2 * Copyright (c) 2014 Muhammad Faiz <mfcc64@gmail.com>
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
22 #include "libavcodec/avfft.h"
23 #include "libavutil/avassert.h"
24 #include "libavutil/channel_layout.h"
25 #include "libavutil/opt.h"
26 #include "libavutil/xga_font_data.h"
27 #include "libavutil/qsort.h"
28 #include "libavutil/time.h"
29 #include "libavutil/eval.h"
36 #if CONFIG_LIBFREETYPE
38 #include FT_FREETYPE_H
41 /* this filter is designed to do 16 bins/semitones constant Q transform with Brown-Puckette algorithm
42 * start from E0 to D#10 (10 octaves)
43 * so there are 16 bins/semitones * 12 semitones/octaves * 10 octaves = 1920 bins
44 * match with full HD resolution */
46 #define VIDEO_WIDTH 1920
47 #define VIDEO_HEIGHT 1080
48 #define FONT_HEIGHT 32
49 #define SPECTOGRAM_HEIGHT ((VIDEO_HEIGHT-FONT_HEIGHT)/2)
50 #define SPECTOGRAM_START (VIDEO_HEIGHT-SPECTOGRAM_HEIGHT)
51 #define BASE_FREQ 20.051392800492
52 #define COEFF_CLAMP 1.0e-4
53 #define TLENGTH_MIN 0.001
54 #define TLENGTH_DEFAULT "384/f*tc/(384/f+tc)"
55 #define VOLUME_MIN 1e-10
56 #define VOLUME_MAX 100.0
66 FFTContext *fft_context;
68 FFTComplex *fft_result_left;
69 FFTComplex *fft_result_right;
71 SparseCoeff *coeff_sort;
72 SparseCoeff *coeffs[VIDEO_WIDTH];
74 char *fontfile; /* using freetype */
75 int coeffs_len[VIDEO_WIDTH];
76 uint8_t font_color[VIDEO_WIDTH];
85 double timeclamp; /* lower timeclamp, time-accurate, higher timeclamp, freq-accurate (at low freq)*/
86 float coeffclamp; /* lower coeffclamp, more precise, higher coeffclamp, faster */
87 int fullhd; /* if true, output video is at full HD resolution, otherwise it will be halved */
88 float gamma; /* lower gamma, more contrast, higher gamma, more range */
89 int fps; /* the required fps is so strict, so it's enough to be int, but 24000/1001 etc cannot be encoded */
90 int count; /* fps * count = transform rate */
93 #define OFFSET(x) offsetof(ShowCQTContext, x)
94 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
96 static const AVOption showcqt_options[] = {
97 { "volume", "set volume", OFFSET(volume), AV_OPT_TYPE_STRING, { .str = "16" }, CHAR_MIN, CHAR_MAX, FLAGS },
98 { "tlength", "set transform length", OFFSET(tlength), AV_OPT_TYPE_STRING, { .str = TLENGTH_DEFAULT }, CHAR_MIN, CHAR_MAX, FLAGS },
99 { "timeclamp", "set timeclamp", OFFSET(timeclamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.17 }, 0.1, 1.0, FLAGS },
100 { "coeffclamp", "set coeffclamp", OFFSET(coeffclamp), AV_OPT_TYPE_FLOAT, { .dbl = 1 }, 0.1, 10, FLAGS },
101 { "gamma", "set gamma", OFFSET(gamma), AV_OPT_TYPE_FLOAT, { .dbl = 3 }, 1, 7, FLAGS },
102 { "fullhd", "set full HD resolution", OFFSET(fullhd), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, FLAGS },
103 { "fps", "set video fps", OFFSET(fps), AV_OPT_TYPE_INT, { .i64 = 25 }, 10, 100, FLAGS },
104 { "count", "set number of transform per frame", OFFSET(count), AV_OPT_TYPE_INT, { .i64 = 6 }, 1, 30, FLAGS },
105 { "fontfile", "set font file", OFFSET(fontfile), AV_OPT_TYPE_STRING, { .str = NULL }, CHAR_MIN, CHAR_MAX, FLAGS },
109 AVFILTER_DEFINE_CLASS(showcqt);
111 static av_cold void uninit(AVFilterContext *ctx)
