falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
81350 Commits
32 Branches
449MB
Tree: fc3eb173f4
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'fc3eb173f4'
${ noResults }
FFmpeg/libavfilter/avf_showspectrum.c

1278 lines
54KB
Raw Blame History 

  1. /*

  2.  * Copyright (c) 2012-2013 Clément Bœsch

  3.  * Copyright (c) 2013 Rudolf Polzer <divverent@xonotic.org>

  4.  * Copyright (c) 2015 Paul B Mahol

  5.  *

  6.  * This file is part of FFmpeg.

  7.  *

  8.  * FFmpeg is free software; you can redistribute it and/or

  9.  * modify it under the terms of the GNU Lesser General Public

  10.  * License as published by the Free Software Foundation; either

  11.  * version 2.1 of the License, or (at your option) any later version.

  12.  *

  13.  * FFmpeg is distributed in the hope that it will be useful,

  14.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  15.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  16.  * Lesser General Public License for more details.

  17.  *

  18.  * You should have received a copy of the GNU Lesser General Public

  19.  * License along with FFmpeg; if not, write to the Free Software

  20.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  21.  */

  22. 

  23. /**

  24.  * @file

  25.  * audio to spectrum (video) transmedia filter, based on ffplay rdft showmode

  26.  * (by Michael Niedermayer) and lavfi/avf_showwaves (by Stefano Sabatini).

  27.  */

  28. 

  29. #include <math.h>

  30. 

  31. #include "libavcodec/avfft.h"

  32. #include "libavutil/audio_fifo.h"

  33. #include "libavutil/avassert.h"

  34. #include "libavutil/avstring.h"

  35. #include "libavutil/channel_layout.h"

  36. #include "libavutil/opt.h"

  37. #include "libavutil/xga_font_data.h"

  38. #include "audio.h"

  39. #include "video.h"

  40. #include "avfilter.h"

  41. #include "internal.h"

  42. #include "window_func.h"

  43. 

  44. enum DisplayMode  { COMBINED, SEPARATE, NB_MODES };

  45. enum DataMode     { D_MAGNITUDE, D_PHASE, NB_DMODES };

  46. enum DisplayScale { LINEAR, SQRT, CBRT, LOG, FOURTHRT, FIFTHRT, NB_SCALES };

  47. enum ColorMode    { CHANNEL, INTENSITY, RAINBOW, MORELAND, NEBULAE, FIRE, FIERY, FRUIT, COOL, NB_CLMODES };

  48. enum SlideMode    { REPLACE, SCROLL, FULLFRAME, RSCROLL, NB_SLIDES };

  49. enum Orientation  { VERTICAL, HORIZONTAL, NB_ORIENTATIONS };

  50. 

  51. typedef struct {

  52.     const AVClass *class;

  53.     int w, h;

  54.     AVFrame *outpicref;

  55.     int nb_display_channels;

  56.     int orientation;

  57.     int channel_width;

  58.     int channel_height;

  59.     int sliding;                ///< 1 if sliding mode, 0 otherwise

  60.     int mode;                   ///< channel display mode

  61.     int color_mode;             ///< display color scheme

  62.     int scale;

  63.     float saturation;           ///< color saturation multiplier

  64.     float rotation;             ///< color rotation

  65.     int data;

  66.     int xpos;                   ///< x position (current column)

  67.     FFTContext **fft;           ///< Fast Fourier Transform context

  68.     int fft_bits;               ///< number of bits (FFT window size = 1<<fft_bits)

  69.     FFTComplex **fft_data;      ///< bins holder for each (displayed) channels

  70.     float *window_func_lut;     ///< Window function LUT

  71.     float **magnitudes;

  72.     float **phases;

  73.     int win_func;

  74.     int win_size;

  75.     double win_scale;

  76.     float overlap;

  77.     float gain;

  78.     int hop_size;

  79.     float *combine_buffer;      ///< color combining buffer (3 * h items)

  80.     AVAudioFifo *fifo;

  81.     int64_t pts;

  82.     int single_pic;

  83.     int legend;

  84.     int start_x, start_y;

  85. } ShowSpectrumContext;

  86. 

  87. #define OFFSET(x) offsetof(ShowSpectrumContext, x)

  88. #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM

  89. 

  90. static const AVOption showspectrum_options[] = {

  91.     { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },

  92.     { "s",    "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },

  93.     { "slide", "set sliding mode", OFFSET(sliding), AV_OPT_TYPE_INT, {.i64 = 0}, 0, NB_SLIDES-1, FLAGS, "slide" },

  94.         { "replace", "replace old columns with new", 0, AV_OPT_TYPE_CONST, {.i64=REPLACE}, 0, 0, FLAGS, "slide" },

  95.         { "scroll", "scroll from right to left", 0, AV_OPT_TYPE_CONST, {.i64=SCROLL}, 0, 0, FLAGS, "slide" },

  96.         { "rscroll", "scroll from left to right", 0, AV_OPT_TYPE_CONST, {.i64=RSCROLL}, 0, 0, FLAGS, "slide" },

  97.         { "fullframe", "return full frames", 0, AV_OPT_TYPE_CONST, {.i64=FULLFRAME}, 0, 0, FLAGS, "slide" },

  98.     { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, COMBINED, NB_MODES-1, FLAGS, "mode" },

  99.         { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },

  100.         { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },

  101.     { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=CHANNEL}, CHANNEL, NB_CLMODES-1, FLAGS, "color" },

  102.         { "channel",   "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL},   0, 0, FLAGS, "color" },

  103.         { "intensity", "intensity based coloring",        0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },

  104.         { "rainbow",   "rainbow based coloring",          0, AV_OPT_TYPE_CONST, {.i64=RAINBOW},   0, 0, FLAGS, "color" },

  105.         { "moreland",  "moreland based coloring",         0, AV_OPT_TYPE_CONST, {.i64=MORELAND},  0, 0, FLAGS, "color" },

  106.         { "nebulae",   "nebulae based coloring",          0, AV_OPT_TYPE_CONST, {.i64=NEBULAE},   0, 0, FLAGS, "color" },

  107.         { "fire",      "fire based coloring",             0, AV_OPT_TYPE_CONST, {.i64=FIRE},      0, 0, FLAGS, "color" },

  108.         { "fiery",     "fiery based coloring",            0, AV_OPT_TYPE_CONST, {.i64=FIERY},     0, 0, FLAGS, "color" },

  109.         { "fruit",     "fruit based coloring",            0, AV_OPT_TYPE_CONST, {.i64=FRUIT},     0, 0, FLAGS, "color" },

  110.         { "cool",      "cool based coloring",             0, AV_OPT_TYPE_CONST, {.i64=COOL},      0, 0, FLAGS, "color" },

  111.     { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=SQRT}, LINEAR, NB_SCALES-1, FLAGS, "scale" },

  112.         { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT},   0, 0, FLAGS, "scale" },

  113.         { "cbrt", "cubic root",  0, AV_OPT_TYPE_CONST, {.i64=CBRT},   0, 0, FLAGS, "scale" },

  114.         { "4thrt","4th root",    0, AV_OPT_TYPE_CONST, {.i64=FOURTHRT}, 0, 0, FLAGS, "scale" },

  115.         { "5thrt","5th root",    0, AV_OPT_TYPE_CONST, {.i64=FIFTHRT},  0, 0, FLAGS, "scale" },

  116.         { "log",  "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG},    0, 0, FLAGS, "scale" },

  117.         { "lin",  "linear",      0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },

  118.     { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },

  119.     { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },

  120.         { "rect",     "Rectangular",      0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT},     0, 0, FLAGS, "win_func" },

  121.         { "bartlett", "Bartlett",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },

  122.         { "hann",     "Hann",             0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING},  0, 0, FLAGS, "win_func" },

  123.         { "hanning",  "Hanning",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING},  0, 0, FLAGS, "win_func" },

  124.         { "hamming",  "Hamming",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING},  0, 0, FLAGS, "win_func" },

  125.         { "blackman", "Blackman",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },

  126.         { "welch",    "Welch",            0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH},    0, 0, FLAGS, "win_func" },

  127.         { "flattop",  "Flat-top",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP},  0, 0, FLAGS, "win_func" },

  128.         { "bharris",  "Blackman-Harris",  0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS},  0, 0, FLAGS, "win_func" },

  129.         { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },

  130.         { "bhann",    "Bartlett-Hann",    0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN},    0, 0, FLAGS, "win_func" },

  131.         { "sine",     "Sine",             0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE},     0, 0, FLAGS, "win_func" },

