2 * Copyright (c) 2014-2015 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
25 #include "libavcodec/avfft.h"
26 #include "libavutil/avassert.h"
27 #include "libavutil/opt.h"
28 #include "libavutil/xga_font_data.h"
29 #include "libavutil/eval.h"
30 #include "libavutil/pixdesc.h"
33 #include "lavfutils.h"
34 #include "lswsutils.h"
36 #if CONFIG_LIBFREETYPE
38 #include FT_FREETYPE_H
41 #include "avf_showcqt.h"
43 #define BASEFREQ 20.01523126408007475
44 #define ENDFREQ 20495.59681441799654
45 #define TLENGTH "384*tc/(384+tc*f)"
46 #define TLENGTH_MIN 0.001
47 #define VOLUME_MAX 100.0
48 #define FONTCOLOR "st(0, (midi(f)-59.5)/12);" \
49 "st(1, if(between(ld(0),0,1), 0.5-0.5*cos(2*PI*ld(0)), 0));" \
50 "r(1-ld(1)) + b(ld(1))"
52 #define PTS_TOLERANCE 1
54 #define OFFSET(x) offsetof(ShowCQTContext, x)
55 #define FLAGS (AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM)
57 static const AVOption showcqt_options[] = {
58 { "size", "set video size", OFFSET(width), AV_OPT_TYPE_IMAGE_SIZE, { .str = "1920x1080" }, 0, 0, FLAGS },
59 { "s", "set video size", OFFSET(width), AV_OPT_TYPE_IMAGE_SIZE, { .str = "1920x1080" }, 0, 0, FLAGS },
60 { "fps", "set video rate", OFFSET(rate), AV_OPT_TYPE_VIDEO_RATE, { .str = "25" }, 0, 0, FLAGS },
61 { "rate", "set video rate", OFFSET(rate), AV_OPT_TYPE_VIDEO_RATE, { .str = "25" }, 0, 0, FLAGS },
62 { "r", "set video rate", OFFSET(rate), AV_OPT_TYPE_VIDEO_RATE, { .str = "25" }, 0, 0, FLAGS },
63 { "bar_h", "set bargraph height", OFFSET(bar_h), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, FLAGS },
64 { "axis_h", "set axis height", OFFSET(axis_h), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, FLAGS },
65 { "sono_h", "set sonogram height", OFFSET(sono_h), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, FLAGS },
66 { "fullhd", "set fullhd size", OFFSET(fullhd), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS },
67 { "sono_v", "set sonogram volume", OFFSET(sono_v), AV_OPT_TYPE_STRING, { .str = "16" }, CHAR_MIN, CHAR_MAX, FLAGS },
68 { "volume", "set sonogram volume", OFFSET(sono_v), AV_OPT_TYPE_STRING, { .str = "16" }, CHAR_MIN, CHAR_MAX, FLAGS },
69 { "bar_v", "set bargraph volume", OFFSET(bar_v), AV_OPT_TYPE_STRING, { .str = "sono_v" }, CHAR_MIN, CHAR_MAX, FLAGS },
70 { "volume2", "set bargraph volume", OFFSET(bar_v), AV_OPT_TYPE_STRING, { .str = "sono_v" }, CHAR_MIN, CHAR_MAX, FLAGS },
71 { "sono_g", "set sonogram gamma", OFFSET(sono_g), AV_OPT_TYPE_FLOAT, { .dbl = 3.0 }, 1.0, 7.0, FLAGS },
72 { "gamma", "set sonogram gamma", OFFSET(sono_g), AV_OPT_TYPE_FLOAT, { .dbl = 3.0 }, 1.0, 7.0, FLAGS },
73 { "bar_g", "set bargraph gamma", OFFSET(bar_g), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 }, 1.0, 7.0, FLAGS },
74 { "gamma2", "set bargraph gamma", OFFSET(bar_g), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 }, 1.0, 7.0, FLAGS },
75 { "timeclamp", "set timeclamp", OFFSET(timeclamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.17 }, 0.1, 1.0, FLAGS },
76 { "tc", "set timeclamp", OFFSET(timeclamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.17 }, 0.1, 1.0, FLAGS },
77 { "basefreq", "set base frequency", OFFSET(basefreq), AV_OPT_TYPE_DOUBLE, { .dbl = BASEFREQ }, 10.0, 100000.0, FLAGS },
78 { "endfreq", "set end frequency", OFFSET(endfreq), AV_OPT_TYPE_DOUBLE, { .dbl = ENDFREQ }, 10.0, 100000.0, FLAGS },
79 { "coeffclamp", "set coeffclamp", OFFSET(coeffclamp), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 }, 0.1, 10.0, FLAGS },
