FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
avf_showspectrum.c
Go to the documentation of this file.
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"
36 #include "libavutil/opt.h"
38 #include "audio.h"
39 #include "video.h"
40 #include "avfilter.h"
41 #include "internal.h"
42 #include "window_func.h"
43 
50 
51 typedef struct {
52  const AVClass *class;
53  int w, h;
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  int data;
65  int xpos; ///< x position (current column)
66  FFTContext *fft; ///< Fast Fourier Transform context
67  int fft_bits; ///< number of bits (FFT window size = 1<<fft_bits)
68  FFTComplex **fft_data; ///< bins holder for each (displayed) channels
69  float *window_func_lut; ///< Window function LUT
70  float **magnitudes;
71  float **phases;
72  int win_func;
73  int win_size;
74  double win_scale;
75  float overlap;
76  float gain;
77  int hop_size;
78  float *combine_buffer; ///< color combining buffer (3 * h items)
80  int64_t pts;
82  int legend;
83  int start_x, start_y;
85 
86 #define OFFSET(x) offsetof(ShowSpectrumContext, x)
87 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
88 
89 static const AVOption showspectrum_options[] = {
90  { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },
91  { "s", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },
92  { "slide", "set sliding mode", OFFSET(sliding), AV_OPT_TYPE_INT, {.i64 = 0}, 0, NB_SLIDES-1, FLAGS, "slide" },
93  { "replace", "replace old columns with new", 0, AV_OPT_TYPE_CONST, {.i64=REPLACE}, 0, 0, FLAGS, "slide" },
94  { "scroll", "scroll from right to left", 0, AV_OPT_TYPE_CONST, {.i64=SCROLL}, 0, 0, FLAGS, "slide" },
95  { "rscroll", "scroll from left to right", 0, AV_OPT_TYPE_CONST, {.i64=RSCROLL}, 0, 0, FLAGS, "slide" },
96  { "fullframe", "return full frames", 0, AV_OPT_TYPE_CONST, {.i64=FULLFRAME}, 0, 0, FLAGS, "slide" },
97  { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, COMBINED, NB_MODES-1, FLAGS, "mode" },
98  { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },
99  { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },
100  { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=CHANNEL}, CHANNEL, NB_CLMODES-1, FLAGS, "color" },
101  { "channel", "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL}, 0, 0, FLAGS, "color" },
102  { "intensity", "intensity based coloring", 0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },
103  { "rainbow", "rainbow based coloring", 0, AV_OPT_TYPE_CONST, {.i64=RAINBOW}, 0, 0, FLAGS, "color" },
104  { "moreland", "moreland based coloring", 0, AV_OPT_TYPE_CONST, {.i64=MORELAND}, 0, 0, FLAGS, "color" },
105  { "nebulae", "nebulae based coloring", 0, AV_OPT_TYPE_CONST, {.i64=NEBULAE}, 0, 0, FLAGS, "color" },
106  { "fire", "fire based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FIRE}, 0, 0, FLAGS, "color" },
107  { "fiery", "fiery based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FIERY}, 0, 0, FLAGS, "color" },
108  { "fruit", "fruit based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FRUIT}, 0, 0, FLAGS, "color" },
109  { "cool", "cool based coloring", 0, AV_OPT_TYPE_CONST, {.i64=COOL}, 0, 0, FLAGS, "color" },
110  { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=SQRT}, LINEAR, NB_SCALES-1, FLAGS, "scale" },
111  { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT}, 0, 0, FLAGS, "scale" },
112  { "cbrt", "cubic root", 0, AV_OPT_TYPE_CONST, {.i64=CBRT}, 0, 0, FLAGS, "scale" },
113  { "4thrt","4th root", 0, AV_OPT_TYPE_CONST, {.i64=FOURTHRT}, 0, 0, FLAGS, "scale" },
114  { "5thrt","5th root", 0, AV_OPT_TYPE_CONST, {.i64=FIFTHRT}, 0, 0, FLAGS, "scale" },
115  { "log", "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG}, 0, 0, FLAGS, "scale" },
116  { "lin", "linear", 0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },
117  { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },
118  { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },
119  { "rect", "Rectangular", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT}, 0, 0, FLAGS, "win_func" },
120  { "bartlett", "Bartlett", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },
121  { "hann", "Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
122  { "hanning", "Hanning", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
123  { "hamming", "Hamming", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING}, 0, 0, FLAGS, "win_func" },
124  { "blackman", "Blackman", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },
125  { "welch", "Welch", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH}, 0, 0, FLAGS, "win_func" },
126  { "flattop", "Flat-top", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP}, 0, 0, FLAGS, "win_func" },
127  { "bharris", "Blackman-Harris", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS}, 0, 0, FLAGS, "win_func" },
128  { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },
129  { "bhann", "Bartlett-Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN}, 0, 0, FLAGS, "win_func" },
130  { "sine", "Sine", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE}, 0, 0, FLAGS, "win_func" },
131  { "nuttall", "Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL}, 0, 0, FLAGS, "win_func" },
132  { "lanczos", "Lanczos", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS}, 0, 0, FLAGS, "win_func" },
133  { "gauss", "Gauss", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS}, 0, 0, FLAGS, "win_func" },
134  { "tukey", "Tukey", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_TUKEY}, 0, 0, FLAGS, "win_func" },
135  { "orientation", "set orientation", OFFSET(orientation), AV_OPT_TYPE_INT, {.i64=VERTICAL}, 0, NB_ORIENTATIONS-1, FLAGS, "orientation" },
136  { "vertical", NULL, 0, AV_OPT_TYPE_CONST, {.i64=VERTICAL}, 0, 0, FLAGS, "orientation" },
137  { "horizontal", NULL, 0, AV_OPT_TYPE_CONST, {.i64=HORIZONTAL}, 0, 0, FLAGS, "orientation" },
138  { "overlap", "set window overlap", OFFSET(overlap), AV_OPT_TYPE_FLOAT, {.dbl = 0}, 0, 1, FLAGS },
139  { "gain", "set scale gain", OFFSET(gain), AV_OPT_TYPE_FLOAT, {.dbl = 1}, 0, 128, FLAGS },
140  { "data", "set data mode", OFFSET(data), AV_OPT_TYPE_INT, {.i64 = 0}, 0, NB_DMODES-1, FLAGS, "data" },
141  { "magnitude", NULL, 0, AV_OPT_TYPE_CONST, {.i64=D_MAGNITUDE}, 0, 0, FLAGS, "data" },