  132.         { "nuttall",  "Nuttall",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL},  0, 0, FLAGS, "win_func" },

  133.         { "lanczos",  "Lanczos",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS},  0, 0, FLAGS, "win_func" },

  134.         { "gauss",    "Gauss",            0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS},    0, 0, FLAGS, "win_func" },

  135.         { "tukey",    "Tukey",            0, AV_OPT_TYPE_CONST, {.i64=WFUNC_TUKEY},    0, 0, FLAGS, "win_func" },

  136.     { "orientation", "set orientation", OFFSET(orientation), AV_OPT_TYPE_INT, {.i64=VERTICAL}, 0, NB_ORIENTATIONS-1, FLAGS, "orientation" },

  137.         { "vertical",   NULL, 0, AV_OPT_TYPE_CONST, {.i64=VERTICAL},   0, 0, FLAGS, "orientation" },

  138.         { "horizontal", NULL, 0, AV_OPT_TYPE_CONST, {.i64=HORIZONTAL}, 0, 0, FLAGS, "orientation" },

  139.     { "overlap", "set window overlap", OFFSET(overlap), AV_OPT_TYPE_FLOAT, {.dbl = 0}, 0, 1, FLAGS },

  140.     { "gain", "set scale gain", OFFSET(gain), AV_OPT_TYPE_FLOAT, {.dbl = 1}, 0, 128, FLAGS },

  141.     { "data", "set data mode", OFFSET(data), AV_OPT_TYPE_INT, {.i64 = 0}, 0, NB_DMODES-1, FLAGS, "data" },

  142.         { "magnitude", NULL, 0, AV_OPT_TYPE_CONST, {.i64=D_MAGNITUDE}, 0, 0, FLAGS, "data" },

  143.         { "phase",     NULL, 0, AV_OPT_TYPE_CONST, {.i64=D_PHASE},     0, 0, FLAGS, "data" },

  144.     { "rotation", "color rotation", OFFSET(rotation), AV_OPT_TYPE_FLOAT, {.dbl = 0}, -1, 1, FLAGS },

  145.     { NULL }

  146. };

  147. 

  148. AVFILTER_DEFINE_CLASS(showspectrum);

  149. 

  150. static const struct ColorTable {

  151.     float a, y, u, v;

  152. } color_table[][8] = {

  153.     [INTENSITY] = {

  154.     {    0,                  0,                  0,                   0 },

  155.     { 0.13, .03587126228984074,  .1573300977624594, -.02548747583751842 },

  156.     { 0.30, .18572281794568020,  .1772436246393981,  .17475554840414750 },

  157.     { 0.60, .28184980583656130, -.1593064119945782,  .47132074554608920 },

  158.     { 0.73, .65830621175547810, -.3716070802232764,  .24352759331252930 },

  159.     { 0.78, .76318535758242900, -.4307467689263783,  .16866496622310430 },

  160.     { 0.91, .95336363636363640, -.2045454545454546,  .03313636363636363 },

  161.     {    1,                  1,                  0,                   0 }},

  162.     [RAINBOW] = {

  163.     {    0,                  0,                  0,                   0 },

  164.     { 0.13,            44/256.,     (189-128)/256.,      (138-128)/256. },

  165.     { 0.25,            29/256.,     (186-128)/256.,      (119-128)/256. },

  166.     { 0.38,           119/256.,     (194-128)/256.,       (53-128)/256. },

  167.     { 0.60,           111/256.,      (73-128)/256.,       (59-128)/256. },

  168.     { 0.73,           205/256.,      (19-128)/256.,      (149-128)/256. },

  169.     { 0.86,           135/256.,      (83-128)/256.,      (200-128)/256. },

  170.     {    1,            73/256.,      (95-128)/256.,      (225-128)/256. }},

  171.     [MORELAND] = {

  172.     {    0,            44/256.,     (181-128)/256.,      (112-128)/256. },

  173.     { 0.13,           126/256.,     (177-128)/256.,      (106-128)/256. },

  174.     { 0.25,           164/256.,     (163-128)/256.,      (109-128)/256. },

  175.     { 0.38,           200/256.,     (140-128)/256.,      (120-128)/256. },

  176.     { 0.60,           201/256.,     (117-128)/256.,      (141-128)/256. },

  177.     { 0.73,           177/256.,     (103-128)/256.,      (165-128)/256. },

  178.     { 0.86,           136/256.,     (100-128)/256.,      (183-128)/256. },

  179.     {    1,            68/256.,     (117-128)/256.,      (203-128)/256. }},

  180.     [NEBULAE] = {

  181.     {    0,            10/256.,     (134-128)/256.,      (132-128)/256. },

  182.     { 0.23,            21/256.,     (137-128)/256.,      (130-128)/256. },

  183.     { 0.45,            35/256.,     (134-128)/256.,      (134-128)/256. },

  184.     { 0.57,            51/256.,     (130-128)/256.,      (139-128)/256. },

  185.     { 0.67,           104/256.,     (116-128)/256.,      (162-128)/256. },

  186.     { 0.77,           120/256.,     (105-128)/256.,      (188-128)/256. },

  187.     { 0.87,           140/256.,     (105-128)/256.,      (188-128)/256. },

  188.     {    1,                  1,                  0,                   0 }},

  189.     [FIRE] = {

  190.     {    0,                  0,                  0,                   0 },

  191.     { 0.23,            44/256.,     (132-128)/256.,      (127-128)/256. },

  192.     { 0.45,            62/256.,     (116-128)/256.,      (140-128)/256. },

  193.     { 0.57,            75/256.,     (105-128)/256.,      (152-128)/256. },

  194.     { 0.67,            95/256.,      (91-128)/256.,      (166-128)/256. },

  195.     { 0.77,           126/256.,      (74-128)/256.,      (172-128)/256. },

  196.     { 0.87,           164/256.,      (73-128)/256.,      (162-128)/256. },

  197.     {    1,                  1,                  0,                   0 }},

  198.     [FIERY] = {

  199.     {    0,                  0,                  0,                   0 },

  200.     { 0.23,            36/256.,     (116-128)/256.,      (163-128)/256. },

  201.     { 0.45,            52/256.,     (102-128)/256.,      (200-128)/256. },

  202.     { 0.57,           116/256.,      (84-128)/256.,      (196-128)/256. },

  203.     { 0.67,           157/256.,      (67-128)/256.,      (181-128)/256. },

  204.     { 0.77,           193/256.,      (40-128)/256.,      (155-128)/256. },

  205.     { 0.87,           221/256.,     (101-128)/256.,      (134-128)/256. },

  206.     {    1,                  1,                  0,                   0 }},

  207.     [FRUIT] = {

  208.     {    0,                  0,                  0,                   0 },

  209.     { 0.20,            29/256.,     (136-128)/256.,      (119-128)/256. },

  210.     { 0.30,            60/256.,     (119-128)/256.,       (90-128)/256. },

  211.     { 0.40,            85/256.,      (91-128)/256.,       (85-128)/256. },

  212.     { 0.50,           116/256.,      (70-128)/256.,      (105-128)/256. },

  213.     { 0.60,           151/256.,      (50-128)/256.,      (146-128)/256. },

  214.     { 0.70,           191/256.,      (63-128)/256.,      (178-128)/256. },

  215.     {    1,            98/256.,      (80-128)/256.,      (221-128)/256. }},

  216.     [COOL] = {

  217.     {    0,                  0,                  0,                   0 },

  218.     {  .15,                  0,                 .5,                 -.5 },

  219.     {    1,                  1,                -.5,                  .5 }},

  220. };

  221. 

  222. static av_cold void uninit(AVFilterContext *ctx)

  223. {

  224.     ShowSpectrumContext *s = ctx->priv;

  225.     int i;

  226. 

  227.     av_freep(&s->combine_buffer);

  228.     if (s->fft) {

  229.         for (i = 0; i < s->nb_display_channels; i++)

  230.             av_fft_end(s->fft[i]);

  231.     }

  232.     if (s->fft_data) {

  233.         for (i = 0; i < s->nb_display_channels; i++)

  234.             av_freep(&s->fft_data[i]);

  235.     }

  236.     av_freep(&s->fft);

  237.     av_freep(&s->fft_data);

  238.     av_freep(&s->window_func_lut);

  239.     if (s->magnitudes) {

  240.         for (i = 0; i < s->nb_display_channels; i++)

  241.             av_freep(&s->magnitudes[i]);

  242.     }

  243.     av_freep(&s->magnitudes);

  244.     av_frame_free(&s->outpicref);

  245.     av_audio_fifo_free(s->fifo);

  246.     if (s->phases) {

  247.         for (i = 0; i < s->nb_display_channels; i++)

  248.             av_freep(&s->phases[i]);

  249.     }

  250.     av_freep(&s->phases);

  251. }

  252. 