115 ShowCQTContext *s = ctx->priv;
116 av_fft_end(s->fft_context);
117 s->fft_context = NULL;
118 for (k = 0; k < VIDEO_WIDTH; k++)
119 av_freep(&s->coeffs[k]);
120 av_freep(&s->fft_data);
121 av_freep(&s->fft_result_left);
122 av_freep(&s->fft_result_right);
123 av_freep(&s->coeff_sort);
124 av_freep(&s->spectogram);
125 av_freep(&s->font_alpha);
126 av_frame_free(&s->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_FLT, AV_SAMPLE_FMT_NONE };
136 static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_RGB24, AV_PIX_FMT_NONE };
137 static const int64_t channel_layouts[] = { AV_CH_LAYOUT_STEREO, AV_CH_LAYOUT_STEREO_DOWNMIX, -1 };
138 static const int samplerates[] = { 44100, 48000, -1 };
140 /* set input audio formats */
141 formats = ff_make_format_list(sample_fmts);
143 return AVERROR(ENOMEM);
144 ff_formats_ref(formats, &inlink->out_formats);
146 layouts = avfilter_make_format64_list(channel_layouts);
148 return AVERROR(ENOMEM);
149 ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts);
151 formats = ff_make_format_list(samplerates);
153 return AVERROR(ENOMEM);
154 ff_formats_ref(formats, &inlink->out_samplerates);
156 /* set output video format */
157 formats = ff_make_format_list(pix_fmts);
159 return AVERROR(ENOMEM);
160 ff_formats_ref(formats, &outlink->in_formats);
165 #if CONFIG_LIBFREETYPE
166 static void load_freetype_font(AVFilterContext *ctx)
168 static const char str[] = "EF G A BC D ";
169 ShowCQTContext *s = ctx->priv;
170 FT_Library lib = NULL;
172 int video_scale = s->fullhd ? 2 : 1;
173 int video_width = (VIDEO_WIDTH/2) * video_scale;
174 int font_height = (FONT_HEIGHT/2) * video_scale;
175 int font_width = 8 * video_scale;
176 int font_repeat = font_width * 12;
177 int linear_hori_advance = font_width * 65536;
178 int non_monospace_warning = 0;
181 s->font_alpha = NULL;
186 if (FT_Init_FreeType(&lib))
189 if (FT_New_Face(lib, s->fontfile, 0, &face))
192 if (FT_Set_Char_Size(face, 16*64, 0, 0, 0))
195 if (FT_Load_Char(face, 'A', FT_LOAD_RENDER))
198 if (FT_Set_Char_Size(face, 16*64 * linear_hori_advance / face->glyph->linearHoriAdvance, 0, 0, 0))
201 s->font_alpha = av_malloc(font_height * video_width);
205 memset(s->font_alpha, 0, font_height * video_width);
207 for (x = 0; x < 12; x++) {
208 int sx, sy, rx, bx, by, dx, dy;
213 if (FT_Load_Char(face, str[x], FT_LOAD_RENDER))
216 if (face->glyph->advance.x != font_width*64 && !non_monospace_warning) {
217 av_log(ctx, AV_LOG_WARNING, "Font is not monospace\n");
218 non_monospace_warning = 1;
221 sy = font_height - 4*video_scale - face->glyph->bitmap_top;
222 for (rx = 0; rx < 10; rx++) {
223 sx = rx * font_repeat + x * font_width + face->glyph->bitmap_left;
224 for (by = 0; by < face->glyph->bitmap.rows; by++) {
228 if (dy >= font_height)
231 for (bx = 0; bx < face->glyph->bitmap.width; bx++) {
235 if (dx >= video_width)
237 s->font_alpha[dy*video_width+dx] = face->glyph->bitmap.buffer[by*face->glyph->bitmap.width+bx];
244 FT_Done_FreeType(lib);
248 av_log(ctx, AV_LOG_WARNING, "Error while loading freetype font, using default font instead\n");
250 FT_Done_FreeType(lib);
251 av_freep(&s->font_alpha);
256 static double a_weighting(void *p, double f)
