2 * Copyright (c) 2012 Clément Bœsch
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * audio to spectrum (video) transmedia filter, based on ffplay rdft showmode
24 * (by Michael Niedermayer) and lavfi/avf_showwaves (by Stefano Sabatini).
29 #include "libavcodec/avfft.h"
30 #include "libavutil/avassert.h"
31 #include "libavutil/channel_layout.h"
32 #include "libavutil/opt.h"
36 enum DisplayMode { COMBINED, SEPARATE, NB_MODES };
37 enum DisplayScale { LINEAR, SQRT, CBRT, LOG, NB_SCALES };
38 enum ColorMode { CHANNEL, INTENSITY, NB_CLMODES };
43 AVFilterBufferRef *outpicref;
45 int nb_display_channels;
47 int sliding; ///< 1 if sliding mode, 0 otherwise
48 enum DisplayMode mode; ///< channel display mode
49 enum ColorMode color_mode; ///< display color scheme
50 enum DisplayScale scale;
51 float saturation; ///< color saturation multiplier
52 int xpos; ///< x position (current column)
53 RDFTContext *rdft; ///< Real Discrete Fourier Transform context
54 int rdft_bits; ///< number of bits (RDFT window size = 1<<rdft_bits)
55 FFTSample **rdft_data; ///< bins holder for each (displayed) channels
56 int filled; ///< number of samples (per channel) filled in current rdft_buffer
57 int consumed; ///< number of samples (per channel) consumed from the input frame
58 float *window_func_lut; ///< Window function LUT
59 float *combine_buffer; ///< color combining buffer (3 * h items)
60 } ShowSpectrumContext;
62 #define OFFSET(x) offsetof(ShowSpectrumContext, x)
63 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
65 static const AVOption showspectrum_options[] = {
66 { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },
67 { "s", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },
68 { "slide", "set sliding mode", OFFSET(sliding), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS },
69 { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, COMBINED, NB_MODES-1, FLAGS, "mode" },
70 { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },
71 { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },
72 { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=CHANNEL}, CHANNEL, NB_CLMODES-1, FLAGS, "color" },
73 { "channel", "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL}, 0, 0, FLAGS, "color" },
74 { "intensity", "intensity based coloring", 0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },
75 { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=SQRT}, LINEAR, NB_SCALES-1, FLAGS, "scale" },
76 { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT}, 0, 0, FLAGS, "scale" },
77 { "cbrt", "cubic root", 0, AV_OPT_TYPE_CONST, {.i64=CBRT}, 0, 0, FLAGS, "scale" },
78 { "log", "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG}, 0, 0, FLAGS, "scale" },
79 { "lin", "linear", 0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },
80 { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },
84 AVFILTER_DEFINE_CLASS(showspectrum);
88 } intensity_color_table_item;
89 static const intensity_color_table_item intensity_color_table[] =
92 { 0.13, .03587126228984074, .1573300977624594, -.02548747583751842 },
93 { 0.3, .1857228179456802, .1772436246393981, .1747555484041475 },
94 { 0.6, .2818498058365613, -.1593064119945782, .4713207455460892 },
95 { 0.73, .6583062117554781, -.3716070802232764, .2435275933125293 },
96 { 0.78, 0.763185357582429, -.4307467689263783, .1686649662231043 },
97 { 0.91, .9533636363636364, -.2045454545454546, .03313636363636363 },
101 static av_cold int init(AVFilterContext *ctx, const char *args)
103 ShowSpectrumContext *showspectrum = ctx->priv;
106 showspectrum->class = &showspectrum_class;
107 av_opt_set_defaults(showspectrum);