  253. static int query_formats(AVFilterContext *ctx)

  254. {

  255.     AVFilterFormats *formats = NULL;

  256.     AVFilterChannelLayouts *layouts = NULL;

  257.     AVFilterLink *inlink = ctx->inputs[0];

  258.     AVFilterLink *outlink = ctx->outputs[0];

  259.     static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE };

  260.     static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_NONE };

  261.     int ret;

  262. 

  263.     /* set input audio formats */

  264.     formats = ff_make_format_list(sample_fmts);

  265.     if ((ret = ff_formats_ref(formats, &inlink->out_formats)) < 0)

  266.         return ret;

  267. 

  268.     layouts = ff_all_channel_layouts();

  269.     if ((ret = ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts)) < 0)

  270.         return ret;

  271. 

  272.     formats = ff_all_samplerates();

  273.     if ((ret = ff_formats_ref(formats, &inlink->out_samplerates)) < 0)

  274.         return ret;

  275. 

  276.     /* set output video format */

  277.     formats = ff_make_format_list(pix_fmts);

  278.     if ((ret = ff_formats_ref(formats, &outlink->in_formats)) < 0)

  279.         return ret;

  280. 

  281.     return 0;

  282. }

  283. 

  284. static int config_output(AVFilterLink *outlink)

  285. {

  286.     AVFilterContext *ctx = outlink->src;

  287.     AVFilterLink *inlink = ctx->inputs[0];

  288.     ShowSpectrumContext *s = ctx->priv;

  289.     int i, fft_bits, h, w;

  290.     float overlap;

  291. 

  292.     if (!strcmp(ctx->filter->name, "showspectrumpic"))

  293.         s->single_pic = 1;

  294. 

  295.     outlink->w = s->w;

  296.     outlink->h = s->h;

  297. 

  298.     if (s->legend) {

  299.         s->start_x = log10(inlink->sample_rate) * 25;

  300.         s->start_y = 64;

  301.         outlink->w += s->start_x * 2;

  302.         outlink->h += s->start_y * 2;

  303.     }

  304. 

  305.     h = (s->mode == COMBINED || s->orientation == HORIZONTAL) ? s->h : s->h / inlink->channels;

  306.     w = (s->mode == COMBINED || s->orientation == VERTICAL)   ? s->w : s->w / inlink->channels;

  307.     s->channel_height = h;

  308.     s->channel_width  = w;

  309. 

  310.     if (s->orientation == VERTICAL) {

  311.         /* FFT window size (precision) according to the requested output frame height */

  312.         for (fft_bits = 1; 1 << fft_bits < 2 * h; fft_bits++);

  313.     } else {

  314.         /* FFT window size (precision) according to the requested output frame width */

  315.         for (fft_bits = 1; 1 << fft_bits < 2 * w; fft_bits++);

  316.     }

  317.     s->win_size = 1 << fft_bits;

  318. 

  319.     if (!s->fft) {

  320.         s->fft = av_calloc(inlink->channels, sizeof(*s->fft));

  321.         if (!s->fft)

  322.             return AVERROR(ENOMEM);

  323.     }

  324. 

  325.     /* (re-)configuration if the video output changed (or first init) */

  326.     if (fft_bits != s->fft_bits) {

  327.         AVFrame *outpicref;

  328. 

  329.         s->fft_bits = fft_bits;

  330. 

  331.         /* FFT buffers: x2 for each (display) channel buffer.

  332.          * Note: we use free and malloc instead of a realloc-like function to

  333.          * make sure the buffer is aligned in memory for the FFT functions. */

  334.         for (i = 0; i < s->nb_display_channels; i++) {

  335.             av_fft_end(s->fft[i]);

  336.             av_freep(&s->fft_data[i]);

  337.         }

  338.         av_freep(&s->fft_data);

  339. 

  340.         s->nb_display_channels = inlink->channels;

  341.         for (i = 0; i < s->nb_display_channels; i++) {

  342.             s->fft[i] = av_fft_init(fft_bits, 0);

  343.             if (!s->fft[i]) {

  344.                 av_log(ctx, AV_LOG_ERROR, "Unable to create FFT context. "

  345.                        "The window size might be too high.\n");

  346.                 return AVERROR(EINVAL);

  347.             }

  348.         }

  349. 

  350.         s->magnitudes = av_calloc(s->nb_display_channels, sizeof(*s->magnitudes));

  351.         if (!s->magnitudes)

  352.             return AVERROR(ENOMEM);

  353.         for (i = 0; i < s->nb_display_channels; i++) {

  354.             s->magnitudes[i] = av_calloc(s->orientation == VERTICAL ? s->h : s->w, sizeof(**s->magnitudes));

  355.             if (!s->magnitudes[i])

  356.                 return AVERROR(ENOMEM);

  357.         }

  358. 

  359.         s->phases = av_calloc(s->nb_display_channels, sizeof(*s->phases));

  360.         if (!s->phases)

  361.             return AVERROR(ENOMEM);

  362.         for (i = 0; i < s->nb_display_channels; i++) {

  363.             s->phases[i] = av_calloc(s->orientation == VERTICAL ? s->h : s->w, sizeof(**s->phases));

  364.             if (!s->phases[i])

  365.                 return AVERROR(ENOMEM);

  366.         }

  367. 

  368.         s->fft_data = av_calloc(s->nb_display_channels, sizeof(*s->fft_data));

  369.         if (!s->fft_data)

  370.             return AVERROR(ENOMEM);

  371.         for (i = 0; i < s->nb_display_channels; i++) {

  372.             s->fft_data[i] = av_calloc(s->win_size, sizeof(**s->fft_data));

  373.             if (!s->fft_data[i])

  374.                 return AVERROR(ENOMEM);

  375.         }

  376. 

  377.         /* pre-calc windowing function */

  378.         s->window_func_lut =

  379.             av_realloc_f(s->window_func_lut, s->win_size,

  380.                          sizeof(*s->window_func_lut));

  381.         if (!s->window_func_lut)

  382.             return AVERROR(ENOMEM);

  383.         ff_generate_window_func(s->window_func_lut, s->win_size, s->win_func, &overlap);

  384.         if (s->overlap == 1)

  385.             s->overlap = overlap;

  386.         s->hop_size = (1. - s->overlap) * s->win_size;

  387.         if (s->hop_size < 1) {

  388.             av_log(ctx, AV_LOG_ERROR, "overlap %f too big\n", s->overlap);

  389.             return AVERROR(EINVAL);

  390.         }

  391. 

  392.         for (s->win_scale = 0, i = 0; i < s->win_size; i++) {

  393.             s->win_scale += s->window_func_lut[i] * s->window_func_lut[i];

  394.         }

  395.         s->win_scale = 1. / sqrt(s->win_scale);

  396. 

  397.         /* prepare the initial picref buffer (black frame) */

  398.         av_frame_free(&s->outpicref);

  399.         s->outpicref = outpicref =

  400.             ff_get_video_buffer(outlink, outlink->w, outlink->h);

  401.         if (!outpicref)

  402.             return AVERROR(ENOMEM);

  403.         outlink->sample_aspect_ratio = (AVRational){1,1};

  404.         for (i = 0; i < outlink->h; i++) {

  405.             memset(outpicref->data[0] + i * outpicref->linesize[0],   0, outlink->w);

  406.             memset(outpicref->data[1] + i * outpicref->linesize[1], 128, outlink->w);

  407.             memset(outpicref->data[2] + i * outpicref->linesize[2], 128, outlink->w);

  408.         }

  409.         av_frame_set_color_range(outpicref, AVCOL_RANGE_JPEG);

  410.     }

  411. 

  412.     if ((s->orientation == VERTICAL   && s->xpos >= s->w) ||

  413.         (s->orientation == HORIZONTAL && s->xpos >= s->h))

  414.         s->xpos = 0;

  415. 

  416.     outlink->frame_rate = av_make_q(inlink->sample_rate, s->win_size * (1.-s->overlap));

  417.     if (s->orientation == VERTICAL && s->sliding == FULLFRAME)

  418.         outlink->frame_rate.den *= s->w;

  419.     if (s->orientation == HORIZONTAL && s->sliding == FULLFRAME)

  420.         outlink->frame_rate.den *= s->h;

  421. 

