• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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  * Audio resampling, sample format conversion and mixing library.
35  *
36  * Interaction with lswr is done through SwrContext, which is
37  * allocated with swr_alloc() or swr_alloc_set_opts(). It is opaque, so all parameters
38  * must be set with the @ref avoptions API.
39  *
40  * The first thing you will need to do in order to use lswr is to allocate
41  * SwrContext. This can be done with swr_alloc() or swr_alloc_set_opts(). If you
42  * are using the former, you must set options through the @ref avoptions API.
43  * The latter function provides the same feature, but it allows you to set some
44  * common options in the same statement.
45  *
46  * For example the following code will setup conversion from planar float sample
47  * format to interleaved signed 16-bit integer, downsampling from 48kHz to
48  * 44.1kHz and downmixing from 5.1 channels to stereo (using the default mixing
49  * matrix). This is using the swr_alloc() function.
50  * @code
51  * SwrContext *swr = swr_alloc();
52  * av_opt_set_channel_layout(swr, "in_channel_layout",  AV_CH_LAYOUT_5POINT1, 0);
53  * av_opt_set_channel_layout(swr, "out_channel_layout", AV_CH_LAYOUT_STEREO,  0);
54  * av_opt_set_int(swr, "in_sample_rate",     48000,                0);
55  * av_opt_set_int(swr, "out_sample_rate",    44100,                0);
56  * av_opt_set_sample_fmt(swr, "in_sample_fmt",  AV_SAMPLE_FMT_FLTP, 0);
57  * av_opt_set_sample_fmt(swr, "out_sample_fmt", AV_SAMPLE_FMT_S16,  0);
58  * @endcode
59  *
60  * The same job can be done using swr_alloc_set_opts() as well:
61  * @code
62  * SwrContext *swr = swr_alloc_set_opts(NULL,  // we're allocating a new context
63  *                       AV_CH_LAYOUT_STEREO,  // out_ch_layout
64  *                       AV_SAMPLE_FMT_S16,    // out_sample_fmt
65  *                       44100,                // out_sample_rate
66  *                       AV_CH_LAYOUT_5POINT1, // in_ch_layout
67  *                       AV_SAMPLE_FMT_FLTP,   // in_sample_fmt
68  *                       48000,                // in_sample_rate
69  *                       0,                    // log_offset
70  *                       NULL);                // log_ctx
71  * @endcode
72  *
73  * Once all values have been set, it must be initialized with swr_init(). If
74  * you need to change the conversion parameters, you can change the parameters
75  * using @ref AVOptions, as described above in the first example; or by using
76  * swr_alloc_set_opts(), but with the first argument the allocated context.
77  * You must then call swr_init() again.
78  *
79  * The conversion itself is done by repeatedly calling swr_convert().
80  * Note that the samples may get buffered in swr if you provide insufficient
81  * output space or if sample rate conversion is done, which requires "future"
82  * samples. Samples that do not require future input can be retrieved at any
83  * time by using swr_convert() (in_count can be set to 0).
84  * At the end of conversion the resampling buffer can be flushed by calling
85  * swr_convert() with NULL in and 0 in_count.
86  *
87  * The samples used in the conversion process can be managed with the libavutil
88  * @ref lavu_sampmanip "samples manipulation" API, including av_samples_alloc()
89  * function used in the following example.
90  *
91  * The delay between input and output, can at any time be found by using
92  * swr_get_delay().
93  *
94  * The following code demonstrates the conversion loop assuming the parameters
95  * from above and caller-defined functions get_input() and handle_output():
96  * @code
97  * uint8_t **input;
98  * int in_samples;
99  *
100  * while (get_input(&input, &in_samples)) {
101  *     uint8_t *output;
102  *     int out_samples = av_rescale_rnd(swr_get_delay(swr, 48000) +
103  *                                      in_samples, 44100, 48000, AV_ROUND_UP);
104  *     av_samples_alloc(&output, NULL, 2, out_samples,
105  *                      AV_SAMPLE_FMT_S16, 0);
106  *     out_samples = swr_convert(swr, &output, out_samples,
107  *                                      input, in_samples);
108  *     handle_output(output, out_samples);
109  *     av_freep(&output);
110  * }
111  * @endcode
112  *
113  * When the conversion is finished, the conversion
114  * context and everything associated with it must be freed with swr_free().