109 if ((err = av_set_options_string(showspectrum, args, "=", ":")) < 0)
115 static av_cold void uninit(AVFilterContext *ctx)
117 ShowSpectrumContext *showspectrum = ctx->priv;
120 av_freep(&showspectrum->combine_buffer);
121 av_rdft_end(showspectrum->rdft);
122 for (i = 0; i < showspectrum->nb_display_channels; i++)
123 av_freep(&showspectrum->rdft_data[i]);
124 av_freep(&showspectrum->rdft_data);
125 av_freep(&showspectrum->window_func_lut);
126 avfilter_unref_bufferp(&showspectrum->outpicref);
129 static int query_formats(AVFilterContext *ctx)
131 AVFilterFormats *formats = NULL;
132 AVFilterChannelLayouts *layouts = NULL;
133 AVFilterLink *inlink = ctx->inputs[0];
134 AVFilterLink *outlink = ctx->outputs[0];
135 static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_NONE };
136 static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_NONE };
138 /* set input audio formats */
139 formats = ff_make_format_list(sample_fmts);
141 return AVERROR(ENOMEM);
142 ff_formats_ref(formats, &inlink->out_formats);
144 layouts = ff_all_channel_layouts();
146 return AVERROR(ENOMEM);
147 ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts);
149 formats = ff_all_samplerates();
151 return AVERROR(ENOMEM);
152 ff_formats_ref(formats, &inlink->out_samplerates);
154 /* set output video format */
155 formats = ff_make_format_list(pix_fmts);
157 return AVERROR(ENOMEM);
158 ff_formats_ref(formats, &outlink->in_formats);
163 static int config_output(AVFilterLink *outlink)
165 AVFilterContext *ctx = outlink->src;
166 AVFilterLink *inlink = ctx->inputs[0];
167 ShowSpectrumContext *showspectrum = ctx->priv;
168 int i, rdft_bits, win_size, h;
170 outlink->w = showspectrum->w;
171 outlink->h = showspectrum->h;
173 h = (showspectrum->mode == COMBINED) ? outlink->h : outlink->h / inlink->channels;
174 showspectrum->channel_height = h;
176 /* RDFT window size (precision) according to the requested output frame height */
177 for (rdft_bits = 1; 1 << rdft_bits < 2 * h; rdft_bits++);
178 win_size = 1 << rdft_bits;
180 /* (re-)configuration if the video output changed (or first init) */
181 if (rdft_bits != showspectrum->rdft_bits) {
182 size_t rdft_size, rdft_listsize;
183 AVFilterBufferRef *outpicref;
185 av_rdft_end(showspectrum->rdft);
186 showspectrum->rdft = av_rdft_init(rdft_bits, DFT_R2C);
187 showspectrum->rdft_bits = rdft_bits;
189 /* RDFT buffers: x2 for each (display) channel buffer.
190 * Note: we use free and malloc instead of a realloc-like function to
191 * make sure the buffer is aligned in memory for the FFT functions. */
192 for (i = 0; i < showspectrum->nb_display_channels; i++)
193 av_freep(&showspectrum->rdft_data[i]);
194 av_freep(&showspectrum->rdft_data);
195 showspectrum->nb_display_channels = inlink->channels;
197 if (av_size_mult(sizeof(*showspectrum->rdft_data),
198 showspectrum->nb_display_channels, &rdft_listsize) < 0)
199 return AVERROR(EINVAL);
200 if (av_size_mult(sizeof(**showspectrum->rdft_data),
201 win_size, &rdft_size) < 0)
202 return AVERROR(EINVAL);
203 showspectrum->rdft_data = av_malloc(rdft_listsize);
204 if (!showspectrum->rdft_data)
205 return AVERROR(ENOMEM);
206 for (i = 0; i < showspectrum->nb_display_channels; i++) {
207 showspectrum->rdft_data[i] = av_malloc(rdft_size);
208 if (!showspectrum->rdft_data[i])
209 return AVERROR(ENOMEM);
211 showspectrum->filled = 0;
213 /* pre-calc windowing function (hann here) */
214 showspectrum->window_func_lut =
215 av_realloc_f(showspectrum->window_func_lut, win_size,
216 sizeof(*showspectrum->window_func_lut));
217 if (!showspectrum->window_func_lut)
218 return AVERROR(ENOMEM);
219 for (i = 0; i < win_size; i++)