  422.     if (s->orientation == VERTICAL) {

  423.         s->combine_buffer =

  424.             av_realloc_f(s->combine_buffer, s->h * 3,

  425.                          sizeof(*s->combine_buffer));

  426.     } else {

  427.         s->combine_buffer =

  428.             av_realloc_f(s->combine_buffer, s->w * 3,

  429.                          sizeof(*s->combine_buffer));

  430.     }

  431. 

  432.     av_log(ctx, AV_LOG_VERBOSE, "s:%dx%d FFT window size:%d\n",

  433.            s->w, s->h, s->win_size);

  434. 

  435.     av_audio_fifo_free(s->fifo);

  436.     s->fifo = av_audio_fifo_alloc(inlink->format, inlink->channels, s->win_size);

  437.     if (!s->fifo)

  438.         return AVERROR(ENOMEM);

  439.     return 0;

  440. }

  441. 

  442. static int run_channel_fft(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)

  443. {

  444.     ShowSpectrumContext *s = ctx->priv;

  445.     const float *window_func_lut = s->window_func_lut;

  446.     AVFrame *fin = arg;

  447.     const int ch = jobnr;

  448.     int n;

  449. 

  450.     /* fill FFT input with the number of samples available */

  451.     const float *p = (float *)fin->extended_data[ch];

  452. 

  453.     for (n = 0; n < s->win_size; n++) {

  454.         s->fft_data[ch][n].re = p[n] * window_func_lut[n];

  455.         s->fft_data[ch][n].im = 0;

  456.     }

  457. 

  458.     /* run FFT on each samples set */

  459.     av_fft_permute(s->fft[ch], s->fft_data[ch]);

  460.     av_fft_calc(s->fft[ch], s->fft_data[ch]);

  461. 

  462.     return 0;

  463. }

  464. 

  465. #define RE(y, ch) s->fft_data[ch][y].re

  466. #define IM(y, ch) s->fft_data[ch][y].im

  467. #define MAGNITUDE(y, ch) hypot(RE(y, ch), IM(y, ch))

  468. #define PHASE(y, ch) atan2(IM(y, ch), RE(y, ch))

  469. 

  470. static int calc_channel_magnitudes(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)

  471. {

  472.     ShowSpectrumContext *s = ctx->priv;

  473.     const double w = s->win_scale * (s->scale == LOG ? s->win_scale : 1);

  474.     int y, h = s->orientation == VERTICAL ? s->h : s->w;

  475.     const float f = s->gain * w;

  476.     const int ch = jobnr;

  477.     float *magnitudes = s->magnitudes[ch];

  478. 

  479.     for (y = 0; y < h; y++)

  480.         magnitudes[y] = MAGNITUDE(y, ch) * f;

  481. 

  482.     return 0;

  483. }

  484. 

  485. static int calc_channel_phases(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)

  486. {

  487.     ShowSpectrumContext *s = ctx->priv;

  488.     const int h = s->orientation == VERTICAL ? s->h : s->w;

  489.     const int ch = jobnr;

  490.     float *phases = s->phases[ch];

  491.     int y;

  492. 

  493.     for (y = 0; y < h; y++)

  494.         phases[y] = (PHASE(y, ch) / M_PI + 1) / 2;

  495. 

  496.     return 0;

  497. }

  498. 

  499. static void acalc_magnitudes(ShowSpectrumContext *s)

  500. {

  501.     const double w = s->win_scale * (s->scale == LOG ? s->win_scale : 1);

  502.     int ch, y, h = s->orientation == VERTICAL ? s->h : s->w;

  503.     const float f = s->gain * w;

  504. 

  505.     for (ch = 0; ch < s->nb_display_channels; ch++) {

  506.         float *magnitudes = s->magnitudes[ch];

  507. 

  508.         for (y = 0; y < h; y++)

  509.             magnitudes[y] += MAGNITUDE(y, ch) * f;

  510.     }

  511. }

  512. 

  513. static void scale_magnitudes(ShowSpectrumContext *s, float scale)

  514. {

  515.     int ch, y, h = s->orientation == VERTICAL ? s->h : s->w;

  516. 

  517.     for (ch = 0; ch < s->nb_display_channels; ch++) {

  518.         float *magnitudes = s->magnitudes[ch];

  519. 

  520.         for (y = 0; y < h; y++)

  521.             magnitudes[y] *= scale;

  522.     }

  523. }

  524. 

  525. static void color_range(ShowSpectrumContext *s, int ch,

  526.                         float *yf, float *uf, float *vf)

  527. {

  528.     switch (s->mode) {

  529.     case COMBINED:

  530.         // reduce range by channel count

  531.         *yf = 256.0f / s->nb_display_channels;

  532.         switch (s->color_mode) {

  533.         case RAINBOW:

  534.         case MORELAND:

  535.         case NEBULAE:

  536.         case FIRE:

  537.         case FIERY:

  538.         case FRUIT:

  539.         case COOL:

  540.         case INTENSITY:

  541.             *uf = *yf;

  542.             *vf = *yf;

  543.             break;

  544.         case CHANNEL:

  545.             /* adjust saturation for mixed UV coloring */

  546.             /* this factor is correct for infinite channels, an approximation otherwise */

  547.             *uf = *yf * M_PI;

  548.             *vf = *yf * M_PI;

  549.             break;

  550.         default:

  551.             av_assert0(0);

  552.         }

  553.         break;

  554.     case SEPARATE:

  555.         // full range

  556.         *yf = 256.0f;

  557.         *uf = 256.0f;

  558.         *vf = 256.0f;

  559.         break;

  560.     default:

  561.         av_assert0(0);

  562.     }

  563. 

  564.     if (s->color_mode == CHANNEL) {

  565.         if (s->nb_display_channels > 1) {

  566.             *uf *= 0.5 * sin((2 * M_PI * ch) / s->nb_display_channels + M_PI * s->rotation);

  567.             *vf *= 0.5 * cos((2 * M_PI * ch) / s->nb_display_channels + M_PI * s->rotation);

  568.         } else {

  569.             *uf *= 0.5 * sin(M_PI * s->rotation);

  570.             *vf *= 0.5 * cos(M_PI * s->rotation + M_PI_2);

  571.         }

  572.     } else {

  573.         *uf += *uf * sin(M_PI * s->rotation);

  574.         *vf += *vf * cos(M_PI * s->rotation + M_PI_2);

  575.     }

  576. 

  577.     *uf *= s->saturation;

  578.     *vf *= s->saturation;

  579. }

  580. 

  581. static void pick_color(ShowSpectrumContext *s,

  582.                        float yf, float uf, float vf,

  583.                        float a, float *out)

  584. {

  585.     if (s->color_mode > CHANNEL) {

  586.         const int cm = s->color_mode;

  587.         float y, u, v;

  588.         int i;

  589. 

  590.         for (i = 1; i < FF_ARRAY_ELEMS(color_table[cm]) - 1; i++)

  591.             if (color_table[cm][i].a >= a)

  592.                 break;

  593.         // i now is the first item >= the color

  594.         // now we know to interpolate between item i - 1 and i

  595.         if (a <= color_table[cm][i - 1].a) {

  596.             y = color_table[cm][i - 1].y;

  597.             u = color_table[cm][i - 1].u;

  598.             v = color_table[cm][i - 1].v;

  599.         } else if (a >= color_table[cm][i].a) {

  600.             y = color_table[cm][i].y;

  601.             u = color_table[cm][i].u;

  602.             v = color_table[cm][i].v;

  603.         } else {

  604.             float start = color_table[cm][i - 1].a;

  605.             float end = color_table[cm][i].a;

  606.             float lerpfrac = (a - start) / (end - start);

  607.             y = color_table[cm][i - 1].y * (1.0f - lerpfrac)

  608.               + color_table[cm][i].y * lerpfrac;

  609.             u = color_table[cm][i - 1].u * (1.0f - lerpfrac)

  610.               + color_table[cm][i].u * lerpfrac;

  611.             v = color_table[cm][i - 1].v * (1.0f - lerpfrac)

  612.               + color_table[cm][i].v * lerpfrac;

  613.         }

  614. 

  615.         out[0] += y * yf;

  616.         out[1] += u * uf;

  617.         out[2] += v * vf;

  618.     } else {

  619.         out[0] += a * yf;

  620.         out[1] += a * uf;

  621.         out[2] += a * vf;

  622.     }

  623. }

  624. 