80 { "tlength", "set tlength", OFFSET(tlength), AV_OPT_TYPE_STRING, { .str = TLENGTH }, CHAR_MIN, CHAR_MAX, FLAGS },
81 { "count", "set transform count", OFFSET(count), AV_OPT_TYPE_INT, { .i64 = 6 }, 1, 30, FLAGS },
82 { "fcount", "set frequency count", OFFSET(fcount), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 10, FLAGS },
83 { "fontfile", "set axis font", OFFSET(fontfile), AV_OPT_TYPE_STRING, { .str = NULL }, CHAR_MIN, CHAR_MAX, FLAGS },
84 { "fontcolor", "set font color", OFFSET(fontcolor), AV_OPT_TYPE_STRING, { .str = FONTCOLOR }, CHAR_MIN, CHAR_MAX, FLAGS },
85 { "axisfile", "set axis image", OFFSET(axisfile), AV_OPT_TYPE_STRING, { .str = NULL }, CHAR_MIN, CHAR_MAX, FLAGS },
86 { "axis", "draw axis", OFFSET(axis), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS },
87 { "text", "draw axis", OFFSET(axis), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS },
91 AVFILTER_DEFINE_CLASS(showcqt);
93 static void common_uninit(ShowCQTContext *s)
97 /* axis_frame may be non reference counted frame */
98 if (s->axis_frame && !s->axis_frame->buf[0]) {
99 av_freep(s->axis_frame->data);
100 for (k = 0; k < 4; k++)
101 s->axis_frame->data[k] = NULL;
104 av_frame_free(&s->axis_frame);
105 av_frame_free(&s->sono_frame);
106 av_fft_end(s->fft_ctx);
109 for (k = 0; k < s->cqt_len * 2; k++)
110 av_freep(&s->coeffs[k].val);
111 av_freep(&s->coeffs);
112 av_freep(&s->fft_data);
113 av_freep(&s->fft_result);
114 av_freep(&s->cqt_result);
117 av_freep(&s->rcp_h_buf);
119 av_freep(&s->sono_v_buf);
120 av_freep(&s->bar_v_buf);
123 static double *create_freq_table(double base, double end, int n)
125 double log_base, log_end;
126 double rcp_n = 1.0 / n;
130 freq = av_malloc_array(n, sizeof(*freq));
134 log_base = log(base);
136 for (x = 0; x < n; x++) {
137 double log_freq = log_base + (x + 0.5) * (log_end - log_base) * rcp_n;
138 freq[x] = exp(log_freq);
143 static double clip_with_log(void *log_ctx, const char *name,
144 double val, double min, double max,
145 double nan_replace, int idx)
147 int level = AV_LOG_WARNING;
149 av_log(log_ctx, level, "[%d] %s is nan, setting it to %g.\n",
150 idx, name, nan_replace);
152 } else if (val < min) {
153 av_log(log_ctx, level, "[%d] %s is too low (%g), setting it to %g.\n",
154 idx, name, val, min);
156 } else if (val > max) {
157 av_log(log_ctx, level, "[%d] %s it too high (%g), setting it to %g.\n",
158 idx, name, val, max);
164 static double a_weighting(void *p, double f)
166 double ret = 12200.0*12200.0 * (f*f*f*f);
167 ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0) *
168 sqrt((f*f + 107.7*107.7) * (f*f + 737.9*737.9));
172 static double b_weighting(void *p, double f)
174 double ret = 12200.0*12200.0 * (f*f*f);
175 ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0) * sqrt(f*f + 158.5*158.5);
179 static double c_weighting(void *p, double f)
181 double ret = 12200.0*12200.0 * (f*f);
182 ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0);
186 static int init_volume(ShowCQTContext *s)
188 const char *func_names[] = { "a_weighting", "b_weighting", "c_weighting", NULL };
189 const char *sono_names[] = { "timeclamp", "tc", "frequency", "freq", "f", "bar_v", NULL };
190 const char *bar_names[] = { "timeclamp", "tc", "frequency", "freq", "f", "sono_v", NULL };
191 double (*funcs[])(void *, double) = { a_weighting, b_weighting, c_weighting };
192 AVExpr *sono = NULL, *bar = NULL;
193 int x, ret = AVERROR(ENOMEM);
195 s->sono_v_buf = av_malloc_array(s->cqt_len, sizeof(*s->sono_v_buf));
196 s->bar_v_buf = av_malloc_array(s->cqt_len, sizeof(*s->bar_v_buf));
