FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
Data Structures | Macros | Enumerations | Functions | Variables
af_atempo.c File Reference

tempo scaling audio filter – an implementation of WSOLA algorithm More...

#include <float.h>
#include "libavcodec/avfft.h"
#include "libavutil/avassert.h"
#include "libavutil/avstring.h"
#include "libavutil/channel_layout.h"
#include "libavutil/eval.h"
#include "libavutil/opt.h"
#include "libavutil/samplefmt.h"
#include "avfilter.h"
#include "audio.h"
#include "internal.h"

Go to the source code of this file.

Data Structures

struct  AudioFragment
 A fragment of audio waveform. More...
 
struct  ATempoContext
 Filter state machine. More...
 

Macros

#define OFFSET(x)   offsetof(ATempoContext, x)
 
#define RE_MALLOC_OR_FAIL(field, field_size)
 
#define yae_init_xdat(scalar_type, scalar_max)
 A helper macro for initializing complex data buffer with scalar data of a given type. More...
 
#define yae_blend(scalar_type)
 A helper macro for blending the overlap region of previous and current audio fragment. More...
 

Enumerations

enum  FilterState {
  YAE_LOAD_FRAGMENT, YAE_ADJUST_POSITION, YAE_RELOAD_FRAGMENT, YAE_OUTPUT_OVERLAP_ADD,
  YAE_FLUSH_OUTPUT
}
 Filter state machine states. More...
 

Functions

 AVFILTER_DEFINE_CLASS (atempo)
 
static AudioFragmentyae_curr_frag (ATempoContext *atempo)
 
static AudioFragmentyae_prev_frag (ATempoContext *atempo)
 
static void yae_clear (ATempoContext *atempo)
 Reset filter to initial state, do not deallocate existing local buffers. More...
 
static void yae_release_buffers (ATempoContext *atempo)
 Reset filter to initial state and deallocate all buffers. More...
 
static int yae_reset (ATempoContext *atempo, enum AVSampleFormat format, int sample_rate, int channels)
 Prepare filter for processing audio data of given format, sample rate and number of channels. More...
 
static int yae_set_tempo (AVFilterContext *ctx, const char *arg_tempo)
 
static void yae_downmix (ATempoContext *atempo, AudioFragment *frag)
 Initialize complex data buffer of a given audio fragment with down-mixed mono data of appropriate scalar type. More...
 
static int yae_load_data (ATempoContext *atempo, const uint8_t **src_ref, const uint8_t *src_end, int64_t stop_here)
 Populate the internal data buffer on as-needed basis. More...
 
static int yae_load_frag (ATempoContext *atempo, const uint8_t **src_ref, const uint8_t *src_end)
 Populate current audio fragment data buffer. More...
 
static void yae_advance_to_next_frag (ATempoContext *atempo)
 Prepare for loading next audio fragment. More...
 
static void yae_xcorr_via_rdft (FFTSample *xcorr, RDFTContext *complex_to_real, const FFTComplex *xa, const FFTComplex *xb, const int window)
 Calculate cross-correlation via rDFT. More...
 
static int yae_align (AudioFragment *frag, const AudioFragment *prev, const int window, const int delta_max, const int drift, FFTSample *correlation, RDFTContext *complex_to_real)
 Calculate alignment offset for given fragment relative to the previous fragment. More...
 
static int yae_adjust_position (ATempoContext *atempo)
 Adjust current fragment position for better alignment with previous fragment. More...
 
static int yae_overlap_add (ATempoContext *atempo, uint8_t **dst_ref, uint8_t *dst_end)
 Blend the overlap region of previous and current audio fragment and output the results to the given destination buffer. More...
 
static void yae_apply (ATempoContext *atempo, const uint8_t **src_ref, const uint8_t *src_end, uint8_t **dst_ref, uint8_t *dst_end)
 Feed as much data to the filter as it is able to consume and receive as much processed data in the destination buffer as it is able to produce or store. More...
 