  625. static void clear_combine_buffer(ShowSpectrumContext *s, int size)

  626. {

  627.     int y;

  628. 

  629.     for (y = 0; y < size; y++) {

  630.         s->combine_buffer[3 * y    ] = 0;

  631.         s->combine_buffer[3 * y + 1] = 127.5;

  632.         s->combine_buffer[3 * y + 2] = 127.5;

  633.     }

  634. }

  635. 

  636. static int plot_channel(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)

  637. {

  638.     ShowSpectrumContext *s = ctx->priv;

  639.     const int h = s->orientation == VERTICAL ? s->channel_height : s->channel_width;

  640.     const int ch = jobnr;

  641.     float *magnitudes = s->magnitudes[ch];

  642.     float *phases = s->phases[ch];

  643.     float yf, uf, vf;

  644.     int y;

  645. 

  646.     /* decide color range */

  647.     color_range(s, ch, &yf, &uf, &vf);

  648. 

  649.     /* draw the channel */

  650.     for (y = 0; y < h; y++) {

  651.         int row = (s->mode == COMBINED) ? y : ch * h + y;

  652.         float *out = &s->combine_buffer[3 * row];

  653.         float a;

  654. 

  655.         switch (s->data) {

  656.         case D_MAGNITUDE:

  657.             /* get magnitude */

  658.             a = magnitudes[y];

  659.             break;

  660.         case D_PHASE:

  661.             /* get phase */

  662.             a = phases[y];

  663.             break;

  664.         default:

  665.             av_assert0(0);

  666.         }

  667. 

  668.         /* apply scale */

  669.         switch (s->scale) {

  670.         case LINEAR:

  671.             a = av_clipf(a, 0, 1);

  672.             break;

  673.         case SQRT:

  674.             a = av_clipf(sqrt(a), 0, 1);

  675.             break;

  676.         case CBRT:

  677.             a = av_clipf(cbrt(a), 0, 1);

  678.             break;

  679.         case FOURTHRT:

  680.             a = av_clipf(sqrt(sqrt(a)), 0, 1);

  681.             break;

  682.         case FIFTHRT:

  683.             a = av_clipf(pow(a, 0.20), 0, 1);

  684.             break;

  685.         case LOG:

  686.             a = 1 + log10(av_clipd(a, 1e-6, 1)) / 6; // zero = -120dBFS

  687.             break;

  688.         default:

  689.             av_assert0(0);

  690.         }

  691. 

  692.         pick_color(s, yf, uf, vf, a, out);

  693.     }

  694. 

  695.     return 0;

  696. }

  697. 

  698. static int plot_spectrum_column(AVFilterLink *inlink, AVFrame *insamples)

  699. {

  700.     AVFilterContext *ctx = inlink->dst;

  701.     AVFilterLink *outlink = ctx->outputs[0];

  702.     ShowSpectrumContext *s = ctx->priv;

  703.     AVFrame *outpicref = s->outpicref;

  704.     int ret, plane, x, y;

  705. 

  706.     /* fill a new spectrum column */

  707.     /* initialize buffer for combining to black */

  708.     clear_combine_buffer(s, s->orientation == VERTICAL ? s->h : s->w);

  709. 

  710.     ctx->internal->execute(ctx, plot_channel, NULL, NULL, s->nb_display_channels);

  711. 

  712.     av_frame_make_writable(s->outpicref);

  713.     /* copy to output */

  714.     if (s->orientation == VERTICAL) {

  715.         if (s->sliding == SCROLL) {

  716.             for (plane = 0; plane < 3; plane++) {

  717.                 for (y = 0; y < s->h; y++) {

  718.                     uint8_t *p = outpicref->data[plane] +

  719.                                  y * outpicref->linesize[plane];

  720.                     memmove(p, p + 1, s->w - 1);

  721.                 }

  722.             }

  723.             s->xpos = s->w - 1;

  724.         } else if (s->sliding == RSCROLL) {

  725.             for (plane = 0; plane < 3; plane++) {

  726.                 for (y = 0; y < s->h; y++) {

  727.                     uint8_t *p = outpicref->data[plane] +

  728.                                  y * outpicref->linesize[plane];

  729.                     memmove(p + 1, p, s->w - 1);

  730.                 }

  731.             }

  732.             s->xpos = 0;

  733.         }

  734.         for (plane = 0; plane < 3; plane++) {

  735.             uint8_t *p = outpicref->data[plane] + s->start_x +

  736.                          (outlink->h - 1 - s->start_y) * outpicref->linesize[plane] +

  737.                          s->xpos;

  738.             for (y = 0; y < s->h; y++) {

  739.                 *p = lrintf(av_clipf(s->combine_buffer[3 * y + plane], 0, 255));

  740.                 p -= outpicref->linesize[plane];

  741.             }

  742.         }

  743.     } else {

  744.         if (s->sliding == SCROLL) {

  745.             for (plane = 0; plane < 3; plane++) {

  746.                 for (y = 1; y < s->h; y++) {

  747.                     memmove(outpicref->data[plane] + (y-1) * outpicref->linesize[plane],

  748.                             outpicref->data[plane] + (y  ) * outpicref->linesize[plane],

  749.                             s->w);

  750.                 }

  751.             }

  752.             s->xpos = s->h - 1;

  753.         } else if (s->sliding == RSCROLL) {

  754.             for (plane = 0; plane < 3; plane++) {

  755.                 for (y = s->h - 1; y >= 1; y--) {

  756.                     memmove(outpicref->data[plane] + (y  ) * outpicref->linesize[plane],

  757.                             outpicref->data[plane] + (y-1) * outpicref->linesize[plane],

  758.                             s->w);

  759.                 }

  760.             }

  761.             s->xpos = 0;

  762.         }

  763.         for (plane = 0; plane < 3; plane++) {

  764.             uint8_t *p = outpicref->data[plane] + s->start_x +

  765.                          (s->xpos + s->start_y) * outpicref->linesize[plane];

  766.             for (x = 0; x < s->w; x++) {

  767.                 *p = lrintf(av_clipf(s->combine_buffer[3 * x + plane], 0, 255));

  768.                 p++;

  769.             }

  770.         }

  771.     }

  772. 

  773.     if (s->sliding != FULLFRAME || s->xpos == 0)

  774.         outpicref->pts = insamples->pts;

  775. 

  776.     s->xpos++;

  777.     if (s->orientation == VERTICAL && s->xpos >= s->w)

  778.         s->xpos = 0;

  779.     if (s->orientation == HORIZONTAL && s->xpos >= s->h)

  780.         s->xpos = 0;

  781.     if (!s->single_pic && (s->sliding != FULLFRAME || s->xpos == 0)) {

  782.         ret = ff_filter_frame(outlink, av_frame_clone(s->outpicref));

  783.         if (ret < 0)

  784.             return ret;

  785.     }

  786. 

  787.     return s->win_size;

  788. }

  789. 

  790. #if CONFIG_SHOWSPECTRUM_FILTER

  791. 

  792. static int request_frame(AVFilterLink *outlink)

  793. {

  794.     ShowSpectrumContext *s = outlink->src->priv;

  795.     AVFilterLink *inlink = outlink->src->inputs[0];

  796.     unsigned i;

  797.     int ret;

  798. 

  799.     ret = ff_request_frame(inlink);

  800.     if (ret == AVERROR_EOF && s->sliding == FULLFRAME && s->xpos > 0 &&

  801.         s->outpicref) {

  802.         if (s->orientation == VERTICAL) {

  803.             for (i = 0; i < outlink->h; i++) {

  804.                 memset(s->outpicref->data[0] + i * s->outpicref->linesize[0] + s->xpos,   0, outlink->w - s->xpos);

  805.                 memset(s->outpicref->data[1] + i * s->outpicref->linesize[1] + s->xpos, 128, outlink->w - s->xpos);

  806.                 memset(s->outpicref->data[2] + i * s->outpicref->linesize[2] + s->xpos, 128, outlink->w - s->xpos);

  807.             }

  808.         } else {

  809.             for (i = s->xpos; i < outlink->h; i++) {

  810.                 memset(s->outpicref->data[0] + i * s->outpicref->linesize[0],   0, outlink->w);

  811.                 memset(s->outpicref->data[1] + i * s->outpicref->linesize[1], 128, outlink->w);

  812.                 memset(s->outpicref->data[2] + i * s->outpicref->linesize[2], 128, outlink->w);

  813.             }

  814.         }

  815.         ret = ff_filter_frame(outlink, s->outpicref);

  816.         s->outpicref = NULL;

  817.     }

  818. 