258 double ret = 12200.0*12200.0 * (f*f*f*f);
259 ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0) *
260 sqrt((f*f + 107.7*107.7) * (f*f + 737.9*737.9));
264 static double b_weighting(void *p, double f)
266 double ret = 12200.0*12200.0 * (f*f*f);
267 ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0) * sqrt(f*f + 158.5*158.5);
271 static double c_weighting(void *p, double f)
273 double ret = 12200.0*12200.0 * (f*f);
274 ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0);
278 static inline int qsort_sparsecoeff(const SparseCoeff *a, const SparseCoeff *b)
280 if (fabsf(a->value) >= fabsf(b->value))
286 static int config_output(AVFilterLink *outlink)
288 AVFilterContext *ctx = outlink->src;
289 AVFilterLink *inlink = ctx->inputs[0];
290 ShowCQTContext *s = ctx->priv;
291 AVExpr *tlength_expr, *volume_expr;
292 static const char * const expr_vars[] = { "timeclamp", "tc", "frequency", "freq", "f", NULL };
293 static const char * const expr_func_names[] = { "a_weighting", "b_weighting", "c_weighting", NULL };
294 static double (* const expr_funcs[])(void *, double) = { a_weighting, b_weighting, c_weighting, NULL };
295 int fft_len, k, x, y, ret;
297 int rate = inlink->sample_rate;
298 double max_len = rate * (double) s->timeclamp;
299 int64_t start_time, end_time;
300 int video_scale = s->fullhd ? 2 : 1;
301 int video_width = (VIDEO_WIDTH/2) * video_scale;
302 int video_height = (VIDEO_HEIGHT/2) * video_scale;
303 int spectogram_height = (SPECTOGRAM_HEIGHT/2) * video_scale;
305 s->fft_bits = ceil(log2(max_len));
306 fft_len = 1 << s->fft_bits;
308 if (rate % (s->fps * s->count)) {
309 av_log(ctx, AV_LOG_ERROR, "Rate (%u) is not divisible by fps*count (%u*%u)\n", rate, s->fps, s->count);
310 return AVERROR(EINVAL);
313 s->fft_data = av_malloc_array(fft_len, sizeof(*s->fft_data));
314 s->coeff_sort = av_malloc_array(fft_len, sizeof(*s->coeff_sort));
315 s->fft_result_left = av_malloc_array(fft_len, sizeof(*s->fft_result_left));
316 s->fft_result_right = av_malloc_array(fft_len, sizeof(*s->fft_result_right));
317 s->fft_context = av_fft_init(s->fft_bits, 0);
319 if (!s->fft_data || !s->coeff_sort || !s->fft_result_left || !s->fft_result_right || !s->fft_context)
320 return AVERROR(ENOMEM);
322 /* initializing font */
323 for (x = 0; x < video_width; x++) {
324 if (x >= (12*3+8)*8*video_scale && x < (12*4+8)*8*video_scale) {
325 float fx = (x-(12*3+8)*8*video_scale) * (2.0f/(192.0f*video_scale));
326 float sv = sinf(M_PI*fx);
327 s->font_color[x] = sv*sv*255.0f + 0.5f;
329 s->font_color[x] = 0;
333 #if CONFIG_LIBFREETYPE
334 load_freetype_font(ctx);
337 av_log(ctx, AV_LOG_WARNING, "Freetype is not available, ignoring fontfile option\n");
338 s->font_alpha = NULL;
341 av_log(ctx, AV_LOG_INFO, "Calculating spectral kernel, please wait\n");
342 start_time = av_gettime_relative();
343 ret = av_expr_parse(&tlength_expr, s->tlength, expr_vars, NULL, NULL, NULL, NULL, 0, ctx);
346 ret = av_expr_parse(&volume_expr, s->volume, expr_vars, expr_func_names,
347 expr_funcs, NULL, NULL, 0, ctx);
349 av_expr_free(tlength_expr);
352 for (k = 0; k < VIDEO_WIDTH; k++) {
353 int hlen = fft_len >> 1;