142  { "phase", NULL, 0, AV_OPT_TYPE_CONST, {.i64=D_PHASE}, 0, 0, FLAGS, "data" },
143  { NULL }
144 };
145 
146 AVFILTER_DEFINE_CLASS(showspectrum);
147 
148 static const struct ColorTable {
149  float a, y, u, v;
150 } color_table[][8] = {
151  [INTENSITY] = {
152  { 0, 0, 0, 0 },
153  { 0.13, .03587126228984074, .1573300977624594, -.02548747583751842 },
154  { 0.30, .18572281794568020, .1772436246393981, .17475554840414750 },
155  { 0.60, .28184980583656130, -.1593064119945782, .47132074554608920 },
156  { 0.73, .65830621175547810, -.3716070802232764, .24352759331252930 },
157  { 0.78, .76318535758242900, -.4307467689263783, .16866496622310430 },
158  { 0.91, .95336363636363640, -.2045454545454546, .03313636363636363 },
159  { 1, 1, 0, 0 }},
160  [RAINBOW] = {
161  { 0, 0, 0, 0 },
162  { 0.13, 44/256., (189-128)/256., (138-128)/256. },
163  { 0.25, 29/256., (186-128)/256., (119-128)/256. },
164  { 0.38, 119/256., (194-128)/256., (53-128)/256. },
165  { 0.60, 111/256., (73-128)/256., (59-128)/256. },
166  { 0.73, 205/256., (19-128)/256., (149-128)/256. },
167  { 0.86, 135/256., (83-128)/256., (200-128)/256. },
168  { 1, 73/256., (95-128)/256., (225-128)/256. }},
169  [MORELAND] = {
170  { 0, 44/256., (181-128)/256., (112-128)/256. },
171  { 0.13, 126/256., (177-128)/256., (106-128)/256. },
172  { 0.25, 164/256., (163-128)/256., (109-128)/256. },
173  { 0.38, 200/256., (140-128)/256., (120-128)/256. },
174  { 0.60, 201/256., (117-128)/256., (141-128)/256. },
175  { 0.73, 177/256., (103-128)/256., (165-128)/256. },
176  { 0.86, 136/256., (100-128)/256., (183-128)/256. },
177  { 1, 68/256., (117-128)/256., (203-128)/256. }},
178  [NEBULAE] = {
179  { 0, 10/256., (134-128)/256., (132-128)/256. },
180  { 0.23, 21/256., (137-128)/256., (130-128)/256. },
181  { 0.45, 35/256., (134-128)/256., (134-128)/256. },
182  { 0.57, 51/256., (130-128)/256., (139-128)/256. },
183  { 0.67, 104/256., (116-128)/256., (162-128)/256. },
184  { 0.77, 120/256., (105-128)/256., (188-128)/256. },
185  { 0.87, 140/256., (105-128)/256., (188-128)/256. },
186  { 1, 1, 0, 0 }},
187  [FIRE] = {
188  { 0, 0, 0, 0 },
189  { 0.23, 44/256., (132-128)/256., (127-128)/256. },
190  { 0.45, 62/256., (116-128)/256., (140-128)/256. },
191  { 0.57, 75/256., (105-128)/256., (152-128)/256. },
192  { 0.67, 95/256., (91-128)/256., (166-128)/256. },
193  { 0.77, 126/256., (74-128)/256., (172-128)/256. },
194  { 0.87, 164/256., (73-128)/256., (162-128)/256. },
195  { 1, 1, 0, 0 }},
196  [FIERY] = {
197  { 0, 0, 0, 0 },
198  { 0.23, 36/256., (116-128)/256., (163-128)/256. },
199  { 0.45, 52/256., (102-128)/256., (200-128)/256. },
200  { 0.57, 116/256., (84-128)/256., (196-128)/256. },
201  { 0.67, 157/256., (67-128)/256., (181-128)/256. },
202  { 0.77, 193/256., (40-128)/256., (155-128)/256. },
203  { 0.87, 221/256., (101-128)/256., (134-128)/256. },
204  { 1, 1, 0, 0 }},
205  [FRUIT] = {
206  { 0, 0, 0, 0 },
207  { 0.20, 29/256., (136-128)/256., (119-128)/256. },
208  { 0.30, 60/256., (119-128)/256., (90-128)/256. },
209  { 0.40, 85/256., (91-128)/256., (85-128)/256. },
210  { 0.50, 116/256., (70-128)/256., (105-128)/256. },
211  { 0.60, 151/256., (50-128)/256., (146-128)/256. },
212  { 0.70, 191/256., (63-128)/256., (178-128)/256. },
213  { 1, 98/256., (80-128)/256., (221-128)/256. }},
214  [COOL] = {
215  { 0, 0, 0, 0 },
216  { .15, 0, .5, -.5 },
217  { 1, 1, -.5, .5 }},
218 };
219 
221 {
222  ShowSpectrumContext *s = ctx->priv;
223  int i;
224 
226  av_fft_end(s->fft);
227  if (s->fft_data) {
228  for (i = 0; i < s->nb_display_channels; i++)
229  av_freep(&s->fft_data[i]);
230  }
231  av_freep(&s->fft_data);
233  if (s->magnitudes) {
234  for (i = 0; i < s->nb_display_channels; i++)
235  av_freep(&s->magnitudes[i]);
236  }
237  av_freep(&s->magnitudes);
240  if (s->phases) {
241  for (i = 0; i < s->nb_display_channels; i++)
242  av_freep(&s->phases[i]);
243  }
244  av_freep(&s->phases);
245 }
246 
248 {
251  AVFilterLink *inlink = ctx->inputs[0];
252  AVFilterLink *outlink = ctx->outputs[0];
255  int ret;
256 
257  /* set input audio formats */
258  formats = ff_make_format_list(sample_fmts);
259  if ((ret = ff_formats_ref(formats, &inlink->out_formats)) < 0)
260  return ret;
261 
262  layouts = ff_all_channel_layouts();
263  if ((ret = ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts)) < 0)
264  return ret;
265 
266  formats = ff_all_samplerates();
267  if ((ret = ff_formats_ref(formats, &inlink->out_samplerates)) < 0)
268  return ret;
269 
270  /* set output video format */
271  formats = ff_make_format_list(pix_fmts);
272  if ((ret = ff_formats_ref(formats, &outlink->in_formats)) < 0)
273  return ret;
274 
275  return 0;
276 }
277 
278 static int config_output(AVFilterLink *outlink)
279 {
280  AVFilterContext *ctx = outlink->src;
281  AVFilterLink *inlink = ctx->inputs[0];
282  ShowSpectrumContext *s = ctx->priv;
283  int i, fft_bits, h, w;
284  float overlap;
285 
286  if (!strcmp(ctx->filter->name, "showspectrumpic"))
287  s->single_pic = 1;
288 
289  outlink->w = s->w;
290  outlink->h = s->h;
291 
292  if (s->legend) {
293  s->start_x = log10(inlink->sample_rate) * 25;
294  s->start_y = 64;
295  outlink->w += s->start_x * 2;
296  outlink->h += s->start_y * 2;
297  }
298 
299  h = (s->mode == COMBINED || s->orientation == HORIZONTAL) ? s->h : s->h / inlink->channels;
300  w = (s->mode == COMBINED || s->orientation == VERTICAL) ? s->w : s->w / inlink->channels;
301  s->channel_height = h;
302  s->channel_width = w;
303 
304  if (s->orientation == VERTICAL) {
305  /* FFT window size (precision) according to the requested output frame height */
306  for (fft_bits = 1; 1 << fft_bits < 2 * h; fft_bits++);
307  } else {
308  /* FFT window size (precision) according to the requested output frame width */
309  for (fft_bits = 1; 1 << fft_bits < 2 * w; fft_bits++);
310  }
311  s->win_size = 1 << fft_bits;
312 
313  /* (re-)configuration if the video output changed (or first init) */
314  if (fft_bits != s->fft_bits) {
315  AVFrame *outpicref;
316 
317  av_fft_end(s->fft);
318  s->fft = av_fft_init(fft_bits, 0);
319  if (!s->fft) {
320  av_log(ctx, AV_LOG_ERROR, "Unable to create FFT context. "
321  "The window size might be too high.\n");
322  return AVERROR(EINVAL);
323  }
324  s->fft_bits = fft_bits;
325 
326  /* FFT buffers: x2 for each (display) channel buffer.