115  * A swr_close() function is also available, but it exists mainly for
116  * compatibility with libavresample, and is not required to be called.
117  *
118  * There will be no memory leak if the data is not completely flushed before
119  * swr_free().
120  */
121 
122 #include <stdint.h>
123 #include "libavutil/channel_layout.h"
124 #include "libavutil/frame.h"
125 #include "libavutil/samplefmt.h"
126 
127 #include "libswresample/version.h"
128 
129 /**
130  * @name Option constants
131  * These constants are used for the @ref avoptions interface for lswr.
132  * @{
133  *
134  */
135 
136 #define SWR_FLAG_RESAMPLE 1 ///< Force resampling even if equal sample rate
137 //TODO use int resample ?
138 //long term TODO can we enable this dynamically?
139 
140 /** Dithering algorithms */
141 enum SwrDitherType {
142     SWR_DITHER_NONE = 0,
143     SWR_DITHER_RECTANGULAR,
144     SWR_DITHER_TRIANGULAR,
145     SWR_DITHER_TRIANGULAR_HIGHPASS,
146 
147     SWR_DITHER_NS = 64,         ///< not part of API/ABI
148     SWR_DITHER_NS_LIPSHITZ,
149     SWR_DITHER_NS_F_WEIGHTED,
150     SWR_DITHER_NS_MODIFIED_E_WEIGHTED,
151     SWR_DITHER_NS_IMPROVED_E_WEIGHTED,
152     SWR_DITHER_NS_SHIBATA,
153     SWR_DITHER_NS_LOW_SHIBATA,
154     SWR_DITHER_NS_HIGH_SHIBATA,
155     SWR_DITHER_NB,              ///< not part of API/ABI
156 };
157 
158 /** Resampling Engines */
159 enum SwrEngine {
160     SWR_ENGINE_SWR,             /**< SW Resampler */
161     SWR_ENGINE_SOXR,            /**< SoX Resampler */
162     SWR_ENGINE_NB,              ///< not part of API/ABI
163 };
164 
165 /** Resampling Filter Types */
166 enum SwrFilterType {
167     SWR_FILTER_TYPE_CUBIC,              /**< Cubic */
168     SWR_FILTER_TYPE_BLACKMAN_NUTTALL,   /**< Blackman Nuttall windowed sinc */
169     SWR_FILTER_TYPE_KAISER,             /**< Kaiser windowed sinc */
170 };
171 
172 /**
173  * @}
174  */
175 
176 /**
177  * The libswresample context. Unlike libavcodec and libavformat, this structure
178  * is opaque. This means that if you would like to set options, you must use
179  * the @ref avoptions API and cannot directly set values to members of the
180  * structure.
181  */
182 typedef struct SwrContext SwrContext;
183 
184 /**
185  * Get the AVClass for SwrContext. It can be used in combination with
186  * AV_OPT_SEARCH_FAKE_OBJ for examining options.
187  *
188  * @see av_opt_find().
189  * @return the AVClass of SwrContext
190  */
191 const AVClass *swr_get_class(void);
192 
193 /**
194  * @name SwrContext constructor functions
195  * @{
196  */
197 
198 /**
199  * Allocate SwrContext.
200  *
201  * If you use this function you will need to set the parameters (manually or
202  * with swr_alloc_set_opts()) before calling swr_init().
203  *
204  * @see swr_alloc_set_opts(), swr_init(), swr_free()
205  * @return NULL on error, allocated context otherwise
206  */
207 struct SwrContext *swr_alloc(void);
208 
209 /**
210  * Initialize context after user parameters have been set.
211  * @note The context must be configured using the AVOption API.
212  *
213  * @see av_opt_set_int()
214  * @see av_opt_set_dict()
215  *
216  * @param[in,out]   s Swr context to initialize
217  * @return AVERROR error code in case of failure.
218  */
219 int swr_init(struct SwrContext *s);
220 
221 /**
222  * Check whether an swr context has been initialized or not.
223  *
224  * @param[in]       s Swr context to check
225  * @see swr_init()
226  * @return positive if it has been initialized, 0 if not initialized
227  */
228 int swr_is_initialized(struct SwrContext *s);
229 
230 /**
231  * Allocate SwrContext if needed and set/reset common parameters.
232  *
233  * This function does not require s to be allocated with swr_alloc(). On the
234  * other hand, swr_alloc() can use swr_alloc_set_opts() to set the parameters
235  * on the allocated context.
