 259 /** 260 * @} 261 * 262 * @name SwrContext destructor functions 263 * @{ 264 */ 265 266 /** 267 * Free the given SwrContext and set the pointer to NULL. 268 * 269 * @param[in] s a pointer to a pointer to Swr context 270 */ 271 void swr_free(struct SwrContext **s); 272 273 /** 274 * Closes the context so that swr_is_initialized() returns 0. 275 * 276 * The context can be brought back to life by running swr_init(), 277 * swr_init() can also be used without swr_close(). 278 * This function is mainly provided for simplifying the usecase 279 * where one tries to support libavresample and libswresample. 280 * 281 * @param[in,out] s Swr context to be closed 282 */ 283 void swr_close(struct SwrContext *s); 284 285 /** 286 * @} 287 * 288 * @name Core conversion functions 289 * @{ 290 */ 291 292 /** Convert audio. 293 * 294 * in and in_count can be set to 0 to flush the last few samples out at the 295 * end. 296 * 297 * If more input is provided than output space then the input will be buffered. 298 * You can avoid this buffering by providing more output space than input. 299 * Conversion will run directly without copying whenever possible. 300 * 301 * @param s allocated Swr context, with parameters set 302 * @param out output buffers, only the first one need be set in case of packed audio 303 * @param out_count amount of space available for output in samples per channel 304 * @param in input buffers, only the first one need to be set in case of packed audio 305 * @param in_count number of input samples available in one channel 306 * 307 * @return number of samples output per channel, negative value on error 308 */ 309 int swr_convert(struct SwrContext *s, uint8_t **out, int out_count, 310 const uint8_t **in , int in_count); 311 312 /** 313 * Convert the next timestamp from input to output 314 * timestamps are in 1/(in_sample_rate * out_sample_rate) units. 315 * 316 * @note There are 2 slightly differently behaving modes. 317 * @li When automatic timestamp compensation is not used, (min_compensation >= FLT_MAX) 318 * in this case timestamps will be passed through with delays compensated 319 * @li When automatic timestamp compensation is used, (min_compensation < FLT_MAX) 320 * in this case the output timestamps will match output sample numbers. 321 * See ffmpeg-resampler(1) for the two modes of compensation. 322 * 323 * @param s[in] initialized Swr context 324 * @param pts[in] timestamp for the next input sample, INT64_MIN if unknown 325 * @see swr_set_compensation(), swr_drop_output(), and swr_inject_silence() are 326 * function used internally for timestamp compensation. 327 * @return the output timestamp for the next output sample 328 */ 329 int64_t swr_next_pts(struct SwrContext *s, int64_t pts); 330 331 /** 332 * @} 333 * 334 * @name Low-level option setting functions 335 * These functons provide a means to set low-level options that is not possible 336 * with the AVOption API. 337 * @{ 338 */ 339 340 /** 341 * Activate resampling compensation ("soft" compensation). This function is 342 * internally called when needed in swr_next_pts(). 343 * 344 * @param[in,out] s allocated Swr context. If it is not initialized, 345 * or SWR_FLAG_RESAMPLE is not set, swr_init() is 346 * called with the flag set. 347 * @param[in] sample_delta delta in PTS per sample 348 * @param[in] compensation_distance number of samples to compensate for 349 * @return >= 0 on success, AVERROR error codes if: 350 * @li @c s is NULL, 351 * @li @c compensation_distance is less than 0, 352 * @li @c compensation_distance is 0 but sample_delta is not, 353 * @li compensation unsupported by resampler, or 354 * @li swr_init() fails when called. 355 */ 356 int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance); 357 358 /** 359 * Set a customized input channel mapping. 360 * 361 * @param[in,out] s allocated Swr context, not yet initialized 362 * @param[in] channel_map customized input channel mapping (array of channel 363 * indexes, -1 for a muted channel) 364 * @return >= 0 on success, or AVERROR error code in case of failure. 365 */ 366 int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map); 367 368 /** 369 * Set a customized remix matrix. 370 * 371 * @param s allocated Swr context, not yet initialized 372 * @param matrix remix coefficients; matrix[i + stride * o] is 373 * the weight of input channel i in output channel o 374 * @param stride offset between lines of the matrix 375 * @return >= 0 on success, or AVERROR error code in case of failure. 376 */ 377 int swr_set_matrix(struct SwrContext *s, const double *matrix, int stride); 378 379 /** 380 * @} 381 * 382 * @name Sample handling functions 383 * @{ 384 */ 385 386 /** 387 * Drops the specified number of output samples. 388 * 389 * This function, along with swr_inject_silence(), is called by swr_next_pts() 390 * if needed for "hard" compensation. 