327  * Note: we use free and malloc instead of a realloc-like function to
328  * make sure the buffer is aligned in memory for the FFT functions. */
329  for (i = 0; i < s->nb_display_channels; i++)
330  av_freep(&s->fft_data[i]);
331  av_freep(&s->fft_data);
332  s->nb_display_channels = inlink->channels;
333 
334  s->magnitudes = av_calloc(s->nb_display_channels, sizeof(*s->magnitudes));
335  if (!s->magnitudes)
336  return AVERROR(ENOMEM);
337  for (i = 0; i < s->nb_display_channels; i++) {
338  s->magnitudes[i] = av_calloc(s->orientation == VERTICAL ? s->h : s->w, sizeof(**s->magnitudes));
339  if (!s->magnitudes[i])
340  return AVERROR(ENOMEM);
341  }
342 
343  s->phases = av_calloc(s->nb_display_channels, sizeof(*s->magnitudes));
344  if (!s->phases)
345  return AVERROR(ENOMEM);
346  for (i = 0; i < s->nb_display_channels; i++) {
347  s->phases[i] = av_calloc(s->orientation == VERTICAL ? s->h : s->w, sizeof(**s->phases));
348  if (!s->phases[i])
349  return AVERROR(ENOMEM);
350  }
351 
352  s->fft_data = av_calloc(s->nb_display_channels, sizeof(*s->fft_data));
353  if (!s->fft_data)
354  return AVERROR(ENOMEM);
355  for (i = 0; i < s->nb_display_channels; i++) {
356  s->fft_data[i] = av_calloc(s->win_size, sizeof(**s->fft_data));
357  if (!s->fft_data[i])
358  return AVERROR(ENOMEM);
359  }
360 
361  /* pre-calc windowing function */
362  s->window_func_lut =
364  sizeof(*s->window_func_lut));
365  if (!s->window_func_lut)
366  return AVERROR(ENOMEM);
368  if (s->overlap == 1)
369  s->overlap = overlap;
370  s->hop_size = (1. - s->overlap) * s->win_size;
371  if (s->hop_size < 1) {
372  av_log(ctx, AV_LOG_ERROR, "overlap %f too big\n", s->overlap);
373  return AVERROR(EINVAL);
374  }
375 
376  for (s->win_scale = 0, i = 0; i < s->win_size; i++) {
377  s->win_scale += s->window_func_lut[i] * s->window_func_lut[i];
378  }
379  s->win_scale = 1. / sqrt(s->win_scale);
380 
381  /* prepare the initial picref buffer (black frame) */
383  s->outpicref = outpicref =
384  ff_get_video_buffer(outlink, outlink->w, outlink->h);
385  if (!outpicref)
386  return AVERROR(ENOMEM);
387  outlink->sample_aspect_ratio = (AVRational){1,1};
388  for (i = 0; i < outlink->h; i++) {
389  memset(outpicref->data[0] + i * outpicref->linesize[0], 0, outlink->w);
390  memset(outpicref->data[1] + i * outpicref->linesize[1], 128, outlink->w);
391  memset(outpicref->data[2] + i * outpicref->linesize[2], 128, outlink->w);
392  }
394  }
395 
396  if ((s->orientation == VERTICAL && s->xpos >= s->w) ||
397  (s->orientation == HORIZONTAL && s->xpos >= s->h))
398  s->xpos = 0;
399 
400  outlink->frame_rate = av_make_q(inlink->sample_rate, s->win_size * (1.-s->overlap));
401  if (s->orientation == VERTICAL && s->sliding == FULLFRAME)
402  outlink->frame_rate.den *= s->w;
403  if (s->orientation == HORIZONTAL && s->sliding == FULLFRAME)
404  outlink->frame_rate.den *= s->h;
405 
406  if (s->orientation == VERTICAL) {
407  s->combine_buffer =
408  av_realloc_f(s->combine_buffer, s->h * 3,
409  sizeof(*s->combine_buffer));
410  } else {
411  s->combine_buffer =
412  av_realloc_f(s->combine_buffer, s->w * 3,
413  sizeof(*s->combine_buffer));
414  }
415 
416  av_log(ctx, AV_LOG_VERBOSE, "s:%dx%d FFT window size:%d\n",
417  s->w, s->h, s->win_size);
418 
420  s->fifo = av_audio_fifo_alloc(inlink->format, inlink->channels, s->win_size);
421  if (!s->fifo)
422  return AVERROR(ENOMEM);
423  return 0;
424 }
425 
427 {
428  int ch, n;
429 
430  /* fill FFT input with the number of samples available */
431  for (ch = 0; ch < s->nb_display_channels; ch++) {
432  const float *p = (float *)fin->extended_data[ch];
433 
434  for (n = 0; n < s->win_size; n++) {
435  s->fft_data[ch][n].re = p[n] * s->window_func_lut[n];
436  s->fft_data[ch][n].im = 0;
437  }
438  }
439 
440  /* run FFT on each samples set */
441  for (ch = 0; ch < s->nb_display_channels; ch++) {
442  av_fft_permute(s->fft, s->fft_data[ch]);
443  av_fft_calc(s->fft, s->fft_data[ch]);
444  }
445 }
446 
447 #define RE(y, ch) s->fft_data[ch][y].re
448 #define IM(y, ch) s->fft_data[ch][y].im
449 #define MAGNITUDE(y, ch) hypot(RE(y, ch), IM(y, ch))
450 #define PHASE(y, ch) atan2(IM(y, ch), RE(y, ch))
451 
453 {
454  const double w = s->win_scale * (s->scale == LOG ? s->win_scale : 1);
455  int ch, y, h = s->orientation == VERTICAL ? s->h : s->w;
456  const float f = s->gain * w;
457 
458  for (ch = 0; ch < s->nb_display_channels; ch++) {
459  float *magnitudes = s->magnitudes[ch];
460 
461  for (y = 0; y < h; y++)
462  magnitudes[y] = MAGNITUDE(y, ch) * f;
463  }
464 }
465 
467 {
468  int ch, y, h = s->orientation == VERTICAL ? s->h : s->w;
469 
470  for (ch = 0; ch < s->nb_display_channels; ch++) {
471  float *phases = s->phases[ch];
472 
473  for (y = 0; y < h; y++)
474  phases[y] = (PHASE(y, ch) / M_PI + 1) / 2;
475  }
476 }
477 
479 {
480  const double w = s->win_scale * (s->scale == LOG ? s->win_scale : 1);
481  int ch, y, h = s->orientation == VERTICAL ? s->h : s->w;
482  const float f = s->gain * w;
483 
484  for (ch = 0; ch < s->nb_display_channels; ch++) {
485  float *magnitudes = s->magnitudes[ch];
486 
487  for (y = 0; y < h; y++)
488  magnitudes[y] += MAGNITUDE(y, ch) * f;
489  }
490 }
491 
492 static void scale_magnitudes(ShowSpectrumContext *s, float scale)
493 {
494  int ch, y, h = s->orientation == VERTICAL ? s->h : s->w;
495 
496  for (ch = 0; ch < s->nb_display_channels; ch++) {
497  float *magnitudes = s->magnitudes[ch];
498 
499  for (y = 0; y < h; y++)
500  magnitudes[y] *= scale;
501  }
502 }
503 
504 static void color_range(ShowSpectrumContext *s, int ch,
505  float *yf, float *uf, float *vf)
506 {
507  switch (s->mode) {
508  case COMBINED:
509  // reduce range by channel count
510  *yf = 256.0f / s->nb_display_channels;
511  switch (s->color_mode) {
512  case RAINBOW:
513  case MORELAND:
514  case NEBULAE:
515  case FIRE:
516  case FIERY:
517  case FRUIT:
518  case COOL:
519  case INTENSITY:
520  *uf = *yf;
521  *vf = *yf;
522  break;
523  case CHANNEL:
524  /* adjust saturation for mixed UV coloring */
525  /* this factor is correct for infinite channels, an approximation otherwise */
526  *uf = *yf * M_PI;
527  *vf = *yf * M_PI;
528  break;
529  default:
530  av_assert0(0);
531  }
532  break;
533  case SEPARATE:
534  // full range
535  *yf = 256.0f;
536  *uf = 256.0f;
537  *vf = 256.0f;
538  break;
539  default:
540  av_assert0(0);
541  }
542 
543  if (s->color_mode == CHANNEL) {
544  if (s->nb_display_channels > 1) {