static int yae_flush (ATempoContext *atempo, uint8_t **dst_ref, uint8_t *dst_end)
 Flush any buffered data from the filter. More...
 
static av_cold int init (AVFilterContext *ctx)
 
static av_cold void uninit (AVFilterContext *ctx)
 
static int query_formats (AVFilterContext *ctx)
 
static int config_props (AVFilterLink *inlink)
 
static int push_samples (ATempoContext *atempo, AVFilterLink *outlink, int n_out)
 
static int filter_frame (AVFilterLink *inlink, AVFrame *src_buffer)
 
static int request_frame (AVFilterLink *outlink)
 
static int process_command (AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
 

Variables

static const AVOption atempo_options []
 
static const AVFilterPad atempo_inputs []
 
static const AVFilterPad atempo_outputs []
 
AVFilter ff_af_atempo
 

Detailed Description

tempo scaling audio filter – an implementation of WSOLA algorithm

Based on MIT licensed yaeAudioTempoFilter.h and yaeAudioFragment.h from Apprentice Video player by Pavel Koshevoy. https://sourceforge.net/projects/apprenticevideo/

An explanation of SOLA algorithm is available at http://www.surina.net/article/time-and-pitch-scaling.html

WSOLA is very similar to SOLA, only one major difference exists between these algorithms. SOLA shifts audio fragments along the output stream, where as WSOLA shifts audio fragments along the input stream.

The advantage of WSOLA algorithm is that the overlap region size is always the same, therefore the blending function is constant and can be precomputed.

Definition in file af_atempo.c.

Macro Definition Documentation

#define OFFSET (   x)    offsetof(ATempoContext, x)

Definition at line 152 of file af_atempo.c.

#define RE_MALLOC_OR_FAIL (   field,
  field_size 
)
Value:
do { \
av_freep(&field); \
field = av_malloc(field_size); \
if (!field) { \
return AVERROR(ENOMEM); \
} \
} while (0)
if()
Definition: avfilter.c:975
#define av_malloc(s)
#define AVERROR(e)
Definition: error.h:43
return
static void yae_release_buffers(ATempoContext *atempo)
Reset filter to initial state and deallocate all buffers.
Definition: af_atempo.c:216
#define av_freep(p)

Definition at line 238 of file af_atempo.c.

Referenced by yae_reset().

#define yae_init_xdat (   scalar_type,
  scalar_max 
)

A helper macro for initializing complex data buffer with scalar data of a given type.

Definition at line 351 of file af_atempo.c.

Referenced by yae_downmix().

#define yae_blend (   scalar_type)
Value:
do { \
const scalar_type *aaa = (const scalar_type *)a; \
const scalar_type *bbb = (const scalar_type *)b; \
\
scalar_type *out = (scalar_type *)dst; \
scalar_type *out_end = (scalar_type *)dst_end; \
int64_t i; \
for (i = 0; i < overlap && out < out_end; \
i++, atempo->position[1]++, wa++, wb++) { \
float w0 = *wa; \
float w1 = *wb; \
int j; \
for (j = 0; j < atempo->channels; \
j++, aaa++, bbb++, out++) { \
float t0 = (float)*aaa; \
float t1 = (float)*bbb; \
\
*out = \
frag->position[0] + i < 0 ? \
*aaa : \
(scalar_type)(t0 * w0 + t1 * w1); \
} \
} \
dst = (uint8_t *)out; \
} while (0)
const char * b
Definition: vf_curves.c:109
uint8_t
#define t0
Definition: regdef.h:28
#define t1
Definition: regdef.h:29
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> out
for(j=16;j >0;--j)

A helper macro for blending the overlap region of previous and current audio fragment.

Definition at line 732 of file af_atempo.c.

Referenced by yae_overlap_add().

Enumeration Type Documentation

Filter state machine states.

Enumerator
YAE_LOAD_FRAGMENT 
YAE_ADJUST_POSITION 
YAE_RELOAD_FRAGMENT 
YAE_OUTPUT_OVERLAP_ADD 
YAE_FLUSH_OUTPUT 

Definition at line 76 of file af_atempo.c.