220 showspectrum->window_func_lut[i] = .5f * (1 - cos(2*M_PI*i / (win_size-1)));
222 /* prepare the initial picref buffer (black frame) */
223 avfilter_unref_bufferp(&showspectrum->outpicref);
224 showspectrum->outpicref = outpicref =
225 ff_get_video_buffer(outlink, AV_PERM_WRITE|AV_PERM_PRESERVE|AV_PERM_REUSE2,
226 outlink->w, outlink->h);
228 return AVERROR(ENOMEM);
229 outlink->sample_aspect_ratio = (AVRational){1,1};
230 memset(outpicref->data[0], 0, outlink->h * outpicref->linesize[0]);
233 if (showspectrum->xpos >= outlink->w)
234 showspectrum->xpos = 0;
236 showspectrum->combine_buffer =
237 av_realloc_f(showspectrum->combine_buffer, outlink->h * 3,
238 sizeof(*showspectrum->combine_buffer));
240 av_log(ctx, AV_LOG_VERBOSE, "s:%dx%d RDFT window size:%d\n",
241 showspectrum->w, showspectrum->h, win_size);
245 inline static void push_frame(AVFilterLink *outlink)
247 ShowSpectrumContext *showspectrum = outlink->src->priv;
249 showspectrum->xpos++;
250 if (showspectrum->xpos >= outlink->w)
251 showspectrum->xpos = 0;
252 showspectrum->filled = 0;
253 showspectrum->req_fullfilled = 1;
255 ff_filter_frame(outlink, avfilter_ref_buffer(showspectrum->outpicref, ~AV_PERM_WRITE));
258 static int request_frame(AVFilterLink *outlink)
260 ShowSpectrumContext *showspectrum = outlink->src->priv;
261 AVFilterLink *inlink = outlink->src->inputs[0];
264 showspectrum->req_fullfilled = 0;
266 ret = ff_request_frame(inlink);
267 } while (!showspectrum->req_fullfilled && ret >= 0);
269 if (ret == AVERROR_EOF && showspectrum->outpicref)
274 static int plot_spectrum_column(AVFilterLink *inlink, AVFilterBufferRef *insamples, int nb_samples)
276 AVFilterContext *ctx = inlink->dst;
277 AVFilterLink *outlink = ctx->outputs[0];
278 ShowSpectrumContext *showspectrum = ctx->priv;
279 AVFilterBufferRef *outpicref = showspectrum->outpicref;
281 /* nb_freq contains the power of two superior or equal to the output image
282 * height (or half the RDFT window size) */
283 const int nb_freq = 1 << (showspectrum->rdft_bits - 1);
284 const int win_size = nb_freq << 1;
285 const double w = 1. / (sqrt(nb_freq) * 32768.);
288 const int start = showspectrum->filled;
289 const int add_samples = FFMIN(win_size - start, nb_samples);
291 /* fill RDFT input with the number of samples available */
292 for (ch = 0; ch < showspectrum->nb_display_channels; ch++) {
293 const int16_t *p = (int16_t *)insamples->extended_data[ch];
295 p += showspectrum->consumed;
296 for (n = 0; n < add_samples; n++)
297 showspectrum->rdft_data[ch][start + n] = p[n] * showspectrum->window_func_lut[start + n];
299 showspectrum->filled += add_samples;
301 /* complete RDFT window size? */
302 if (showspectrum->filled == win_size) {
305 int h = showspectrum->channel_height;
307 /* run RDFT on each samples set */
308 for (ch = 0; ch < showspectrum->nb_display_channels; ch++)
309 av_rdft_calc(showspectrum->rdft, showspectrum->rdft_data[ch]);
311 /* fill a new spectrum column */
312 #define RE(y, ch) showspectrum->rdft_data[ch][2 * y + 0]
313 #define IM(y, ch) showspectrum->rdft_data[ch][2 * y + 1]
314 #define MAGNITUDE(y, ch) hypot(RE(y, ch), IM(y, ch))
316 /* initialize buffer for combining to black */
317 for (y = 0; y < outlink->h; y++) {
318 showspectrum->combine_buffer[3 * y ] = 0;
319 showspectrum->combine_buffer[3 * y + 1] = 127.5;
320 showspectrum->combine_buffer[3 * y + 2] = 127.5;
323 for (ch = 0; ch < showspectrum->nb_display_channels; ch++) {
326 /* decide color range */
327 switch (showspectrum->mode) {
329 // reduce range by channel count
330 yf = 256.0f / showspectrum->nb_display_channels;
331 switch (showspectrum->color_mode) {