545  *uf *= 0.5 * sin((2 * M_PI * ch) / s->nb_display_channels);
546  *vf *= 0.5 * cos((2 * M_PI * ch) / s->nb_display_channels);
547  } else {
548  *uf = 0.0f;
549  *vf = 0.0f;
550  }
551  }
552  *uf *= s->saturation;
553  *vf *= s->saturation;
554 }
555 
557  float yf, float uf, float vf,
558  float a, float *out)
559 {
560  if (s->color_mode > CHANNEL) {
561  const int cm = s->color_mode;
562  float y, u, v;
563  int i;
564 
565  for (i = 1; i < FF_ARRAY_ELEMS(color_table[cm]) - 1; i++)
566  if (color_table[cm][i].a >= a)
567  break;
568  // i now is the first item >= the color
569  // now we know to interpolate between item i - 1 and i
570  if (a <= color_table[cm][i - 1].a) {
571  y = color_table[cm][i - 1].y;
572  u = color_table[cm][i - 1].u;
573  v = color_table[cm][i - 1].v;
574  } else if (a >= color_table[cm][i].a) {
575  y = color_table[cm][i].y;
576  u = color_table[cm][i].u;
577  v = color_table[cm][i].v;
578  } else {
579  float start = color_table[cm][i - 1].a;
580  float end = color_table[cm][i].a;
581  float lerpfrac = (a - start) / (end - start);
582  y = color_table[cm][i - 1].y * (1.0f - lerpfrac)
583  + color_table[cm][i].y * lerpfrac;
584  u = color_table[cm][i - 1].u * (1.0f - lerpfrac)
585  + color_table[cm][i].u * lerpfrac;
586  v = color_table[cm][i - 1].v * (1.0f - lerpfrac)
587  + color_table[cm][i].v * lerpfrac;
588  }
589 
590  out[0] += y * yf;
591  out[1] += u * uf;
592  out[2] += v * vf;
593  } else {
594  out[0] += a * yf;
595  out[1] += a * uf;
596  out[2] += a * vf;
597  }
598 }
599 
601 {
602  int y;
603 
604  for (y = 0; y < size; y++) {
605  s->combine_buffer[3 * y ] = 0;
606  s->combine_buffer[3 * y + 1] = 127.5;
607  s->combine_buffer[3 * y + 2] = 127.5;
608  }
609 }
610 
611 static int plot_spectrum_column(AVFilterLink *inlink, AVFrame *insamples)
612 {
613  int ret;
614  AVFilterContext *ctx = inlink->dst;
615  AVFilterLink *outlink = ctx->outputs[0];
616  ShowSpectrumContext *s = ctx->priv;
617  AVFrame *outpicref = s->outpicref;
618  int h = s->orientation == VERTICAL ? s->channel_height : s->channel_width;
619 
620  int ch, plane, x, y;
621 
622  /* fill a new spectrum column */
623  /* initialize buffer for combining to black */
624  clear_combine_buffer(s, s->orientation == VERTICAL ? s->h : s->w);
625 
626  for (ch = 0; ch < s->nb_display_channels; ch++) {
627  float *magnitudes = s->magnitudes[ch];
628  float *phases = s->phases[ch];
629  float yf, uf, vf;
630 
631  /* decide color range */
632  color_range(s, ch, &yf, &uf, &vf);
633 
634  /* draw the channel */
635  for (y = 0; y < h; y++) {
636  int row = (s->mode == COMBINED) ? y : ch * h + y;
637  float *out = &s->combine_buffer[3 * row];
638  float a;
639 
640  switch (s->data) {
641  case D_MAGNITUDE:
642  /* get magnitude */
643  a = magnitudes[y];
644  break;
645  case D_PHASE:
646  /* get phase */
647  a = phases[y];
648  break;
649  default:
650  av_assert0(0);
651  }
652 
653  /* apply scale */
654  switch (s->scale) {
655  case LINEAR:
656  a = av_clipf(a, 0, 1);
657  break;
658  case SQRT:
659  a = av_clipf(sqrt(a), 0, 1);
660  break;
661  case CBRT:
662  a = av_clipf(cbrt(a), 0, 1);
663  break;
664  case FOURTHRT:
665  a = av_clipf(sqrt(sqrt(a)), 0, 1);
666  break;
667  case FIFTHRT:
668  a = av_clipf(pow(a, 0.20), 0, 1);
669  break;
670  case LOG:
671  a = 1 + log10(av_clipd(a, 1e-6, 1)) / 6; // zero = -120dBFS
672  break;
673  default:
674  av_assert0(0);
675  }
676 
677  pick_color(s, yf, uf, vf, a, out);
678  }
679  }
680 
682  /* copy to output */
683  if (s->orientation == VERTICAL) {
684  if (s->sliding == SCROLL) {
685  for (plane = 0; plane < 3; plane++) {
686  for (y = 0; y < s->h; y++) {
687  uint8_t *p = outpicref->data[plane] +
688  y * outpicref->linesize[plane];
689  memmove(p, p + 1, s->w - 1);
690  }
691  }
692  s->xpos = s->w - 1;
693  } else if (s->sliding == RSCROLL) {
694  for (plane = 0; plane < 3; plane++) {
695  for (y = 0; y < s->h; y++) {
696  uint8_t *p = outpicref->data[plane] +
697  y * outpicref->linesize[plane];
698  memmove(p + 1, p, s->w - 1);
699  }
700  }
701  s->xpos = 0;
702  }
703  for (plane = 0; plane < 3; plane++) {
704  uint8_t *p = outpicref->data[plane] + s->start_x +
705  (outlink->h - 1 - s->start_y) * outpicref->linesize[plane] +
706  s->xpos;
707  for (y = 0; y < s->h; y++) {
708  *p = lrintf(av_clipf(s->combine_buffer[3 * y + plane], 0, 255));
709  p -= outpicref->linesize[plane];
710  }
711  }
712  } else {
713  if (s->sliding == SCROLL) {
714  for (plane = 0; plane < 3; plane++) {
715  for (y = 1; y < s->h; y++) {
716  memmove(outpicref->data[plane] + (y-1) * outpicref->linesize[plane],
717  outpicref->data[plane] + (y ) * outpicref->linesize[plane],
718  s->w);
719  }
720  }
721  s->xpos = s->h - 1;
722  } else if (s->sliding == RSCROLL) {
723  for (plane = 0; plane < 3; plane++) {
724  for (y = s->h - 1; y >= 1; y--) {
725  memmove(outpicref->data[plane] + (y ) * outpicref->linesize[plane],
726  outpicref->data[plane] + (y-1) * outpicref->linesize[plane],
727  s->w);
728  }
729  }
730  s->xpos = 0;
731  }
732  for (plane = 0; plane < 3; plane++) {
733  uint8_t *p = outpicref->data[plane] + s->start_x +
734  (s->xpos + s->start_y) * outpicref->linesize[plane];
735  for (x = 0; x < s->w; x++) {
736  *p = lrintf(av_clipf(s->combine_buffer[3 * x + plane], 0, 255));
737  p++;
738  }
739  }
740  }
741 
742  if (s->sliding != FULLFRAME || s->xpos == 0)
743  outpicref->pts = insamples->pts;
744 
745  s->xpos++;
746  if (s->orientation == VERTICAL && s->xpos >= s->w)
747  s->xpos = 0;
748  if (s->orientation == HORIZONTAL && s->xpos >= s->h)
749  s->xpos = 0;
750  if (!s->single_pic && (s->sliding != FULLFRAME || s->xpos == 0)) {
751  ret = ff_filter_frame(outlink, av_frame_clone(s->outpicref));
752  if (ret < 0)
753  return ret;
754  }
755 
756  return s->win_size;
757 }
758 
759 #if CONFIG_SHOWSPECTRUM_FILTER
760 
761 static int request_frame(AVFilterLink *outlink)
762 {
763  ShowSpectrumContext *s = outlink->src->priv;
764  AVFilterLink *inlink = outlink->src->inputs[0];
765  unsigned i;
766  int ret;
767 
768  ret = ff_request_frame(inlink);
769  if (ret == AVERROR_EOF && s->sliding == FULLFRAME && s->xpos > 0 &&
770  s->outpicref) {
771  if (s->orientation == VERTICAL) {
772  for (i = 0; i < outlink->h; i++) {
773  memset(s->outpicref->data[0] + i * s->outpicref->linesize[0] + s->xpos, 0, outlink->w - s->xpos);
774  memset(s->outpicref->data[1] + i * s->outpicref->linesize[1] + s->xpos, 128, outlink->w - s->xpos);