  819.     return ret;

  820. }

  821. 

  822. static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)

  823. {

  824.     AVFilterContext *ctx = inlink->dst;

  825.     ShowSpectrumContext *s = ctx->priv;

  826.     AVFrame *fin = NULL;

  827.     int ret = 0, consumed = 0;

  828. 

  829.     if (s->pts == AV_NOPTS_VALUE)

  830.         s->pts = insamples->pts - av_audio_fifo_size(s->fifo);

  831. 

  832.     av_audio_fifo_write(s->fifo, (void **)insamples->extended_data, insamples->nb_samples);

  833.     av_frame_free(&insamples);

  834.     while (av_audio_fifo_size(s->fifo) >= s->win_size) {

  835.         fin = ff_get_audio_buffer(inlink, s->win_size);

  836.         if (!fin) {

  837.             ret = AVERROR(ENOMEM);

  838.             goto fail;

  839.         }

  840. 

  841.         fin->pts = s->pts + consumed;

  842.         consumed += s->hop_size;

  843.         ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data, s->win_size);

  844.         if (ret < 0)

  845.             goto fail;

  846. 

  847.         av_assert0(fin->nb_samples == s->win_size);

  848. 

  849.         ctx->internal->execute(ctx, run_channel_fft, fin, NULL, s->nb_display_channels);

  850. 

  851.         if (s->data == D_MAGNITUDE)

  852.             ctx->internal->execute(ctx, calc_channel_magnitudes, NULL, NULL, s->nb_display_channels);

  853. 

  854.         if (s->data == D_PHASE)

  855.             ctx->internal->execute(ctx, calc_channel_phases, NULL, NULL, s->nb_display_channels);

  856. 

  857.         ret = plot_spectrum_column(inlink, fin);

  858.         av_frame_free(&fin);

  859.         av_audio_fifo_drain(s->fifo, s->hop_size);

  860.         if (ret < 0)

  861.             goto fail;

  862.     }

  863. 

  864. fail:

  865.     s->pts = AV_NOPTS_VALUE;

  866.     av_frame_free(&fin);

  867.     return ret;

  868. }

  869. 

  870. static const AVFilterPad showspectrum_inputs[] = {

  871.     {

  872.         .name         = "default",

  873.         .type         = AVMEDIA_TYPE_AUDIO,

  874.         .filter_frame = filter_frame,

  875.     },

  876.     { NULL }

  877. };

  878. 

  879. static const AVFilterPad showspectrum_outputs[] = {

  880.     {

  881.         .name          = "default",

  882.         .type          = AVMEDIA_TYPE_VIDEO,

  883.         .config_props  = config_output,

  884.         .request_frame = request_frame,

  885.     },

  886.     { NULL }

  887. };

  888. 

  889. AVFilter ff_avf_showspectrum = {

  890.     .name          = "showspectrum",

  891.     .description   = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output."),

  892.     .uninit        = uninit,

  893.     .query_formats = query_formats,

  894.     .priv_size     = sizeof(ShowSpectrumContext),

  895.     .inputs        = showspectrum_inputs,

  896.     .outputs       = showspectrum_outputs,

  897.     .priv_class    = &showspectrum_class,

  898.     .flags         = AVFILTER_FLAG_SLICE_THREADS,

  899. };

  900. #endif // CONFIG_SHOWSPECTRUM_FILTER

  901. 

  902. #if CONFIG_SHOWSPECTRUMPIC_FILTER

  903. 

  904. static const AVOption showspectrumpic_options[] = {

  905.     { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "4096x2048"}, 0, 0, FLAGS },

  906.     { "s",    "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "4096x2048"}, 0, 0, FLAGS },

  907.     { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, 0, NB_MODES-1, FLAGS, "mode" },

  908.         { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },

  909.         { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },

  910.     { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=INTENSITY}, 0, NB_CLMODES-1, FLAGS, "color" },

  911.         { "channel",   "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL},   0, 0, FLAGS, "color" },

  912.         { "intensity", "intensity based coloring",        0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },

  913.         { "rainbow",   "rainbow based coloring",          0, AV_OPT_TYPE_CONST, {.i64=RAINBOW},   0, 0, FLAGS, "color" },

  914.         { "moreland",  "moreland based coloring",         0, AV_OPT_TYPE_CONST, {.i64=MORELAND},  0, 0, FLAGS, "color" },

  915.         { "nebulae",   "nebulae based coloring",          0, AV_OPT_TYPE_CONST, {.i64=NEBULAE},   0, 0, FLAGS, "color" },

  916.         { "fire",      "fire based coloring",             0, AV_OPT_TYPE_CONST, {.i64=FIRE},      0, 0, FLAGS, "color" },

  917.         { "fiery",     "fiery based coloring",            0, AV_OPT_TYPE_CONST, {.i64=FIERY},     0, 0, FLAGS, "color" },

  918.         { "fruit",     "fruit based coloring",            0, AV_OPT_TYPE_CONST, {.i64=FRUIT},     0, 0, FLAGS, "color" },

  919.         { "cool",      "cool based coloring",             0, AV_OPT_TYPE_CONST, {.i64=COOL},      0, 0, FLAGS, "color" },

  920.     { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=LOG}, 0, NB_SCALES-1, FLAGS, "scale" },

  921.         { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT},   0, 0, FLAGS, "scale" },

  922.         { "cbrt", "cubic root",  0, AV_OPT_TYPE_CONST, {.i64=CBRT},   0, 0, FLAGS, "scale" },

  923.         { "4thrt","4th root",    0, AV_OPT_TYPE_CONST, {.i64=FOURTHRT}, 0, 0, FLAGS, "scale" },

  924.         { "5thrt","5th root",    0, AV_OPT_TYPE_CONST, {.i64=FIFTHRT},  0, 0, FLAGS, "scale" },

  925.         { "log",  "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG},    0, 0, FLAGS, "scale" },

  926.         { "lin",  "linear",      0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },

  927.     { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },

  928.     { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },

  929.         { "rect",     "Rectangular",      0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT},     0, 0, FLAGS, "win_func" },

  930.         { "bartlett", "Bartlett",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },

  931.         { "hann",     "Hann",             0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING},  0, 0, FLAGS, "win_func" },

  932.         { "hanning",  "Hanning",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING},  0, 0, FLAGS, "win_func" },

  933.         { "hamming",  "Hamming",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING},  0, 0, FLAGS, "win_func" },

  934.         { "blackman", "Blackman",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },

  935.         { "welch",    "Welch",            0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH},    0, 0, FLAGS, "win_func" },

  936.         { "flattop",  "Flat-top",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP},  0, 0, FLAGS, "win_func" },

  937.         { "bharris",  "Blackman-Harris",  0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS},  0, 0, FLAGS, "win_func" },

  938.         { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },

  939.         { "bhann",    "Bartlett-Hann",    0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN},    0, 0, FLAGS, "win_func" },

  940.         { "sine",     "Sine",             0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE},     0, 0, FLAGS, "win_func" },

  941.         { "nuttall",  "Nuttall",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL},  0, 0, FLAGS, "win_func" },

  942.         { "lanczos",  "Lanczos",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS},  0, 0, FLAGS, "win_func" },

  943.         { "gauss",    "Gauss",            0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS},    0, 0, FLAGS, "win_func" },

  944.         { "tukey",    "Tukey",            0, AV_OPT_TYPE_CONST, {.i64=WFUNC_TUKEY},    0, 0, FLAGS, "win_func" },

  945.     { "orientation", "set orientation", OFFSET(orientation), AV_OPT_TYPE_INT, {.i64=VERTICAL}, 0, NB_ORIENTATIONS-1, FLAGS, "orientation" },

  946.         { "vertical",   NULL, 0, AV_OPT_TYPE_CONST, {.i64=VERTICAL},   0, 0, FLAGS, "orientation" },

  947.         { "horizontal", NULL, 0, AV_OPT_TYPE_CONST, {.i64=HORIZONTAL}, 0, 0, FLAGS, "orientation" },

  948.     { "gain", "set scale gain", OFFSET(gain), AV_OPT_TYPE_FLOAT, {.dbl = 1}, 0, 128, FLAGS },

  949.     { "legend", "draw legend", OFFSET(legend), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, FLAGS },

  950.     { "rotation", "color rotation", OFFSET(rotation), AV_OPT_TYPE_FLOAT, {.dbl = 0}, -1, 1, FLAGS },

  951.     { NULL }

  952. };

  953. 