197 if (!s->sono_v_buf || !s->bar_v_buf)
200 if ((ret = av_expr_parse(&sono, s->sono_v, sono_names, func_names, funcs, NULL, NULL, 0, s->ctx)) < 0)
203 if ((ret = av_expr_parse(&bar, s->bar_v, bar_names, func_names, funcs, NULL, NULL, 0, s->ctx)) < 0)
206 for (x = 0; x < s->cqt_len; x++) {
207 double vars[] = { s->timeclamp, s->timeclamp, s->freq[x], s->freq[x], s->freq[x], 0.0 };
208 double vol = clip_with_log(s->ctx, "sono_v", av_expr_eval(sono, vars, NULL), 0.0, VOLUME_MAX, 0.0, x);
210 vol = clip_with_log(s->ctx, "bar_v", av_expr_eval(bar, vars, NULL), 0.0, VOLUME_MAX, 0.0, x);
211 s->bar_v_buf[x] = vol * vol;
213 vol = clip_with_log(s->ctx, "sono_v", av_expr_eval(sono, vars, NULL), 0.0, VOLUME_MAX, 0.0, x);
214 s->sono_v_buf[x] = vol * vol;
221 av_freep(&s->sono_v_buf);
222 av_freep(&s->bar_v_buf);
228 static void cqt_calc(FFTComplex *dst, const FFTComplex *src, const Coeffs *coeffs,
229 int len, int fft_len)
232 for (k = 0; k < len; k++) {
233 FFTComplex l, r, a = {0,0}, b = {0,0};
235 for (x = 0; x < coeffs[k].len; x++) {
236 FFTSample u = coeffs[k].val[x];
237 i = coeffs[k].start + x;
239 a.re += u * src[i].re;
240 a.im += u * src[i].im;
241 b.re += u * src[j].re;
242 b.im += u * src[j].im;
245 /* separate left and right, (and multiply by 2.0) */
250 dst[k].re = l.re * l.re + l.im * l.im;
251 dst[k].im = r.re * r.re + r.im * r.im;
256 static void cqt_calc_interleave(FFTComplex *dst, const FFTComplex *src, const Coeffs *coeffs,
257 int len, int fft_len)
261 for (k = 0; k < len; k++) {
262 FFTComplex l, r, a = {0,0}, b = {0,0};
265 for (x = 0; x < coeffs[m].len; x++) {
266 FFTSample u = coeffs[m].val[x];
267 i = coeffs[m].start + x;
268 a.re += u * src[i].re;
269 a.im += u * src[i].im;
273 for (x = 0; x < coeffs[m].len; x++) {
274 FFTSample u = coeffs[m].val[x];
275 i = coeffs[m].start + x;
276 b.re += u * src[i].re;
277 b.im += u * src[i].im;
280 /* separate left and right, (and multiply by 2.0) */
285 dst[k].re = l.re * l.re + l.im * l.im;
286 dst[k].im = r.re * r.re + r.im * r.im;
291 static int init_cqt(ShowCQTContext *s)
293 const char *var_names[] = { "timeclamp", "tc", "frequency", "freq", "f", NULL };
295 int rate = s->ctx->inputs[0]->sample_rate;
296 int nb_cqt_coeffs = 0, nb_cqt_coeffs_r = 0;
299 if ((ret = av_expr_parse(&expr, s->tlength, var_names, NULL, NULL, NULL, NULL, 0, s->ctx)) < 0)
302 ret = AVERROR(ENOMEM);
303 if (!(s->coeffs = av_calloc(s->cqt_len * 2, sizeof(*s->coeffs))))
306 for (k = 0; k < s->cqt_len; k++) {
307 double vars[] = { s->timeclamp, s->timeclamp, s->freq[k], s->freq[k], s->freq[k] };
308 double flen, center, tlength;
309 int start, end, m = (s->cqt_coeffs_type == COEFFS_TYPE_INTERLEAVE) ? (2 * k) : k;
311 if (s->freq[k] > 0.5 * rate)
313 tlength = clip_with_log(s->ctx, "tlength", av_expr_eval(expr, vars, NULL),
314 TLENGTH_MIN, s->timeclamp, s->timeclamp, k);
316 flen = 8.0 * s->fft_len / (tlength * rate);
317 center = s->freq[k] * s->fft_len / rate;
318 start = FFMAX(0, ceil(center - 0.5 * flen));
319 end = FFMIN(s->fft_len, floor(center + 0.5 * flen));
321 s->coeffs[m].start = start & ~(s->cqt_align - 1);
322 s->coeffs[m].len = (end | (s->cqt_align - 1)) + 1 - s->coeffs[m].start;
323 nb_cqt_coeffs += s->coeffs[m].len;
324 if (!(s->coeffs[m].val = av_calloc(s->coeffs[m].len, sizeof(*s->coeffs[m].val))))
327 if (s->cqt_coeffs_type == COEFFS_TYPE_INTERLEAVE) {
328 s->coeffs[m+1].start = (s->fft_len - end) & ~(s->cqt_align - 1);