775  memset(s->outpicref->data[2] + i * s->outpicref->linesize[2] + s->xpos, 128, outlink->w - s->xpos);
776  }
777  } else {
778  for (i = s->xpos; i < outlink->h; i++) {
779  memset(s->outpicref->data[0] + i * s->outpicref->linesize[0], 0, outlink->w);
780  memset(s->outpicref->data[1] + i * s->outpicref->linesize[1], 128, outlink->w);
781  memset(s->outpicref->data[2] + i * s->outpicref->linesize[2], 128, outlink->w);
782  }
783  }
784  ret = ff_filter_frame(outlink, s->outpicref);
785  s->outpicref = NULL;
786  }
787 
788  return ret;
789 }
790 
791 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
792 {
793  AVFilterContext *ctx = inlink->dst;
794  ShowSpectrumContext *s = ctx->priv;
795  AVFrame *fin = NULL;
796  int ret = 0, consumed = 0;
797 
798  if (s->pts == AV_NOPTS_VALUE)
799  s->pts = insamples->pts - av_audio_fifo_size(s->fifo);
800 
801  av_audio_fifo_write(s->fifo, (void **)insamples->extended_data, insamples->nb_samples);
802  av_frame_free(&insamples);
803  while (av_audio_fifo_size(s->fifo) >= s->win_size) {
804  fin = ff_get_audio_buffer(inlink, s->win_size);
805  if (!fin) {
806  ret = AVERROR(ENOMEM);
807  goto fail;
808  }
809 
810  fin->pts = s->pts + consumed;
811  consumed += s->hop_size;
812  ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data, s->win_size);
813  if (ret < 0)
814  goto fail;
815 
816  av_assert0(fin->nb_samples == s->win_size);
817 
818  run_fft(s, fin);
819  if (s->data == D_MAGNITUDE)
820  calc_magnitudes(s);
821  if (s->data == D_PHASE)
822  calc_phases(s);
823 
824  ret = plot_spectrum_column(inlink, fin);
825  av_frame_free(&fin);
827  if (ret < 0)
828  goto fail;
829  }
830 
831 fail:
832  s->pts = AV_NOPTS_VALUE;
833  av_frame_free(&fin);
834  return ret;
835 }
836 
837 static const AVFilterPad showspectrum_inputs[] = {
838  {
839  .name = "default",
840  .type = AVMEDIA_TYPE_AUDIO,
841  .filter_frame = filter_frame,
842  },
843  { NULL }
844 };
845 
846 static const AVFilterPad showspectrum_outputs[] = {
847  {
848  .name = "default",
849  .type = AVMEDIA_TYPE_VIDEO,
850  .config_props = config_output,
851  .request_frame = request_frame,
852  },
853  { NULL }
854 };
855 
856 AVFilter ff_avf_showspectrum = {
857  .name = "showspectrum",
858  .description = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output."),
859  .uninit = uninit,
860  .query_formats = query_formats,
861  .priv_size = sizeof(ShowSpectrumContext),
862  .inputs = showspectrum_inputs,
863  .outputs = showspectrum_outputs,
864  .priv_class = &showspectrum_class,
865 };
866 #endif // CONFIG_SHOWSPECTRUM_FILTER
867 
868 #if CONFIG_SHOWSPECTRUMPIC_FILTER
869 
870 static const AVOption showspectrumpic_options[] = {
871  { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "4096x2048"}, 0, 0, FLAGS },
872  { "s", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "4096x2048"}, 0, 0, FLAGS },
873  { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, 0, NB_MODES-1, FLAGS, "mode" },
874  { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },
875  { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },
876  { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=INTENSITY}, 0, NB_CLMODES-1, FLAGS, "color" },
877  { "channel", "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL}, 0, 0, FLAGS, "color" },
878  { "intensity", "intensity based coloring", 0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },
879  { "rainbow", "rainbow based coloring", 0, AV_OPT_TYPE_CONST, {.i64=RAINBOW}, 0, 0, FLAGS, "color" },
880  { "moreland", "moreland based coloring", 0, AV_OPT_TYPE_CONST, {.i64=MORELAND}, 0, 0, FLAGS, "color" },
881  { "nebulae", "nebulae based coloring", 0, AV_OPT_TYPE_CONST, {.i64=NEBULAE}, 0, 0, FLAGS, "color" },
882  { "fire", "fire based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FIRE}, 0, 0, FLAGS, "color" },
883  { "fiery", "fiery based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FIERY}, 0, 0, FLAGS, "color" },
884  { "fruit", "fruit based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FRUIT}, 0, 0, FLAGS, "color" },
885  { "cool", "cool based coloring", 0, AV_OPT_TYPE_CONST, {.i64=COOL}, 0, 0, FLAGS, "color" },
886  { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=LOG}, 0, NB_SCALES-1, FLAGS, "scale" },
887  { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT}, 0, 0, FLAGS, "scale" },
888  { "cbrt", "cubic root", 0, AV_OPT_TYPE_CONST, {.i64=CBRT}, 0, 0, FLAGS, "scale" },
889  { "4thrt","4th root", 0, AV_OPT_TYPE_CONST, {.i64=FOURTHRT}, 0, 0, FLAGS, "scale" },
890  { "5thrt","5th root", 0, AV_OPT_TYPE_CONST, {.i64=FIFTHRT}, 0, 0, FLAGS, "scale" },
891  { "log", "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG}, 0, 0, FLAGS, "scale" },
892  { "lin", "linear", 0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },
893  { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },
894  { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },
895  { "rect", "Rectangular", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT}, 0, 0, FLAGS, "win_func" },
896  { "bartlett", "Bartlett", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },
897  { "hann", "Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
898  { "hanning", "Hanning", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
899  { "hamming", "Hamming", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING}, 0, 0, FLAGS, "win_func" },
900  { "blackman", "Blackman", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },
901  { "welch", "Welch", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH}, 0, 0, FLAGS, "win_func" },
902  { "flattop", "Flat-top", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP}, 0, 0, FLAGS, "win_func" },
903  { "bharris", "Blackman-Harris", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS}, 0, 0, FLAGS, "win_func" },
904  { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },
905  { "bhann", "Bartlett-Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN}, 0, 0, FLAGS, "win_func" },
906  { "sine", "Sine", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE}, 0, 0, FLAGS, "win_func" },
907  { "nuttall", "Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL}, 0, 0, FLAGS, "win_func" },
908  { "lanczos", "Lanczos", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS}, 0, 0, FLAGS, "win_func" },
909  { "gauss", "Gauss", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS}, 0, 0, FLAGS, "win_func" },