  954. AVFILTER_DEFINE_CLASS(showspectrumpic);

  955. 

  956. static void drawtext(AVFrame *pic, int x, int y, const char *txt, int o)

  957. {

  958.     const uint8_t *font;

  959.     int font_height;

  960.     int i;

  961. 

  962.     font = avpriv_cga_font,   font_height =  8;

  963. 

  964.     for (i = 0; txt[i]; i++) {

  965.         int char_y, mask;

  966. 

  967.         if (o) {

  968.             for (char_y = font_height - 1; char_y >= 0; char_y--) {

  969.                 uint8_t *p = pic->data[0] + (y + i * 10) * pic->linesize[0] + x;

  970.                 for (mask = 0x80; mask; mask >>= 1) {

  971.                     if (font[txt[i] * font_height + font_height - 1 - char_y] & mask)

  972.                         p[char_y] = ~p[char_y];

  973.                     p += pic->linesize[0];

  974.                 }

  975.             }

  976.         } else {

  977.             uint8_t *p = pic->data[0] + y*pic->linesize[0] + (x + i*8);

  978.             for (char_y = 0; char_y < font_height; char_y++) {

  979.                 for (mask = 0x80; mask; mask >>= 1) {

  980.                     if (font[txt[i] * font_height + char_y] & mask)

  981.                         *p = ~(*p);

  982.                     p++;

  983.                 }

  984.                 p += pic->linesize[0] - 8;

  985.             }

  986.         }

  987.     }

  988. }

  989. 

  990. static int showspectrumpic_request_frame(AVFilterLink *outlink)

  991. {

  992.     AVFilterContext *ctx = outlink->src;

  993.     ShowSpectrumContext *s = ctx->priv;

  994.     AVFilterLink *inlink = ctx->inputs[0];

  995.     int ret;

  996. 

  997.     ret = ff_request_frame(inlink);

  998.     if (ret == AVERROR_EOF && s->outpicref) {

  999.         int samples = av_audio_fifo_size(s->fifo);

  1000.         int consumed = 0;

  1001.         int y, x = 0, sz = s->orientation == VERTICAL ? s->w : s->h;

  1002.         int ch, spf, spb;

  1003.         AVFrame *fin;

  1004. 

  1005.         spf = s->win_size * (samples / ((s->win_size * sz) * ceil(samples / (float)(s->win_size * sz))));

  1006.         spb = (samples / (spf * sz)) * spf;

  1007. 

  1008.         fin = ff_get_audio_buffer(inlink, s->win_size);

  1009.         if (!fin)

  1010.             return AVERROR(ENOMEM);

  1011. 

  1012.         while (x < sz) {

  1013.             ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data, s->win_size);

  1014.             if (ret < 0) {

  1015.                 av_frame_free(&fin);

  1016.                 return ret;

  1017.             }

  1018. 

  1019.             av_audio_fifo_drain(s->fifo, spf);

  1020. 

  1021.             if (ret < s->win_size) {

  1022.                 for (ch = 0; ch < s->nb_display_channels; ch++) {

  1023.                     memset(fin->extended_data[ch] + ret * sizeof(float), 0,

  1024.                            (s->win_size - ret) * sizeof(float));

  1025.                 }

  1026.             }

  1027. 

  1028.             ctx->internal->execute(ctx, run_channel_fft, fin, NULL, s->nb_display_channels);

  1029.             acalc_magnitudes(s);

  1030. 

  1031.             consumed += spf;

  1032.             if (consumed >= spb) {

  1033.                 int h = s->orientation == VERTICAL ? s->h : s->w;

  1034. 

  1035.                 scale_magnitudes(s, 1. / (consumed / spf));

  1036.                 plot_spectrum_column(inlink, fin);

  1037.                 consumed = 0;

  1038.                 x++;

  1039.                 for (ch = 0; ch < s->nb_display_channels; ch++)

  1040.                     memset(s->magnitudes[ch], 0, h * sizeof(float));

  1041.             }

  1042.         }

  1043. 

  1044.         av_frame_free(&fin);

  1045.         s->outpicref->pts = 0;

  1046. 

  1047.         if (s->legend) {

  1048.             int multi = (s->mode == SEPARATE && s->color_mode == CHANNEL);

  1049.             float spp = samples / (float)sz;

  1050.             uint8_t *dst;

  1051. 

  1052.             drawtext(s->outpicref, 2, outlink->h - 10, "CREATED BY LIBAVFILTER", 0);

  1053. 

  1054.             dst = s->outpicref->data[0] + (s->start_y - 1) * s->outpicref->linesize[0] + s->start_x - 1;

  1055.             for (x = 0; x < s->w + 1; x++)

  1056.                 dst[x] = 200;

  1057.             dst = s->outpicref->data[0] + (s->start_y + s->h) * s->outpicref->linesize[0] + s->start_x - 1;

  1058.             for (x = 0; x < s->w + 1; x++)

  1059.                 dst[x] = 200;

  1060.             for (y = 0; y < s->h + 2; y++) {

  1061.                 dst = s->outpicref->data[0] + (y + s->start_y - 1) * s->outpicref->linesize[0];

  1062.                 dst[s->start_x - 1] = 200;

  1063.                 dst[s->start_x + s->w] = 200;

  1064.             }

  1065.             if (s->orientation == VERTICAL) {

  1066.                 int h = s->mode == SEPARATE ? s->h / s->nb_display_channels : s->h;

  1067.                 for (ch = 0; ch < (s->mode == SEPARATE ? s->nb_display_channels : 1); ch++) {

  1068.                     for (y = 0; y < h; y += 20) {

  1069.                         dst = s->outpicref->data[0] + (s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[0];

  1070.                         dst[s->start_x - 2] = 200;

  1071.                         dst[s->start_x + s->w + 1] = 200;

  1072.                     }

  1073.                     for (y = 0; y < h; y += 40) {

  1074.                         dst = s->outpicref->data[0] + (s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[0];

  1075.                         dst[s->start_x - 3] = 200;

  1076.                         dst[s->start_x + s->w + 2] = 200;

  1077.                     }

  1078.                     dst = s->outpicref->data[0] + (s->start_y - 2) * s->outpicref->linesize[0] + s->start_x;

  1079.                     for (x = 0; x < s->w; x+=40)

  1080.                         dst[x] = 200;

  1081.                     dst = s->outpicref->data[0] + (s->start_y - 3) * s->outpicref->linesize[0] + s->start_x;

  1082.                     for (x = 0; x < s->w; x+=80)

  1083.                         dst[x] = 200;

  1084.                     dst = s->outpicref->data[0] + (s->h + s->start_y + 1) * s->outpicref->linesize[0] + s->start_x;

  1085.                     for (x = 0; x < s->w; x+=40) {

  1086.                         dst[x] = 200;

  1087.                     }

  1088.                     dst = s->outpicref->data[0] + (s->h + s->start_y + 2) * s->outpicref->linesize[0] + s->start_x;

  1089.                     for (x = 0; x < s->w; x+=80) {

  1090.                         dst[x] = 200;

  1091.                     }

  1092.                     for (y = 0; y < h; y += 40) {

  1093.                         float hz = y * (inlink->sample_rate / 2) / (float)(1 << (int)ceil(log2(h)));

  1094.                         char *units;

  1095. 

  1096.                         if (hz == 0)

  1097.                             units = av_asprintf("DC");

  1098.                         else

  1099.                             units = av_asprintf("%.2f", hz);

  1100.                         if (!units)

  1101.                             return AVERROR(ENOMEM);

  1102. 

  1103.                         drawtext(s->outpicref, s->start_x - 8 * strlen(units) - 4, h * (ch + 1) + s->start_y - y - 4, units, 0);

  1104.                         av_free(units);

  1105.                     }

  1106.                 }

  1107. 

  1108.                 for (x = 0; x < s->w; x+=80) {

  1109.                     float seconds = x * spp / inlink->sample_rate;

  1110.                     char *units;

  1111. 

  1112.                     if (x == 0)

  1113.                         units = av_asprintf("0");

  1114.                     else if (log10(seconds) > 6)

  1115.                         units = av_asprintf("%.2fh", seconds / (60 * 60));

  1116.                     else if (log10(seconds) > 3)

  1117.                         units = av_asprintf("%.2fm", seconds / 60);

  1118.                     else

  1119.                         units = av_asprintf("%.2fs", seconds);

  1120.                     if (!units)

  1121.                         return AVERROR(ENOMEM);

  1122. 