356 double freq = BASE_FREQ * exp2(k * (1.0/192.0));
357 double tlen, tlength, volume;
358 double expr_vars_val[] = { s->timeclamp, s->timeclamp, freq, freq, freq, 0 };
359 /* a window function from Albert H. Nuttall,
360 * "Some Windows with Very Good Sidelobe Behavior"
361 * -93.32 dB peak sidelobe and 18 dB/octave asymptotic decay
362 * coefficient normalized to a0 = 1 */
363 double a0 = 0.355768;
364 double a1 = 0.487396/a0;
365 double a2 = 0.144232/a0;
366 double a3 = 0.012604/a0;
367 double sv_step, cv_step, sv, cv;
368 double sw_step, cw_step, sw, cw, w;
370 tlength = av_expr_eval(tlength_expr, expr_vars_val, NULL);
371 if (isnan(tlength)) {
372 av_log(ctx, AV_LOG_WARNING, "at freq %g: tlength is nan, setting it to %g\n", freq, s->timeclamp);
373 tlength = s->timeclamp;
374 } else if (tlength < TLENGTH_MIN) {
375 av_log(ctx, AV_LOG_WARNING, "at freq %g: tlength is %g, setting it to %g\n", freq, tlength, TLENGTH_MIN);
376 tlength = TLENGTH_MIN;
377 } else if (tlength > s->timeclamp) {
378 av_log(ctx, AV_LOG_WARNING, "at freq %g: tlength is %g, setting it to %g\n", freq, tlength, s->timeclamp);
379 tlength = s->timeclamp;
382 volume = FFABS(av_expr_eval(volume_expr, expr_vars_val, NULL));
384 av_log(ctx, AV_LOG_WARNING, "at freq %g: volume is nan, setting it to 0\n", freq);
386 } else if (volume < VOLUME_MIN) {
388 } else if (volume > VOLUME_MAX) {
389 av_log(ctx, AV_LOG_WARNING, "at freq %g: volume is %g, setting it to %g\n", freq, volume, VOLUME_MAX);
393 tlen = tlength * rate;
394 s->fft_data[0].re = 0;
395 s->fft_data[0].im = 0;
396 s->fft_data[hlen].re = (1.0 + a1 + a2 + a3) * (1.0/tlen) * volume * (1.0/fft_len);
397 s->fft_data[hlen].im = 0;
398 sv_step = sv = sin(2.0*M_PI*freq*(1.0/rate));
399 cv_step = cv = cos(2.0*M_PI*freq*(1.0/rate));
400 /* also optimizing window func */
401 sw_step = sw = sin(2.0*M_PI*(1.0/tlen));
402 cw_step = cw = cos(2.0*M_PI*(1.0/tlen));
403 for (x = 1; x < 0.5 * tlen; x++) {
404 double cv_tmp, cw_tmp;
405 double cw2, cw3, sw2;
407 cw2 = cw * cw - sw * sw;
408 sw2 = cw * sw + sw * cw;
409 cw3 = cw * cw2 - sw * sw2;
410 w = (1.0 + a1 * cw + a2 * cw2 + a3 * cw3) * (1.0/tlen) * volume * (1.0/fft_len);
411 s->fft_data[hlen + x].re = w * cv;
412 s->fft_data[hlen + x].im = w * sv;
413 s->fft_data[hlen - x].re = s->fft_data[hlen + x].re;
414 s->fft_data[hlen - x].im = -s->fft_data[hlen + x].im;
416 cv_tmp = cv * cv_step - sv * sv_step;
417 sv = sv * cv_step + cv * sv_step;
419 cw_tmp = cw * cw_step - sw * sw_step;
420 sw = sw * cw_step + cw * sw_step;
423 for (; x < hlen; x++) {
424 s->fft_data[hlen + x].re = 0;
425 s->fft_data[hlen + x].im = 0;
426 s->fft_data[hlen - x].re = 0;
427 s->fft_data[hlen - x].im = 0;
429 av_fft_permute(s->fft_context, s->fft_data);
430 av_fft_calc(s->fft_context, s->fft_data);
432 for (x = 0; x < fft_len; x++) {
433 s->coeff_sort[x].index = x;
434 s->coeff_sort[x].value = s->fft_data[x].re;
437 AV_QSORT(s->coeff_sort, fft_len, SparseCoeff, qsort_sparsecoeff);
438 for (x = 0; x < fft_len; x++)
439 total += fabsf(s->coeff_sort[x].value);
441 for (x = 0; x < fft_len; x++) {
442 partial += fabsf(s->coeff_sort[x].value);
443 if (partial > total * s->coeffclamp * COEFF_CLAMP) {