Function Documentation

AVFILTER_DEFINE_CLASS ( atempo  )
static AudioFragment* yae_curr_frag ( ATempoContext atempo)
inlinestatic
static AudioFragment* yae_prev_frag ( ATempoContext atempo)
inlinestatic
static void yae_clear ( ATempoContext atempo)
static

Reset filter to initial state, do not deallocate existing local buffers.

Definition at line 176 of file af_atempo.c.

Referenced by yae_release_buffers(), and yae_reset().

static void yae_release_buffers ( ATempoContext atempo)
static

Reset filter to initial state and deallocate all buffers.

Definition at line 216 of file af_atempo.c.

Referenced by uninit(), and yae_reset().

static int yae_reset ( ATempoContext atempo,
enum AVSampleFormat  format,
int  sample_rate,
int  channels 
)
static

Prepare filter for processing audio data of given format, sample rate and number of channels.

Definition at line 252 of file af_atempo.c.

Referenced by config_props().

static int yae_set_tempo ( AVFilterContext ctx,
const char *  arg_tempo 
)
static

Definition at line 322 of file af_atempo.c.

Referenced by process_command().

static void yae_downmix ( ATempoContext atempo,
AudioFragment frag 
)
static

Initialize complex data buffer of a given audio fragment with down-mixed mono data of appropriate scalar type.

Definition at line 401 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static int yae_load_data ( ATempoContext atempo,
const uint8_t **  src_ref,
const uint8_t src_end,
int64_t  stop_here 
)
static

Populate the internal data buffer on as-needed basis.

Returns
0 if requested data was already available or was successfully loaded, AVERROR(EAGAIN) if more input data is required.

Definition at line 429 of file af_atempo.c.

Referenced by yae_load_frag().

static int yae_load_frag ( ATempoContext atempo,
const uint8_t **  src_ref,
const uint8_t src_end 
)
static

Populate current audio fragment data buffer.

Returns
0 when the fragment is ready, AVERROR(EAGAIN) if more input data is required.

Definition at line 504 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static void yae_advance_to_next_frag ( ATempoContext atempo)
static

Prepare for loading next audio fragment.

Definition at line 584 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static void yae_xcorr_via_rdft ( FFTSample xcorr,
RDFTContext complex_to_real,
const FFTComplex xa,
const FFTComplex xb,
const int  window 
)
static

Calculate cross-correlation via rDFT.

Multiply two vectors of complex numbers (result of real_to_complex rDFT) and transform back via complex_to_real rDFT.

Definition at line 606 of file af_atempo.c.

Referenced by yae_align().

static int yae_align ( AudioFragment frag,
const AudioFragment prev,
const int  window,
const int  delta_max,
const int  drift,
FFTSample correlation,
RDFTContext complex_to_real 
)
static

Calculate alignment offset for given fragment relative to the previous fragment.

Returns
alignment offset of current fragment relative to previous.

Definition at line 640 of file af_atempo.c.

Referenced by yae_adjust_position().

static int yae_adjust_position ( ATempoContext atempo)
static

Adjust current fragment position for better alignment with previous fragment.

Returns
alignment correction.

Definition at line 694 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static int yae_overlap_add ( ATempoContext atempo,
uint8_t **  dst_ref,
uint8_t dst_end 
)
static

Blend the overlap region of previous and current audio fragment and output the results to the given destination buffer.

Returns
0 if the overlap region was completely stored in the dst buffer, AVERROR(EAGAIN) if more destination buffer space is required.

Definition at line 769 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static void yae_apply ( ATempoContext atempo,
const uint8_t **  src_ref,
const uint8_t src_end,
uint8_t **  dst_ref,
uint8_t dst_end 
)
static

Feed as much data to the filter as it is able to consume and receive as much processed data in the destination buffer as it is able to produce or store.

Definition at line 824 of file af_atempo.c.