236  *
237  * @param s               existing Swr context if available, or NULL if not
238  * @param out_ch_layout   output channel layout (AV_CH_LAYOUT_*)
239  * @param out_sample_fmt  output sample format (AV_SAMPLE_FMT_*).
240  * @param out_sample_rate output sample rate (frequency in Hz)
241  * @param in_ch_layout    input channel layout (AV_CH_LAYOUT_*)
242  * @param in_sample_fmt   input sample format (AV_SAMPLE_FMT_*).
243  * @param in_sample_rate  input sample rate (frequency in Hz)
244  * @param log_offset      logging level offset
245  * @param log_ctx         parent logging context, can be NULL
246  *
247  * @see swr_init(), swr_free()
248  * @return NULL on error, allocated context otherwise
249  */
250 struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
251                                       int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate,
252                                       int64_t  in_ch_layout, enum AVSampleFormat  in_sample_fmt, int  in_sample_rate,
253                                       int log_offset, void *log_ctx);
254 
255 /**
256  * @}
257  *
258  * @name SwrContext destructor functions
259  * @{
260  */
261 
262 /**
263  * Free the given SwrContext and set the pointer to NULL.
264  *
265  * @param[in] s a pointer to a pointer to Swr context
266  */
267 void swr_free(struct SwrContext **s);
268 
269 /**
270  * Closes the context so that swr_is_initialized() returns 0.
271  *
272  * The context can be brought back to life by running swr_init(),
273  * swr_init() can also be used without swr_close().
274  * This function is mainly provided for simplifying the usecase
275  * where one tries to support libavresample and libswresample.
276  *
277  * @param[in,out] s Swr context to be closed
278  */
279 void swr_close(struct SwrContext *s);
280 
281 /**
282  * @}
283  *
284  * @name Core conversion functions
285  * @{
286  */
287 
288 /** Convert audio.
289  *
290  * in and in_count can be set to 0 to flush the last few samples out at the
291  * end.
292  *
293  * If more input is provided than output space, then the input will be buffered.
294  * You can avoid this buffering by using swr_get_out_samples() to retrieve an
295  * upper bound on the required number of output samples for the given number of
296  * input samples. Conversion will run directly without copying whenever possible.
297  *
298  * @param s         allocated Swr context, with parameters set
299  * @param out       output buffers, only the first one need be set in case of packed audio
300  * @param out_count amount of space available for output in samples per channel
301  * @param in        input buffers, only the first one need to be set in case of packed audio
302  * @param in_count  number of input samples available in one channel
303  *
304  * @return number of samples output per channel, negative value on error
305  */
306 int swr_convert(struct SwrContext *s, uint8_t **out, int out_count,
307                                 const uint8_t **in , int in_count);
308 
309 /**
310  * Convert the next timestamp from input to output
311  * timestamps are in 1/(in_sample_rate * out_sample_rate) units.
312  *
313  * @note There are 2 slightly differently behaving modes.
314  *       @li When automatic timestamp compensation is not used, (min_compensation >= FLT_MAX)
315  *              in this case timestamps will be passed through with delays compensated
316  *       @li When automatic timestamp compensation is used, (min_compensation < FLT_MAX)
317  *              in this case the output timestamps will match output sample numbers.
318  *              See ffmpeg-resampler(1) for the two modes of compensation.
319  *
320  * @param s[in]     initialized Swr context
321  * @param pts[in]   timestamp for the next input sample, INT64_MIN if unknown
322  * @see swr_set_compensation(), swr_drop_output(), and swr_inject_silence() are
323  *      function used internally for timestamp compensation.
324  * @return the output timestamp for the next output sample
325  */
326 int64_t swr_next_pts(struct SwrContext *s, int64_t pts);
327 
328 /**
329  * @}
330  *
331  * @name Low-level option setting functions
332  * These functons provide a means to set low-level options that is not possible
333  * with the AVOption API.
334  * @{
335  */
336 
337 /**
338  * Activate resampling compensation ("soft" compensation). This function is
339  * internally called when needed in swr_next_pts().
340  *
341  * @param[in,out] s             allocated Swr context. If it is not initialized,
342  *                              or SWR_FLAG_RESAMPLE is not set, swr_init() is
343  *                              called with the flag set.