391 * 392 * @param s allocated Swr context 393 * @param count number of samples to be dropped 394 * 395 * @return >= 0 on success, or a negative AVERROR code on failure 396 */ 397 int swr_drop_output(struct SwrContext *s, int count); 398 399 /** 400 * Injects the specified number of silence samples. 401 * 402 * This function, along with swr_drop_output(), is called by swr_next_pts() 403 * if needed for "hard" compensation. 404 * 405 * @param s allocated Swr context 406 * @param count number of samples to be dropped 407 * 408 * @return >= 0 on success, or a negative AVERROR code on failure 409 */ 410 int swr_inject_silence(struct SwrContext *s, int count); 411 412 /** 413 * Gets the delay the next input sample will experience relative to the next output sample. 414 * 415 * Swresample can buffer data if more input has been provided than available 416 * output space, also converting between sample rates needs a delay. 417 * This function returns the sum of all such delays. 418 * The exact delay is not necessarily an integer value in either input or 419 * output sample rate. Especially when downsampling by a large value, the 420 * output sample rate may be a poor choice to represent the delay, similarly 421 * for upsampling and the input sample rate. 422 * 423 * @param s swr context 424 * @param base timebase in which the returned delay will be: 425 * @li if it's set to 1 the returned delay is in seconds 426 * @li if it's set to 1000 the returned delay is in milliseconds 427 * @li if it's set to the input sample rate then the returned 428 * delay is in input samples 429 * @li if it's set to the output sample rate then the returned 430 * delay is in output samples 431 * @li if it's the least common multiple of in_sample_rate and 432 * out_sample_rate then an exact rounding-free delay will be 433 * returned 434 * @returns the delay in 1 / @c base units. 435 */ 436 int64_t swr_get_delay(struct SwrContext *s, int64_t base); 437 438 /** 439 * @} 440 * 441 * @name Configuration accessors 442 * @{ 443 */ 444 445 /** 446 * Return the @ref LIBSWRESAMPLE_VERSION_INT constant. 447 * 448 * This is useful to check if the build-time libswresample has the same version 449 * as the run-time one. 450 * 451 * @returns the unsigned int-typed version 452 */ 453 unsigned swresample_version(void); 454 455 /** 456 * Return the swr build-time configuration. 457 * 458 * @returns the build-time @c ./configure flags 459 */ 460 const char *swresample_configuration(void); 461 462 /** 463 * Return the swr license. 464 * 465 * @returns the license of libswresample, determined at build-time 466 */ 467 const char *swresample_license(void); 468 469 /** 470 * @} 471 * 472 * @name AVFrame based API 473 * @{ 474 */ 475 476 /** 477 * Convert the samples in the input AVFrame and write them to the output AVFrame. 478 * 479 * Input and output AVFrames must have channel_layout, sample_rate and format set. 480 * 481 * If the output AVFrame does not have the data pointers allocated the nb_samples 482 * field will be set using av_frame_get_buffer() 483 * is called to allocate the frame. 484 * 485 * The output AVFrame can be NULL or have fewer allocated samples than required. 486 * In this case, any remaining samples not written to the output will be added 487 * to an internal FIFO buffer, to be returned at the next call to this function 488 * or to swr_convert(). 489 * 490 * If converting sample rate, there may be data remaining in the internal 491 * resampling delay buffer. swr_get_delay() tells the number of 492 * remaining samples. To get this data as output, call this function or 493 * swr_convert() with NULL input. 494 * 495 * If the SwrContext configuration does not match the output and 496 * input AVFrame settings the conversion does not take place and depending on 497 * which AVFrame is not matching AVERROR_OUTPUT_CHANGED, AVERROR_INPUT_CHANGED 498 * or the result of a bitwise-OR of them is returned. 499 * 500 * @see swr_delay() 501 * @see swr_convert() 502 * @see swr_get_delay() 503 * 504 * @param swr audio resample context 505 * @param output output AVFrame 506 * @param input input AVFrame 507 * @return 0 on success, AVERROR on failure or nonmatching 508 * configuration. 509 */ 510 int swr_convert_frame(SwrContext *swr, 511 AVFrame *output, const AVFrame *input); 512 513 /** 514 * Configure or reconfigure the SwrContext using the information 515 * provided by the AVFrames. 516 * 517 * The original resampling context is reset even on failure. 518 * The function calls swr_close() internally if the context is open. 519 * 520 * @see swr_close(); 521 * 522 * @param swr audio resample context 523 * @param output output AVFrame 524 * @param input input AVFrame 525 * @return 0 on success, AVERROR on failure. 526 */ 527 int swr_config_frame(SwrContext *swr, const AVFrame *out, const AVFrame *in); 528 529 /** 530 * @} 531 * @} 532 */ 533 534 #endif /* SWRESAMPLE_SWRESAMPLE_H */ 535