910  { "tukey", "Tukey", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_TUKEY}, 0, 0, FLAGS, "win_func" },
911  { "orientation", "set orientation", OFFSET(orientation), AV_OPT_TYPE_INT, {.i64=VERTICAL}, 0, NB_ORIENTATIONS-1, FLAGS, "orientation" },
912  { "vertical", NULL, 0, AV_OPT_TYPE_CONST, {.i64=VERTICAL}, 0, 0, FLAGS, "orientation" },
913  { "horizontal", NULL, 0, AV_OPT_TYPE_CONST, {.i64=HORIZONTAL}, 0, 0, FLAGS, "orientation" },
914  { "gain", "set scale gain", OFFSET(gain), AV_OPT_TYPE_FLOAT, {.dbl = 1}, 0, 128, FLAGS },
915  { "legend", "draw legend", OFFSET(legend), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, FLAGS },
916  { NULL }
917 };
918 
919 AVFILTER_DEFINE_CLASS(showspectrumpic);
920 
921 static void drawtext(AVFrame *pic, int x, int y, const char *txt, int o)
922 {
923  const uint8_t *font;
924  int font_height;
925  int i;
926 
927  font = avpriv_cga_font, font_height = 8;
928 
929  for (i = 0; txt[i]; i++) {
930  int char_y, mask;
931 
932  if (o) {
933  for (char_y = font_height - 1; char_y >= 0; char_y--) {
934  uint8_t *p = pic->data[0] + (y + i * 10) * pic->linesize[0] + x;
935  for (mask = 0x80; mask; mask >>= 1) {
936  if (font[txt[i] * font_height + font_height - 1 - char_y] & mask)
937  p[char_y] = ~p[char_y];
938  p += pic->linesize[0];
939  }
940  }
941  } else {
942  uint8_t *p = pic->data[0] + y*pic->linesize[0] + (x + i*8);
943  for (char_y = 0; char_y < font_height; char_y++) {
944  for (mask = 0x80; mask; mask >>= 1) {
945  if (font[txt[i] * font_height + char_y] & mask)
946  *p = ~(*p);
947  p++;
948  }
949  p += pic->linesize[0] - 8;
950  }
951  }
952  }
953 }
954 
955 static int showspectrumpic_request_frame(AVFilterLink *outlink)
956 {
957  ShowSpectrumContext *s = outlink->src->priv;
958  AVFilterLink *inlink = outlink->src->inputs[0];
959  int ret;
960 
961  ret = ff_request_frame(inlink);
962  if (ret == AVERROR_EOF && s->outpicref) {
963  int samples = av_audio_fifo_size(s->fifo);
964  int consumed = 0;
965  int y, x = 0, sz = s->orientation == VERTICAL ? s->w : s->h;
966  int ch, spf, spb;
967  AVFrame *fin;
968 
969  spf = s->win_size * (samples / ((s->win_size * sz) * ceil(samples / (float)(s->win_size * sz))));
970  spb = (samples / (spf * sz)) * spf;
971 
972  fin = ff_get_audio_buffer(inlink, s->win_size);
973  if (!fin)
974  return AVERROR(ENOMEM);
975 
976  while (x < sz) {
977  ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data, s->win_size);
978  if (ret < 0) {
979  av_frame_free(&fin);
980  return ret;
981  }
982 
983  av_audio_fifo_drain(s->fifo, spf);
984 
985  if (ret < s->win_size) {
986  for (ch = 0; ch < s->nb_display_channels; ch++) {
987  memset(fin->extended_data[ch] + ret * sizeof(float), 0,
988  (s->win_size - ret) * sizeof(float));
989  }
990  }
991 
992  run_fft(s, fin);
993  acalc_magnitudes(s);
994 
995  consumed += spf;
996  if (consumed >= spb) {
997  int h = s->orientation == VERTICAL ? s->h : s->w;
998 
999  scale_magnitudes(s, 1. / (consumed / spf));
1000  plot_spectrum_column(inlink, fin);
1001  consumed = 0;
1002  x++;
1003  for (ch = 0; ch < s->nb_display_channels; ch++)
1004  memset(s->magnitudes[ch], 0, h * sizeof(float));
1005  }
1006  }
1007 
1008  av_frame_free(&fin);
1009  s->outpicref->pts = 0;
1010 
1011  if (s->legend) {
1012  int multi = (s->mode == SEPARATE && s->color_mode == CHANNEL);
1013  float spp = samples / (float)sz;
1014  uint8_t *dst;
1015 
1016  drawtext(s->outpicref, 2, outlink->h - 10, "CREATED BY LIBAVFILTER", 0);
1017 
1018  dst = s->outpicref->data[0] + (s->start_y - 1) * s->outpicref->linesize[0] + s->start_x - 1;
1019  for (x = 0; x < s->w + 1; x++)
1020  dst[x] = 200;
1021  dst = s->outpicref->data[0] + (s->start_y + s->h) * s->outpicref->linesize[0] + s->start_x - 1;
1022  for (x = 0; x < s->w + 1; x++)
1023  dst[x] = 200;
1024  for (y = 0; y < s->h + 2; y++) {
1025  dst = s->outpicref->data[0] + (y + s->start_y - 1) * s->outpicref->linesize[0];
1026  dst[s->start_x - 1] = 200;
1027  dst[s->start_x + s->w] = 200;
1028  }
1029  if (s->orientation == VERTICAL) {
1030  int h = s->mode == SEPARATE ? s->h / s->nb_display_channels : s->h;
1031  for (ch = 0; ch < (s->mode == SEPARATE ? s->nb_display_channels : 1); ch++) {
1032  for (y = 0; y < h; y += 20) {
1033  dst = s->outpicref->data[0] + (s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[0];
1034  dst[s->start_x - 2] = 200;
1035  dst[s->start_x + s->w + 1] = 200;
1036  }
1037  for (y = 0; y < h; y += 40) {
1038  dst = s->outpicref->data[0] + (s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[0];
1039  dst[s->start_x - 3] = 200;
1040  dst[s->start_x + s->w + 2] = 200;
1041  }
1042  dst = s->outpicref->data[0] + (s->start_y - 2) * s->outpicref->linesize[0] + s->start_x;
1043  for (x = 0; x < s->w; x+=40)
1044  dst[x] = 200;
1045  dst = s->outpicref->data[0] + (s->start_y - 3) * s->outpicref->linesize[0] + s->start_x;
1046  for (x = 0; x < s->w; x+=80)
1047  dst[x] = 200;
1048  dst = s->outpicref->data[0] + (s->h + s->start_y + 1) * s->outpicref->linesize[0] + s->start_x;
1049  for (x = 0; x < s->w; x+=40) {
1050  dst[x] = 200;
1051  }
1052  dst = s->outpicref->data[0] + (s->h + s->start_y + 2) * s->outpicref->linesize[0] + s->start_x;
1053  for (x = 0; x < s->w; x+=80) {
1054  dst[x] = 200;
1055  }
1056  for (y = 0; y < h; y += 40) {
1057  float hz = y * (inlink->sample_rate / 2) / (float)(1 << (int)ceil(log2(h)));
1058  char *units;
1059 
1060  if (hz == 0)
1061  units = av_asprintf("DC");
1062  else
1063  units = av_asprintf("%.2f", hz);
1064  if (!units)
1065  return AVERROR(ENOMEM);
1066 
1067  drawtext(s->outpicref, s->start_x - 8 * strlen(units) - 4, h * (ch + 1) + s->start_y - y - 4, units, 0);
1068  av_free(units);
1069  }
1070  }
1071 
1072  for (x = 0; x < s->w; x+=80) {
1073  float seconds = x * spp / inlink->sample_rate;
1074  char *units;
1075 
1076  if (x == 0)
1077  units = av_asprintf("0");
1078  else if (log10(seconds) > 6)
1079  units = av_asprintf("%.2fh", seconds / (60 * 60));
1080  else if (log10(seconds) > 3)
1081  units = av_asprintf("%.2fm", seconds / 60);
1082  else
1083  units = av_asprintf("%.2fs", seconds);
1084  if (!units)
1085  return AVERROR(ENOMEM);
1086 
1087  drawtext(s->outpicref, s->start_x + x - 4 * strlen(units), s->h + s->start_y + 6, units, 0);
1088  drawtext(s->outpicref, s->start_x + x - 4 * strlen(units), s->start_y - 12, units, 0);
1089  av_free(units);
1090  }
1091 
1092  drawtext(s->outpicref, outlink->w / 2 - 4 * 4, outlink->h - s->start_y / 2, "TIME", 0);