344  * @param[in]     sample_delta  delta in PTS per sample
345  * @param[in]     compensation_distance number of samples to compensate for
346  * @return    >= 0 on success, AVERROR error codes if:
347  *            @li @c s is NULL,
348  *            @li @c compensation_distance is less than 0,
349  *            @li @c compensation_distance is 0 but sample_delta is not,
350  *            @li compensation unsupported by resampler, or
351  *            @li swr_init() fails when called.
352  */
353 int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance);
354 
355 /**
356  * Set a customized input channel mapping.
357  *
358  * @param[in,out] s           allocated Swr context, not yet initialized
359  * @param[in]     channel_map customized input channel mapping (array of channel
360  *                            indexes, -1 for a muted channel)
361  * @return >= 0 on success, or AVERROR error code in case of failure.
362  */
363 int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map);
364 
365 /**
366  * Generate a channel mixing matrix.
367  *
368  * This function is the one used internally by libswresample for building the
369  * default mixing matrix. It is made public just as a utility function for
370  * building custom matrices.
371  *
372  * @param in_layout           input channel layout
373  * @param out_layout          output channel layout
374  * @param center_mix_level    mix level for the center channel
375  * @param surround_mix_level  mix level for the surround channel(s)
376  * @param lfe_mix_level       mix level for the low-frequency effects channel
377  * @param rematrix_maxval     if 1.0, coefficients will be normalized to prevent
378  *                            overflow. if INT_MAX, coefficients will not be
379  *                            normalized.
380  * @param[out] matrix         mixing coefficients; matrix[i + stride * o] is
381  *                            the weight of input channel i in output channel o.
382  * @param stride              distance between adjacent input channels in the
383  *                            matrix array
384  * @param matrix_encoding     matrixed stereo downmix mode (e.g. dplii)
385  * @param log_ctx             parent logging context, can be NULL
386  * @return                    0 on success, negative AVERROR code on failure
387  */
388 int swr_build_matrix(uint64_t in_layout, uint64_t out_layout,
389                      double center_mix_level, double surround_mix_level,
390                      double lfe_mix_level, double rematrix_maxval,
391                      double rematrix_volume, double *matrix,
392                      int stride, enum AVMatrixEncoding matrix_encoding,
393                      void *log_ctx);
394 
395 /**
396  * Set a customized remix matrix.
397  *
398  * @param s       allocated Swr context, not yet initialized
399  * @param matrix  remix coefficients; matrix[i + stride * o] is
400  *                the weight of input channel i in output channel o
401  * @param stride  offset between lines of the matrix
402  * @return  >= 0 on success, or AVERROR error code in case of failure.
403  */
404 int swr_set_matrix(struct SwrContext *s, const double *matrix, int stride);
405 
406 /**
407  * @}
408  *
409  * @name Sample handling functions
410  * @{
411  */
412 
413 /**
414  * Drops the specified number of output samples.
415  *
416  * This function, along with swr_inject_silence(), is called by swr_next_pts()
417  * if needed for "hard" compensation.
418  *
419  * @param s     allocated Swr context
420  * @param count number of samples to be dropped
421  *
422  * @return >= 0 on success, or a negative AVERROR code on failure
423  */
424 int swr_drop_output(struct SwrContext *s, int count);
425 
426 /**
427  * Injects the specified number of silence samples.
428  *
429  * This function, along with swr_drop_output(), is called by swr_next_pts()
430  * if needed for "hard" compensation.
431  *
432  * @param s     allocated Swr context
433  * @param count number of samples to be dropped
434  *
435  * @return >= 0 on success, or a negative AVERROR code on failure
436  */
437 int swr_inject_silence(struct SwrContext *s, int count);
438 
439 /**
440  * Gets the delay the next input sample will experience relative to the next output sample.
441  *
442  * Swresample can buffer data if more input has been provided than available
443  * output space, also converting between sample rates needs a delay.
444  * This function returns the sum of all such delays.
445  * The exact delay is not necessarily an integer value in either input or
446  * output sample rate. Especially when downsampling by a large value, the
447  * output sample rate may be a poor choice to represent the delay, similarly
448  * for upsampling and the input sample rate.