444 s->coeffs_len[k] = fft_len - x;
445 num_coeffs += s->coeffs_len[k];
446 s->coeffs[k] = av_malloc_array(s->coeffs_len[k], sizeof(*s->coeffs[k]));
448 av_expr_free(volume_expr);
449 av_expr_free(tlength_expr);
450 return AVERROR(ENOMEM);
452 for (y = 0; y < s->coeffs_len[k]; y++)
453 s->coeffs[k][y] = s->coeff_sort[x+y];
458 av_expr_free(volume_expr);
459 av_expr_free(tlength_expr);
460 end_time = av_gettime_relative();
461 av_log(ctx, AV_LOG_INFO, "Elapsed time %.6f s (fft_len=%u, num_coeffs=%u)\n", 1e-6 * (end_time-start_time), fft_len, num_coeffs);
463 outlink->w = video_width;
464 outlink->h = video_height;
466 s->req_fullfilled = 0;
467 s->spectogram_index = 0;
469 s->spectogram_count = 0;
470 s->remaining_fill = fft_len >> 1;
471 memset(s->fft_data, 0, fft_len * sizeof(*s->fft_data));
473 s->outpicref = ff_get_video_buffer(outlink, outlink->w, outlink->h);
475 return AVERROR(ENOMEM);
477 s->spectogram = av_calloc(spectogram_height, s->outpicref->linesize[0]);
479 return AVERROR(ENOMEM);
481 outlink->sample_aspect_ratio = av_make_q(1, 1);
482 outlink->time_base = av_make_q(1, s->fps);
483 outlink->frame_rate = av_make_q(s->fps, 1);
487 static int plot_cqt(AVFilterLink *inlink)
489 AVFilterContext *ctx = inlink->dst;
490 ShowCQTContext *s = ctx->priv;
491 AVFilterLink *outlink = ctx->outputs[0];
492 int fft_len = 1 << s->fft_bits;
493 FFTSample result[VIDEO_WIDTH][4];
495 int linesize = s->outpicref->linesize[0];
496 int video_scale = s->fullhd ? 2 : 1;
497 int video_width = (VIDEO_WIDTH/2) * video_scale;
498 int spectogram_height = (SPECTOGRAM_HEIGHT/2) * video_scale;
499 int spectogram_start = (SPECTOGRAM_START/2) * video_scale;
500 int font_height = (FONT_HEIGHT/2) * video_scale;
502 /* real part contains left samples, imaginary part contains right samples */
503 memcpy(s->fft_result_left, s->fft_data, fft_len * sizeof(*s->fft_data));
504 av_fft_permute(s->fft_context, s->fft_result_left);
505 av_fft_calc(s->fft_context, s->fft_result_left);
507 /* separate left and right, (and multiply by 2.0) */
508 s->fft_result_right[0].re = 2.0f * s->fft_result_left[0].im;
509 s->fft_result_right[0].im = 0;
510 s->fft_result_left[0].re = 2.0f * s->fft_result_left[0].re;
511 s->fft_result_left[0].im = 0;
512 for (x = 1; x <= fft_len >> 1; x++) {
513 FFTSample tmpy = s->fft_result_left[fft_len-x].im - s->fft_result_left[x].im;
515 s->fft_result_right[x].re = s->fft_result_left[x].im + s->fft_result_left[fft_len-x].im;
516 s->fft_result_right[x].im = s->fft_result_left[x].re - s->fft_result_left[fft_len-x].re;
517 s->fft_result_right[fft_len-x].re = s->fft_result_right[x].re;
518 s->fft_result_right[fft_len-x].im = -s->fft_result_right[x].im;
520 s->fft_result_left[x].re = s->fft_result_left[x].re + s->fft_result_left[fft_len-x].re;
521 s->fft_result_left[x].im = tmpy;
522 s->fft_result_left[fft_len-x].re = s->fft_result_left[x].re;
523 s->fft_result_left[fft_len-x].im = -s->fft_result_left[x].im;
526 /* calculating cqt */
527 for (x = 0; x < VIDEO_WIDTH; x++) {
529 float g = 1.0f / s->gamma;
530 FFTComplex l = {0,0};
531 FFTComplex r = {0,0};
533 for (u = 0; u < s->coeffs_len[x]; u++) {