337 /* adjust saturation for mixed UV coloring */
338 /* this factor is correct for infinite channels, an approximation otherwise */
356 if (showspectrum->color_mode == CHANNEL) {
357 if (showspectrum->nb_display_channels > 1) {
358 uf *= 0.5 * sin((2 * M_PI * ch) / showspectrum->nb_display_channels);
359 vf *= 0.5 * cos((2 * M_PI * ch) / showspectrum->nb_display_channels);
365 uf *= showspectrum->saturation;
366 vf *= showspectrum->saturation;
368 /* draw the channel */
369 for (y = 0; y < h; y++) {
370 int row = (showspectrum->mode == COMBINED) ? y : ch * h + y;
371 float *out = &showspectrum->combine_buffer[3 * row];
374 float a = w * MAGNITUDE(y, ch);
377 switch (showspectrum->scale) {
387 a = 1 - log(FFMAX(FFMIN(1, a), 1e-6)) / log(1e-6); // zero = -120dBFS
393 if (showspectrum->color_mode == INTENSITY) {
397 for (i = 1; i < sizeof(intensity_color_table) / sizeof(*intensity_color_table) - 1; i++)
398 if (intensity_color_table[i].a >= a)
400 // i now is the first item >= the color
401 // now we know to interpolate between item i - 1 and i
402 if (a <= intensity_color_table[i - 1].a) {
403 y = intensity_color_table[i - 1].y;
404 u = intensity_color_table[i - 1].u;
405 v = intensity_color_table[i - 1].v;
406 } else if (a >= intensity_color_table[i].a) {
407 y = intensity_color_table[i].y;
408 u = intensity_color_table[i].u;
409 v = intensity_color_table[i].v;
411 float start = intensity_color_table[i - 1].a;
412 float end = intensity_color_table[i].a;
413 float lerpfrac = (a - start) / (end - start);
414 y = intensity_color_table[i - 1].y * (1.0f - lerpfrac)
415 + intensity_color_table[i].y * lerpfrac;
416 u = intensity_color_table[i - 1].u * (1.0f - lerpfrac)
417 + intensity_color_table[i].u * lerpfrac;
418 v = intensity_color_table[i - 1].v * (1.0f - lerpfrac)
419 + intensity_color_table[i].v * lerpfrac;
434 if (showspectrum->sliding) {
435 for (plane = 0; plane < 3; plane++) {
436 for (y = 0; y < outlink->h; y++) {
437 uint8_t *p = outpicref->data[plane] +
438 y * outpicref->linesize[plane];
439 memmove(p, p + 1, outlink->w - 1);
442 showspectrum->xpos = outlink->w - 1;
444 for (plane = 0; plane < 3; plane++) {
445 uint8_t *p = outpicref->data[plane] +
446 (outlink->h - 1) * outpicref->linesize[plane] +
448 for (y = 0; y < outlink->h; y++) {
449 *p = rint(FFMAX(0, FFMIN(showspectrum->combine_buffer[3 * y + plane], 255)));
450 p -= outpicref->linesize[plane];
454 outpicref->pts = insamples->pts +
455 av_rescale_q(showspectrum->consumed,
456 (AVRational){ 1, inlink->sample_rate },
464 static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *insamples)
466 AVFilterContext *ctx = inlink->dst;
467 ShowSpectrumContext *showspectrum = ctx->priv;
468 int left_samples = insamples->audio->nb_samples;
470 showspectrum->consumed = 0;
471 while (left_samples) {
472 const int added_samples = plot_spectrum_column(inlink, insamples, left_samples);
473 showspectrum->consumed += added_samples;
474 left_samples -= added_samples;
477 avfilter_unref_buffer(insamples);
481 static const AVFilterPad showspectrum_inputs[] = {
484 .type = AVMEDIA_TYPE_AUDIO,
485 .filter_frame = filter_frame,
486 .min_perms = AV_PERM_READ,
491 static const AVFilterPad showspectrum_outputs[] = {
494 .type = AVMEDIA_TYPE_VIDEO,
495 .config_props = config_output,
496 .request_frame = request_frame,
501 AVFilter avfilter_avf_showspectrum = {
502 .name = "showspectrum",
503 .description = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output."),
506 .query_formats = query_formats,
507 .priv_size = sizeof(ShowSpectrumContext),
508 .inputs = showspectrum_inputs,
509 .outputs = showspectrum_outputs,
510 .priv_class = &showspectrum_class,