  1123.                     drawtext(s->outpicref, s->start_x + x - 4 * strlen(units), s->h + s->start_y + 6, units, 0);

  1124.                     drawtext(s->outpicref, s->start_x + x - 4 * strlen(units), s->start_y - 12, units, 0);

  1125.                     av_free(units);

  1126.                 }

  1127. 

  1128.                 drawtext(s->outpicref, outlink->w / 2 - 4 * 4, outlink->h - s->start_y / 2, "TIME", 0);

  1129.                 drawtext(s->outpicref, s->start_x / 7, outlink->h / 2 - 14 * 4, "FREQUENCY (Hz)", 1);

  1130.             } else {

  1131.                 int w = s->mode == SEPARATE ? s->w / s->nb_display_channels : s->w;

  1132.                 for (y = 0; y < s->h; y += 20) {

  1133.                     dst = s->outpicref->data[0] + (s->start_y + y) * s->outpicref->linesize[0];

  1134.                     dst[s->start_x - 2] = 200;

  1135.                     dst[s->start_x + s->w + 1] = 200;

  1136.                 }

  1137.                 for (y = 0; y < s->h; y += 40) {

  1138.                     dst = s->outpicref->data[0] + (s->start_y + y) * s->outpicref->linesize[0];

  1139.                     dst[s->start_x - 3] = 200;

  1140.                     dst[s->start_x + s->w + 2] = 200;

  1141.                 }

  1142.                 for (ch = 0; ch < (s->mode == SEPARATE ? s->nb_display_channels : 1); ch++) {

  1143.                     dst = s->outpicref->data[0] + (s->start_y - 2) * s->outpicref->linesize[0] + s->start_x + w * ch;

  1144.                     for (x = 0; x < w; x+=40)

  1145.                         dst[x] = 200;

  1146.                     dst = s->outpicref->data[0] + (s->start_y - 3) * s->outpicref->linesize[0] + s->start_x + w * ch;

  1147.                     for (x = 0; x < w; x+=80)

  1148.                         dst[x] = 200;

  1149.                     dst = s->outpicref->data[0] + (s->h + s->start_y + 1) * s->outpicref->linesize[0] + s->start_x + w * ch;

  1150.                     for (x = 0; x < w; x+=40) {

  1151.                         dst[x] = 200;

  1152.                     }

  1153.                     dst = s->outpicref->data[0] + (s->h + s->start_y + 2) * s->outpicref->linesize[0] + s->start_x + w * ch;

  1154.                     for (x = 0; x < w; x+=80) {

  1155.                         dst[x] = 200;

  1156.                     }

  1157.                     for (x = 0; x < w; x += 80) {

  1158.                         float hz = x * (inlink->sample_rate / 2) / (float)(1 << (int)ceil(log2(w)));

  1159.                         char *units;

  1160. 

  1161.                         if (hz == 0)

  1162.                             units = av_asprintf("DC");

  1163.                         else

  1164.                             units = av_asprintf("%.2f", hz);

  1165.                         if (!units)

  1166.                             return AVERROR(ENOMEM);

  1167. 

  1168.                         drawtext(s->outpicref, s->start_x - 4 * strlen(units) + x + w * ch, s->start_y - 12, units, 0);

  1169.                         drawtext(s->outpicref, s->start_x - 4 * strlen(units) + x + w * ch, s->h + s->start_y + 6, units, 0);

  1170.                         av_free(units);

  1171.                     }

  1172.                 }

  1173.                 for (y = 0; y < s->h; y+=40) {

  1174.                     float seconds = y * spp / inlink->sample_rate;

  1175.                     char *units;

  1176. 

  1177.                     if (x == 0)

  1178.                         units = av_asprintf("0");

  1179.                     else if (log10(seconds) > 6)

  1180.                         units = av_asprintf("%.2fh", seconds / (60 * 60));

  1181.                     else if (log10(seconds) > 3)

  1182.                         units = av_asprintf("%.2fm", seconds / 60);

  1183.                     else

  1184.                         units = av_asprintf("%.2fs", seconds);

  1185.                     if (!units)

  1186.                         return AVERROR(ENOMEM);

  1187. 

  1188.                     drawtext(s->outpicref, s->start_x - 8 * strlen(units) - 4, s->start_y + y - 4, units, 0);

  1189.                     av_free(units);

  1190.                 }

  1191.                 drawtext(s->outpicref, s->start_x / 7, outlink->h / 2 - 4 * 4, "TIME", 1);

  1192.                 drawtext(s->outpicref, outlink->w / 2 - 14 * 4, outlink->h - s->start_y / 2, "FREQUENCY (Hz)", 0);

  1193.             }

  1194. 

  1195.             for (ch = 0; ch < (multi ? s->nb_display_channels : 1); ch++) {

  1196.                 int h = multi ? s->h / s->nb_display_channels : s->h;

  1197. 

  1198.                 for (y = 0; y < h; y++) {

  1199.                     float out[3] = { 0., 127.5, 127.5};

  1200.                     int chn;

  1201. 

  1202.                     for (chn = 0; chn < (s->mode == SEPARATE ? 1 : s->nb_display_channels); chn++) {

  1203.                         float yf, uf, vf;

  1204.                         int channel = (multi) ? s->nb_display_channels - ch - 1 : chn;

  1205. 

  1206.                         color_range(s, channel, &yf, &uf, &vf);

  1207.                         pick_color(s, yf, uf, vf, y / (float)h, out);

  1208.                     }

  1209.                     memset(s->outpicref->data[0]+(s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[0] + s->w + s->start_x + 20, av_clip_uint8(out[0]), 10);

  1210.                     memset(s->outpicref->data[1]+(s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[1] + s->w + s->start_x + 20, av_clip_uint8(out[1]), 10);

  1211.                     memset(s->outpicref->data[2]+(s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[2] + s->w + s->start_x + 20, av_clip_uint8(out[2]), 10);

  1212.                 }

  1213. 

  1214.                 for (y = 0; ch == 0 && y < h; y += h / 10) {

  1215.                     float value = 120.0 * log10(1. - y / (float)h);

  1216.                     char *text;

  1217. 

  1218.                     if (value < -120)

  1219.                         break;

  1220.                     text = av_asprintf("%.0f dB", value);

  1221.                     if (!text)

  1222.                         continue;

  1223.                     drawtext(s->outpicref, s->w + s->start_x + 35, s->start_y + y - 5, text, 0);

  1224.                 }

  1225.             }

  1226.         }

  1227. 

  1228.         ret = ff_filter_frame(outlink, s->outpicref);

  1229.         s->outpicref = NULL;

  1230.     }

  1231. 

  1232.     return ret;

  1233. }

  1234. 

  1235. static int showspectrumpic_filter_frame(AVFilterLink *inlink, AVFrame *insamples)

  1236. {

  1237.     AVFilterContext *ctx = inlink->dst;

  1238.     ShowSpectrumContext *s = ctx->priv;

  1239.     int ret;

  1240. 

  1241.     ret = av_audio_fifo_write(s->fifo, (void **)insamples->extended_data, insamples->nb_samples);

  1242.     av_frame_free(&insamples);

  1243.     return ret;

  1244. }

  1245. 

  1246. static const AVFilterPad showspectrumpic_inputs[] = {

  1247.     {

  1248.         .name         = "default",

  1249.         .type         = AVMEDIA_TYPE_AUDIO,

  1250.         .filter_frame = showspectrumpic_filter_frame,

  1251.     },

  1252.     { NULL }

  1253. };

  1254. 

  1255. static const AVFilterPad showspectrumpic_outputs[] = {

  1256.     {

  1257.         .name          = "default",

  1258.         .type          = AVMEDIA_TYPE_VIDEO,

  1259.         .config_props  = config_output,

  1260.         .request_frame = showspectrumpic_request_frame,

  1261.     },

  1262.     { NULL }

  1263. };

  1264. 

  1265. AVFilter ff_avf_showspectrumpic = {

  1266.     .name          = "showspectrumpic",

  1267.     .description   = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output single picture."),

  1268.     .uninit        = uninit,

  1269.     .query_formats = query_formats,

  1270.     .priv_size     = sizeof(ShowSpectrumContext),

  1271.     .inputs        = showspectrumpic_inputs,

  1272.     .outputs       = showspectrumpic_outputs,

  1273.     .priv_class    = &showspectrumpic_class,

  1274.     .flags         = AVFILTER_FLAG_SLICE_THREADS,

  1275. };

  1276. 

  1277. #endif // CONFIG_SHOWSPECTRUMPIC_FILTER