1093  drawtext(s->outpicref, s->start_x / 7, outlink->h / 2 - 14 * 4, "FREQUENCY (Hz)", 1);
1094  } else {
1095  int w = s->mode == SEPARATE ? s->w / s->nb_display_channels : s->w;
1096  for (y = 0; y < s->h; y += 20) {
1097  dst = s->outpicref->data[0] + (s->start_y + y) * s->outpicref->linesize[0];
1098  dst[s->start_x - 2] = 200;
1099  dst[s->start_x + s->w + 1] = 200;
1100  }
1101  for (y = 0; y < s->h; y += 40) {
1102  dst = s->outpicref->data[0] + (s->start_y + y) * s->outpicref->linesize[0];
1103  dst[s->start_x - 3] = 200;
1104  dst[s->start_x + s->w + 2] = 200;
1105  }
1106  for (ch = 0; ch < (s->mode == SEPARATE ? s->nb_display_channels : 1); ch++) {
1107  dst = s->outpicref->data[0] + (s->start_y - 2) * s->outpicref->linesize[0] + s->start_x + w * ch;
1108  for (x = 0; x < w; x+=40)
1109  dst[x] = 200;
1110  dst = s->outpicref->data[0] + (s->start_y - 3) * s->outpicref->linesize[0] + s->start_x + w * ch;
1111  for (x = 0; x < w; x+=80)
1112  dst[x] = 200;
1113  dst = s->outpicref->data[0] + (s->h + s->start_y + 1) * s->outpicref->linesize[0] + s->start_x + w * ch;
1114  for (x = 0; x < w; x+=40) {
1115  dst[x] = 200;
1116  }
1117  dst = s->outpicref->data[0] + (s->h + s->start_y + 2) * s->outpicref->linesize[0] + s->start_x + w * ch;
1118  for (x = 0; x < w; x+=80) {
1119  dst[x] = 200;
1120  }
1121  for (x = 0; x < w; x += 80) {
1122  float hz = x * (inlink->sample_rate / 2) / (float)(1 << (int)ceil(log2(w)));
1123  char *units;
1124 
1125  if (hz == 0)
1126  units = av_asprintf("DC");
1127  else
1128  units = av_asprintf("%.2f", hz);
1129  if (!units)
1130  return AVERROR(ENOMEM);
1131 
1132  drawtext(s->outpicref, s->start_x - 4 * strlen(units) + x + w * ch, s->start_y - 12, units, 0);
1133  drawtext(s->outpicref, s->start_x - 4 * strlen(units) + x + w * ch, s->h + s->start_y + 6, units, 0);
1134  av_free(units);
1135  }
1136  }
1137  for (y = 0; y < s->h; y+=40) {
1138  float seconds = y * spp / inlink->sample_rate;
1139  char *units;
1140 
1141  if (x == 0)
1142  units = av_asprintf("0");
1143  else if (log10(seconds) > 6)
1144  units = av_asprintf("%.2fh", seconds / (60 * 60));
1145  else if (log10(seconds) > 3)
1146  units = av_asprintf("%.2fm", seconds / 60);
1147  else
1148  units = av_asprintf("%.2fs", seconds);
1149  if (!units)
1150  return AVERROR(ENOMEM);
1151 
1152  drawtext(s->outpicref, s->start_x - 8 * strlen(units) - 4, s->start_y + y - 4, units, 0);
1153  av_free(units);
1154  }
1155  drawtext(s->outpicref, s->start_x / 7, outlink->h / 2 - 4 * 4, "TIME", 1);
1156  drawtext(s->outpicref, outlink->w / 2 - 14 * 4, outlink->h - s->start_y / 2, "FREQUENCY (Hz)", 0);
1157  }
1158 
1159  for (ch = 0; ch < (multi ? s->nb_display_channels : 1); ch++) {
1160  int h = multi ? s->h / s->nb_display_channels : s->h;
1161 
1162  for (y = 0; y < h; y++) {
1163  float out[3] = { 0., 127.5, 127.5};
1164  int chn;
1165 
1166  for (chn = 0; chn < (s->mode == SEPARATE ? 1 : s->nb_display_channels); chn++) {
1167  float yf, uf, vf;
1168  int channel = (multi) ? s->nb_display_channels - ch - 1 : chn;
1169 
1170  color_range(s, channel, &yf, &uf, &vf);
1171  pick_color(s, yf, uf, vf, y / (float)h, out);
1172  }
1173  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);
1174  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);
1175  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);
1176  }
1177  }
1178  }
1179 
1180  ret = ff_filter_frame(outlink, s->outpicref);
1181  s->outpicref = NULL;
1182  }
1183 
1184  return ret;
1185 }
1186 
1187 static int showspectrumpic_filter_frame(AVFilterLink *inlink, AVFrame *insamples)
1188 {
1189  AVFilterContext *ctx = inlink->dst;
1190  ShowSpectrumContext *s = ctx->priv;
1191  int ret;
1192 
1193  ret = av_audio_fifo_write(s->fifo, (void **)insamples->extended_data, insamples->nb_samples);
1194  av_frame_free(&insamples);
1195  return ret;
1196 }
1197 
1198 static const AVFilterPad showspectrumpic_inputs[] = {
1199  {
1200  .name = "default",
1201  .type = AVMEDIA_TYPE_AUDIO,
1202  .filter_frame = showspectrumpic_filter_frame,
1203  },
1204  { NULL }
1205 };
1206 
1207 static const AVFilterPad showspectrumpic_outputs[] = {
1208  {
1209  .name = "default",
1210  .type = AVMEDIA_TYPE_VIDEO,
1211  .config_props = config_output,
1212  .request_frame = showspectrumpic_request_frame,
1213  },
1214  { NULL }
1215 };
1216 
1217 AVFilter ff_avf_showspectrumpic = {
1218  .name = "showspectrumpic",
1219  .description = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output single picture."),
1220  .uninit = uninit,
1221  .query_formats = query_formats,
1222  .priv_size = sizeof(ShowSpectrumContext),
1223  .inputs = showspectrumpic_inputs,
1224  .outputs = showspectrumpic_outputs,
1225  .priv_class = &showspectrumpic_class,
1226 };
1227 
1228 #endif // CONFIG_SHOWSPECTRUMPIC_FILTER
int plane
Definition: avisynth_c.h:291
float, planar
Definition: samplefmt.h:70
#define NULL
Definition: coverity.c:32
const char * s
Definition: avisynth_c.h:631
AVAudioFifo * av_audio_fifo_alloc(enum AVSampleFormat sample_fmt, int channels, int nb_samples)
Allocate an AVAudioFifo.
Definition: audio_fifo.c:60
#define av_realloc_f(p, o, n)
This structure describes decoded (raw) audio or video data.
Definition: frame.h:181
AVOption.
Definition: opt.h:245
ptrdiff_t const GLvoid * data
Definition: opengl_enc.c:101
av_cold void av_fft_end(FFTContext *s)
Definition: avfft.c:48
AVFormatContext * ctx
Definition: movenc-test.c:48
planar YUV 4:4:4, 24bpp, (1 Cr & Cb sample per 1x1 Y samples)
Definition: pixfmt.h:68
Main libavfilter public API header.
FFTComplex ** fft_data
bins holder for each (displayed) channels
void av_audio_fifo_free(AVAudioFifo *af)
Free an AVAudioFifo.
Definition: audio_fifo.c:45
static void calc_phases(ShowSpectrumContext *s)
static void run_fft(ShowSpectrumContext *s, AVFrame *fin)
float * window_func_lut
Window function LUT.
FFTSample re
Definition: avfft.h:38
color_range
void av_fft_permute(FFTContext *s, FFTComplex *z)
Do the permutation needed BEFORE calling ff_fft_calc().
Definition: avfft.c:38
static enum AVSampleFormat formats[]
int fft_bits
number of bits (FFT window size = 1<<fft_bits)
static int query_formats(AVFilterContext *ctx)
AVFrame * ff_get_video_buffer(AVFilterLink *link, int w, int h)
Request a picture buffer with a specific set of permissions.
Definition: video.c:76
int sliding
1 if sliding mode, 0 otherwise
static AVRational av_make_q(int num, int den)
Create a rational.