Referenced by filter_frame().

static int yae_flush ( ATempoContext atempo,
uint8_t **  dst_ref,
uint8_t dst_end 
)
static

Flush any buffered data from the filter.

Returns
0 if all data was completely stored in the dst buffer, AVERROR(EAGAIN) if more destination buffer space is required.

Definition at line 898 of file af_atempo.c.

Referenced by request_frame().

static av_cold int init ( AVFilterContext ctx)
static

Definition at line 982 of file af_atempo.c.

static av_cold void uninit ( AVFilterContext ctx)
static

Definition at line 990 of file af_atempo.c.

static int query_formats ( AVFilterContext ctx)
static

Definition at line 996 of file af_atempo.c.

static int config_props ( AVFilterLink inlink)
static

Definition at line 1040 of file af_atempo.c.

static int push_samples ( ATempoContext atempo,
AVFilterLink outlink,
int  n_out 
)
static

Definition at line 1054 of file af_atempo.c.

Referenced by filter_frame(), and request_frame().

static int filter_frame ( AVFilterLink inlink,
AVFrame src_buffer 
)
static

Definition at line 1080 of file af_atempo.c.

static int request_frame ( AVFilterLink outlink)
static

Definition at line 1121 of file af_atempo.c.

static int process_command ( AVFilterContext ctx,
const char *  cmd,
const char *  arg,
char *  res,
int  res_len,
int  flags 
)
static

Definition at line 1165 of file af_atempo.c.

Variable Documentation

const AVOption atempo_options[]
static
Initial value:
= {
{ "tempo", "set tempo scale factor",
OFFSET(tempo), AV_OPT_TYPE_DOUBLE, { .dbl = 1.0 }, 0.5, 2.0,
{ NULL }
}
#define NULL
Definition: coverity.c:32
#define AV_OPT_FLAG_AUDIO_PARAM
Definition: opt.h:290
#define AV_OPT_FLAG_FILTERING_PARAM
a generic parameter which can be set by the user for filtering
Definition: opt.h:302
#define OFFSET(x)
Definition: af_atempo.c:152

Definition at line 154 of file af_atempo.c.

const AVFilterPad atempo_inputs[]
static
Initial value:
= {
{
.name = "default",
.filter_frame = filter_frame,
.config_props = config_props,
},
{ NULL }
}
#define NULL
Definition: coverity.c:32
static int config_props(AVFilterLink *inlink)
Definition: af_atempo.c:1040
static int filter_frame(AVFilterLink *inlink, AVFrame *src_buffer)
Definition: af_atempo.c:1080

Definition at line 1175 of file af_atempo.c.

const AVFilterPad atempo_outputs[]
static
Initial value:
= {
{
.name = "default",
.request_frame = request_frame,
},
{ NULL }
}
#define NULL
Definition: coverity.c:32
static int request_frame(AVFilterLink *outlink)
Definition: af_atempo.c:1121

Definition at line 1185 of file af_atempo.c.

AVFilter ff_af_atempo
Initial value:
= {
.name = "atempo",
.description = NULL_IF_CONFIG_SMALL("Adjust audio tempo."),
.init = init,
.uninit = uninit,
.query_formats = query_formats,
.process_command = process_command,
.priv_size = sizeof(ATempoContext),
.priv_class = &atempo_class,
}
static const AVFilterPad outputs[]
Definition: af_ashowinfo.c:248
static const AVFilterPad atempo_outputs[]
Definition: af_atempo.c:1185
Filter state machine.
Definition: af_atempo.c:87
static av_cold int init(AVFilterContext *ctx)
Definition: af_atempo.c:982
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:175
static const AVFilterPad atempo_inputs[]
Definition: af_atempo.c:1175
static int process_command(AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
Definition: af_atempo.c:1165
static int query_formats(AVFilterContext *ctx)
Definition: af_atempo.c:996
static const AVFilterPad inputs[]
Definition: af_ashowinfo.c:239
static av_cold void uninit(AVFilterContext *ctx)
Definition: af_atempo.c:990

Definition at line 1194 of file af_atempo.c.