534 FFTSample value = s->coeffs[x][u].value;
535 int index = s->coeffs[x][u].index;
536 l.re += value * s->fft_result_left[index].re;
537 l.im += value * s->fft_result_left[index].im;
538 r.re += value * s->fft_result_right[index].re;
539 r.im += value * s->fft_result_right[index].im;
541 /* result is power, not amplitude */
542 result[x][0] = l.re * l.re + l.im * l.im;
543 result[x][2] = r.re * r.re + r.im * r.im;
544 result[x][1] = 0.5f * (result[x][0] + result[x][2]);
545 result[x][3] = result[x][1];
546 result[x][0] = 255.0f * powf(FFMIN(1.0f,result[x][0]), g);
547 result[x][1] = 255.0f * powf(FFMIN(1.0f,result[x][1]), g);
548 result[x][2] = 255.0f * powf(FFMIN(1.0f,result[x][2]), g);
552 for (x = 0; x < video_width; x++) {
553 result[x][0] = 0.5f * (result[2*x][0] + result[2*x+1][0]);
554 result[x][1] = 0.5f * (result[2*x][1] + result[2*x+1][1]);
555 result[x][2] = 0.5f * (result[2*x][2] + result[2*x+1][2]);
556 result[x][3] = 0.5f * (result[2*x][3] + result[2*x+1][3]);
560 for (x = 0; x < video_width; x++) {
561 s->spectogram[s->spectogram_index*linesize + 3*x] = result[x][0] + 0.5f;
562 s->spectogram[s->spectogram_index*linesize + 3*x + 1] = result[x][1] + 0.5f;
563 s->spectogram[s->spectogram_index*linesize + 3*x + 2] = result[x][2] + 0.5f;
567 if (!s->spectogram_count) {
568 uint8_t *data = (uint8_t*) s->outpicref->data[0];
569 float rcp_result[VIDEO_WIDTH];
570 int total_length = linesize * spectogram_height;
571 int back_length = linesize * s->spectogram_index;
573 for (x = 0; x < video_width; x++)
574 rcp_result[x] = 1.0f / (result[x][3]+0.0001f);
577 for (y = 0; y < spectogram_height; y++) {
578 float height = (spectogram_height - y) * (1.0f/spectogram_height);
579 uint8_t *lineptr = data + y * linesize;
580 for (x = 0; x < video_width; x++) {
582 if (result[x][3] <= height) {
587 mul = (result[x][3] - height) * rcp_result[x];
588 *lineptr++ = mul * result[x][0] + 0.5f;
589 *lineptr++ = mul * result[x][1] + 0.5f;
590 *lineptr++ = mul * result[x][2] + 0.5f;
597 for (y = 0; y < font_height; y++) {
598 uint8_t *lineptr = data + (spectogram_height + y) * linesize;
599 uint8_t *spectogram_src = s->spectogram + s->spectogram_index * linesize;
600 for (x = 0; x < video_width; x++) {
601 uint8_t alpha = s->font_alpha[y*video_width+x];
602 uint8_t color = s->font_color[x];
603 lineptr[3*x] = (spectogram_src[3*x] * (255-alpha) + (255-color) * alpha + 255) >> 8;
604 lineptr[3*x+1] = (spectogram_src[3*x+1] * (255-alpha) + 255) >> 8;
605 lineptr[3*x+2] = (spectogram_src[3*x+2] * (255-alpha) + color * alpha + 255) >> 8;
609 for (y = 0; y < font_height; y++) {
610 uint8_t *lineptr = data + (spectogram_height + y) * linesize;
611 memcpy(lineptr, s->spectogram + s->spectogram_index * linesize, video_width*3);
613 for (x = 0; x < video_width; x += video_width/10) {
615 static const char str[] = "EF G A BC D ";
616 uint8_t *startptr = data + spectogram_height * linesize + x * 3;
617 for (u = 0; str[u]; u++) {
619 for (v = 0; v < 16; v++) {
620 uint8_t *p = startptr + v * linesize * video_scale + 8 * 3 * u * video_scale;
621 int ux = x + 8 * u * video_scale;
623 for (mask = 0x80; mask; mask >>= 1) {
624 if (mask & avpriv_vga16_font[str[u] * 16 + v]) {