Definition: rational.h:53
int mode
channel display mode
#define log2(x)
Definition: libm.h:404
AVFilterFormats * ff_make_format_list(const int *fmts)
Create a list of supported formats.
Definition: formats.c:283
DisplayScale
const char * name
Pad name.
Definition: internal.h:59
AVFilterLink ** inputs
array of pointers to input links
Definition: avfilter.h:312
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Definition: avassert.h:37
int ff_channel_layouts_ref(AVFilterChannelLayouts *f, AVFilterChannelLayouts **ref)
Add *ref as a new reference to f.
Definition: formats.c:435
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:1163
uint8_t
#define av_cold
Definition: attributes.h:82
mode
Definition: f_perms.c:27
static const AVOption showspectrum_options[]
AVOptions.
static av_cold int end(AVCodecContext *avctx)
Definition: avrndec.c:90
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:262
#define AVERROR_EOF
End of file.
Definition: error.h:55
#define AV_LOG_VERBOSE
Detailed information.
Definition: log.h:192
#define lrintf(x)
Definition: libm_mips.h:70
ptrdiff_t size
Definition: opengl_enc.c:101
#define av_log(a,...)
SlideMode
#define cm
Definition: dvbsubdec.c:36
void av_frame_set_color_range(AVFrame *frame, enum AVColorRange val)
A filter pad used for either input or output.
Definition: internal.h:53
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:176
#define PHASE(y, ch)
const uint8_t avpriv_cga_font[2048]
Definition: xga_font_data.c:29
AVFrame * ff_get_audio_buffer(AVFilterLink *link, int nb_samples)
Request an audio samples buffer with a specific set of permissions.
Definition: audio.c:65
static const uint16_t mask[17]
Definition: lzw.c:38
#define AVERROR(e)
Definition: error.h:43
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:154
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:176
void * priv
private data for use by the filter
Definition: avfilter.h:319
float saturation
color saturation multiplier
#define cbrt
Definition: tablegen.h:35
float * combine_buffer
color combining buffer (3 * h items)
simple assert() macros that are a bit more flexible than ISO C assert().
FFTContext * av_fft_init(int nbits, int inverse)
Set up a complex FFT.
Definition: avfft.c:28
#define fail()
Definition: checkasm.h:80
Context for an Audio FIFO Buffer.
Definition: audio_fifo.c:34
char * av_asprintf(const char *fmt,...)
Definition: avstring.c:113
int av_audio_fifo_size(AVAudioFifo *af)
Get the current number of samples in the AVAudioFifo available for reading.
Definition: audio_fifo.c:205
Definition: fft.h:88
audio channel layout utility functions
static int request_frame(AVFilterLink *outlink)
Definition: aeval.c:274
int ff_formats_ref(AVFilterFormats *f, AVFilterFormats **ref)
Add *ref as a new reference to formats.
Definition: formats.c:440
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
Definition: aeval.c:413
ColorMode
static void acalc_magnitudes(ShowSpectrumContext *s)
#define FLAGS
int n
Definition: avisynth_c.h:547
#define MAGNITUDE(y, ch)
static const AVFilterPad outputs[]
Definition: af_afftfilt.c:385
AVFilterChannelLayouts * ff_all_channel_layouts(void)
Construct an empty AVFilterChannelLayouts/AVFilterFormats struct – representing any channel layout (w...
Definition: formats.c:401
AVFrame * av_frame_clone(const AVFrame *src)
Create a new frame that references the same data as src.
Definition: frame.c:461
#define FF_ARRAY_ELEMS(a)
the normal 2^n-1 "JPEG" YUV ranges
Definition: pixfmt.h:443
A list of supported channel layouts.
Definition: formats.h:85
FILE * out
Definition: movenc-test.c:54
Orientation
static const AVFilterPad inputs[]
Definition: af_afftfilt.c:375
static int config_output(AVFilterLink *outlink)
static void color_range(ShowSpectrumContext *s, int ch, float *yf, float *uf, float *vf)
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:59
static const struct ColorTable color_table[][8]
int linesize[AV_NUM_DATA_POINTERS]
For video, size in bytes of each picture line.
Definition: frame.h:209
FFT functions.
Describe the class of an AVClass context structure.
Definition: log.h:67
Filter definition.
Definition: avfilter.h:141
void ff_generate_window_func(float *lut, int N, int win_func, float *overlap)
Definition: window_func.c:26
AVFILTER_DEFINE_CLASS(showspectrum)
rational number numerator/denominator
Definition: rational.h:43
const char * name
Filter name.
Definition: avfilter.h:145
offset must point to two consecutive integers
Definition: opt.h:232
DataMode
AVFilterLink ** outputs
array of pointers to output links
Definition: avfilter.h:316
enum MovChannelLayoutTag * layouts
Definition: mov_chan.c:434
static enum AVPixelFormat pix_fmts[]
Definition: libkvazaar.c:262
AVFilterFormats * ff_all_samplerates(void)
Definition: formats.c:395
void * av_calloc(size_t nmemb, size_t size)
Allocate a block of nmemb * size bytes with alignment suitable for all memory accesses (including vec...
Definition: mem.c:260
int av_frame_make_writable(AVFrame *frame)
Ensure that the frame data is writable, avoiding data copy if possible.
Definition: frame.c:522
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:192
int av_audio_fifo_write(AVAudioFifo *af, void **data, int nb_samples)
Write data to an AVAudioFifo.
Definition: audio_fifo.c:113
static av_cold void uninit(AVFilterContext *ctx)
int av_audio_fifo_drain(AVAudioFifo *af, int nb_samples)
Drain data from an AVAudioFifo.
Definition: audio_fifo.c:178
static int plot_spectrum_column(AVFilterLink *inlink, AVFrame *insamples)
FFTSample im
Definition: avfft.h:38
if(ret< 0)
Definition: vf_mcdeint.c:282
planar YUV 4:4:4, 24bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV444P and setting col...
Definition: pixfmt.h:77
int den
denominator
Definition: rational.h:45
#define av_free(p)
FFTContext * fft
Fast Fourier Transform context.
Audio FIFO Buffer.
A list of supported formats for one end of a filter link.
Definition: formats.h:64
int av_audio_fifo_peek(AVAudioFifo *af, void **data, int nb_samples)
Peek data from an AVAudioFifo.
Definition: audio_fifo.c:139
An instance of a filter.
Definition: avfilter.h:304
static enum AVSampleFormat sample_fmts[]
Definition: adpcmenc.c:701
#define OFFSET(x)
#define av_freep(p)
void INT64 start
Definition: avisynth_c.h:553
#define M_PI
Definition: mathematics.h:46
static void scale_magnitudes(ShowSpectrumContext *s, float scale)
int ff_request_frame(AVFilterLink *link)
Request an input frame from the filter at the other end of the link.
Definition: avfilter.c:356
static void pick_color(ShowSpectrumContext *s, float yf, float uf, float vf, float a, float *out)
internal API functions
uint8_t ** extended_data
pointers to the data planes/channels.
Definition: frame.h:225
static void drawtext(AVFrame *pic, int x, int y, const char *txt)
int xpos
x position (current column)
void av_fft_calc(FFTContext *s, FFTComplex *z)
Do a complex FFT with the parameters defined in av_fft_init().
Definition: avfft.c:43
AVPixelFormat
Pixel format.
Definition: pixfmt.h:61
static void clear_combine_buffer(ShowSpectrumContext *s, int size)
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:235
const AVFilter * filter
the AVFilter of which this is an instance
Definition: avfilter.h:307
for(j=16;j >0;--j)
static void calc_magnitudes(ShowSpectrumContext *s)
#define AV_NOPTS_VALUE
Undefined timestamp value.
Definition: avutil.h:240
CGA/EGA/VGA ROM font data.
int color_mode
display color scheme
DisplayMode