449  *
450  * @param s     swr context
451  * @param base  timebase in which the returned delay will be:
452  *              @li if it's set to 1 the returned delay is in seconds
453  *              @li if it's set to 1000 the returned delay is in milliseconds
454  *              @li if it's set to the input sample rate then the returned
455  *                  delay is in input samples
456  *              @li if it's set to the output sample rate then the returned
457  *                  delay is in output samples
458  *              @li if it's the least common multiple of in_sample_rate and
459  *                  out_sample_rate then an exact rounding-free delay will be
460  *                  returned
461  * @returns     the delay in 1 / @c base units.
462  */
463 int64_t swr_get_delay(struct SwrContext *s, int64_t base);
464 
465 /**
466  * Find an upper bound on the number of samples that the next swr_convert
467  * call will output, if called with in_samples of input samples. This
468  * depends on the internal state, and anything changing the internal state
469  * (like further swr_convert() calls) will may change the number of samples
470  * swr_get_out_samples() returns for the same number of input samples.
471  *
472  * @param in_samples    number of input samples.
473  * @note any call to swr_inject_silence(), swr_convert(), swr_next_pts()
474  *       or swr_set_compensation() invalidates this limit
475  * @note it is recommended to pass the correct available buffer size
476  *       to all functions like swr_convert() even if swr_get_out_samples()
477  *       indicates that less would be used.
478  * @returns an upper bound on the number of samples that the next swr_convert
479  *          will output or a negative value to indicate an error
480  */
481 int swr_get_out_samples(struct SwrContext *s, int in_samples);
482 
483 /**
484  * @}
485  *
486  * @name Configuration accessors
487  * @{
488  */
489 
490 /**
491  * Return the @ref LIBSWRESAMPLE_VERSION_INT constant.
492  *
493  * This is useful to check if the build-time libswresample has the same version
494  * as the run-time one.
495  *
496  * @returns     the unsigned int-typed version
497  */
498 unsigned swresample_version(void);
499 
500 /**
501  * Return the swr build-time configuration.
502  *
503  * @returns     the build-time @c ./configure flags
504  */
505 const char *swresample_configuration(void);
506 
507 /**
508  * Return the swr license.
509  *
510  * @returns     the license of libswresample, determined at build-time
511  */
512 const char *swresample_license(void);
513 
514 /**
515  * @}
516  *
517  * @name AVFrame based API
518  * @{
519  */
520 
521 /**
522  * Convert the samples in the input AVFrame and write them to the output AVFrame.
523  *
524  * Input and output AVFrames must have channel_layout, sample_rate and format set.
525  *
526  * If the output AVFrame does not have the data pointers allocated the nb_samples
527  * field will be set using av_frame_get_buffer()
528  * is called to allocate the frame.
529  *
530  * The output AVFrame can be NULL or have fewer allocated samples than required.
531  * In this case, any remaining samples not written to the output will be added
532  * to an internal FIFO buffer, to be returned at the next call to this function
533  * or to swr_convert().
534  *
535  * If converting sample rate, there may be data remaining in the internal
536  * resampling delay buffer. swr_get_delay() tells the number of
537  * remaining samples. To get this data as output, call this function or
538  * swr_convert() with NULL input.
539  *
540  * If the SwrContext configuration does not match the output and
541  * input AVFrame settings the conversion does not take place and depending on
542  * which AVFrame is not matching AVERROR_OUTPUT_CHANGED, AVERROR_INPUT_CHANGED
543  * or the result of a bitwise-OR of them is returned.
544  *
545  * @see swr_delay()
546  * @see swr_convert()
547  * @see swr_get_delay()
548  *
549  * @param swr             audio resample context
550  * @param output          output AVFrame
551  * @param input           input AVFrame
552  * @return                0 on success, AVERROR on failure or nonmatching
553  *                        configuration.
554  */
555 int swr_convert_frame(SwrContext *swr,
556                       AVFrame *output, const AVFrame *input);
557 
558 /**
559  * Configure or reconfigure the SwrContext using the information
560  * provided by the AVFrames.
561  *
562  * The original resampling context is reset even on failure.
563  * The function calls swr_close() internally if the context is open.
564  *
565  * @see swr_close();
566  *
567  * @param swr             audio resample context
568  * @param output          output AVFrame
569  * @param input           input AVFrame
570  * @return                0 on success, AVERROR on failure.
571  */
572 int swr_config_frame(SwrContext *swr, const AVFrame *out, const AVFrame *in);
573 
574 /**
575  * @}
576  * @}
577  */
578 
579 #endif /* SWRESAMPLE_SWRESAMPLE_H */
580