625 p[0] = 255 - s->font_color[ux];
627 p[2] = s->font_color[ux];
628 if (video_scale == 2) {
630 p[linesize+1] = p[1];
631 p[linesize+2] = p[2];
632 p[3] = p[linesize+3] = 255 - s->font_color[ux+1];
633 p[4] = p[linesize+4] = 0;
634 p[5] = p[linesize+5] = s->font_color[ux+1];
637 p += 3 * video_scale;
645 /* drawing spectogram/sonogram */
646 data += spectogram_start * linesize;
647 memcpy(data, s->spectogram + s->spectogram_index*linesize, total_length - back_length);
649 data += total_length - back_length;
651 memcpy(data, s->spectogram, back_length);
653 s->outpicref->pts = s->frame_count;
654 ret = ff_filter_frame(outlink, av_frame_clone(s->outpicref));
655 s->req_fullfilled = 1;
658 s->spectogram_count = (s->spectogram_count + 1) % s->count;
659 s->spectogram_index = (s->spectogram_index + spectogram_height - 1) % spectogram_height;
663 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
665 AVFilterContext *ctx = inlink->dst;
666 ShowCQTContext *s = ctx->priv;
667 int step = inlink->sample_rate / (s->fps * s->count);
668 int fft_len = 1 << s->fft_bits;
673 while (s->remaining_fill < (fft_len >> 1)) {
675 memset(&s->fft_data[fft_len - s->remaining_fill], 0, sizeof(*s->fft_data) * s->remaining_fill);
676 ret = plot_cqt(inlink);
679 for (x = 0; x < (fft_len-step); x++)
680 s->fft_data[x] = s->fft_data[x+step];
681 s->remaining_fill += step;
686 remaining = insamples->nb_samples;
687 audio_data = (float*) insamples->data[0];
690 if (remaining >= s->remaining_fill) {
691 int i = insamples->nb_samples - remaining;
692 int j = fft_len - s->remaining_fill;
694 for (m = 0; m < s->remaining_fill; m++) {
695 s->fft_data[j+m].re = audio_data[2*(i+m)];
696 s->fft_data[j+m].im = audio_data[2*(i+m)+1];
698 ret = plot_cqt(inlink);
700 av_frame_free(&insamples);
703 remaining -= s->remaining_fill;
704 for (m = 0; m < fft_len-step; m++)
705 s->fft_data[m] = s->fft_data[m+step];
706 s->remaining_fill = step;
708 int i = insamples->nb_samples - remaining;
709 int j = fft_len - s->remaining_fill;
711 for (m = 0; m < remaining; m++) {
712 s->fft_data[m+j].re = audio_data[2*(i+m)];
713 s->fft_data[m+j].im = audio_data[2*(i+m)+1];
715 s->remaining_fill -= remaining;
719 av_frame_free(&insamples);
723 static int request_frame(AVFilterLink *outlink)
725 ShowCQTContext *s = outlink->src->priv;
726 AVFilterLink *inlink = outlink->src->inputs[0];
729 s->req_fullfilled = 0;
731 ret = ff_request_frame(inlink);
732 } while (!s->req_fullfilled && ret >= 0);
734 if (ret == AVERROR_EOF && s->outpicref)
735 filter_frame(inlink, NULL);
739 static const AVFilterPad showcqt_inputs[] = {
742 .type = AVMEDIA_TYPE_AUDIO,
743 .filter_frame = filter_frame,
748 static const AVFilterPad showcqt_outputs[] = {
751 .type = AVMEDIA_TYPE_VIDEO,
752 .config_props = config_output,
753 .request_frame = request_frame,
758 AVFilter ff_avf_showcqt = {
760 .description = NULL_IF_CONFIG_SMALL("Convert input audio to a CQT (Constant Q Transform) spectrum video output."),
762 .query_formats = query_formats,
763 .priv_size = sizeof(ShowCQTContext),
764 .inputs = showcqt_inputs,
765 .outputs = showcqt_outputs,
766 .priv_class = &showcqt_class,