329 s->coeffs[m+1].len = ((s->fft_len - start) | (s->cqt_align - 1)) + 1 - s->coeffs[m+1].start;
330 nb_cqt_coeffs_r += s->coeffs[m+1].len;
331 if (!(s->coeffs[m+1].val = av_calloc(s->coeffs[m+1].len, sizeof(*s->coeffs[m+1].val))))
335 for (x = start; x <= end; x++) {
336 int sign = (x & 1) ? (-1) : 1;
337 double y = 2.0 * M_PI * (x - center) * (1.0 / flen);
339 double w = 0.355768 + 0.487396 * cos(y) + 0.144232 * cos(2*y) + 0.012604 * cos(3*y);
340 w *= sign * (1.0 / s->fft_len);
341 s->coeffs[m].val[x - s->coeffs[m].start] = w;
342 if (s->cqt_coeffs_type == COEFFS_TYPE_INTERLEAVE)
343 s->coeffs[m+1].val[(s->fft_len - x) - s->coeffs[m+1].start] = w;
348 if (s->cqt_coeffs_type == COEFFS_TYPE_DEFAULT)
349 av_log(s->ctx, AV_LOG_INFO, "nb_cqt_coeffs = %d.\n", nb_cqt_coeffs);
351 av_log(s->ctx, AV_LOG_INFO, "nb_cqt_coeffs = {%d,%d}.\n", nb_cqt_coeffs, nb_cqt_coeffs_r);
357 for (k = 0; k < s->cqt_len * 2; k++)
358 av_freep(&s->coeffs[k].val);
359 av_freep(&s->coeffs);
363 static AVFrame *alloc_frame_empty(enum AVPixelFormat format, int w, int h)
366 out = av_frame_alloc();
369 out->format = format;
372 if (av_frame_get_buffer(out, 32) < 0) {
376 if (format == AV_PIX_FMT_RGB24 || format == AV_PIX_FMT_RGBA) {
377 memset(out->data[0], 0, out->linesize[0] * h);
379 int hh = (format == AV_PIX_FMT_YUV420P || format == AV_PIX_FMT_YUVA420P) ? h / 2 : h;
380 memset(out->data[0], 16, out->linesize[0] * h);
381 memset(out->data[1], 128, out->linesize[1] * hh);
382 memset(out->data[2], 128, out->linesize[2] * hh);
384 memset(out->data[3], 0, out->linesize[3] * h);
389 static enum AVPixelFormat convert_axis_pixel_format(enum AVPixelFormat format)
392 case AV_PIX_FMT_RGB24: format = AV_PIX_FMT_RGBA; break;
393 case AV_PIX_FMT_YUV444P: format = AV_PIX_FMT_YUVA444P; break;
394 case AV_PIX_FMT_YUV422P: format = AV_PIX_FMT_YUVA422P; break;
395 case AV_PIX_FMT_YUV420P: format = AV_PIX_FMT_YUVA420P; break;
400 static int init_axis_empty(ShowCQTContext *s)
402 if (!(s->axis_frame = alloc_frame_empty(convert_axis_pixel_format(s->format), s->width, s->axis_h)))
403 return AVERROR(ENOMEM);
407 static int init_axis_from_file(ShowCQTContext *s)
409 uint8_t *tmp_data[4] = { NULL };
411 enum AVPixelFormat tmp_format;
412 int tmp_w, tmp_h, ret;
414 if ((ret = ff_load_image(tmp_data, tmp_linesize, &tmp_w, &tmp_h, &tmp_format,
415 s->axisfile, s->ctx)) < 0)
418 ret = AVERROR(ENOMEM);
419 if (!(s->axis_frame = av_frame_alloc()))
422 if ((ret = ff_scale_image(s->axis_frame->data, s->axis_frame->linesize, s->width, s->axis_h,
423 convert_axis_pixel_format(s->format), tmp_data, tmp_linesize, tmp_w, tmp_h,
424 tmp_format, s->ctx)) < 0)
427 s->axis_frame->width = s->width;
428 s->axis_frame->height = s->axis_h;
429 s->axis_frame->format = convert_axis_pixel_format(s->format);
434 av_frame_free(&s->axis_frame);
439 static double midi(void *p, double f)
441 return log2(f/440.0) * 12.0 + 69.0;
444 static double r_func(void *p, double x)
446 x = av_clipd(x, 0.0, 1.0);
447 return (int)(x*255.0+0.5) << 16;
450 static double g_func(void *p, double x)
452 x = av_clipd(x, 0.0, 1.0);
453 return (int)(x*255.0+0.5) << 8;
456 static double b_func(void *p, double x)
458 x = av_clipd(x, 0.0, 1.0);
459 return (int)(x*255.0+0.5);
462 static int init_axis_color(ShowCQTContext *s, AVFrame *tmp)
464 const char *var_names[] = { "timeclamp", "tc", "frequency", "freq", "f", NULL };
465 const char *func_names[] = { "midi", "r", "g", "b", NULL };
466 double (*funcs[])(void *, double) = { midi, r_func, g_func, b_func };
471 if (s->basefreq != (double) BASEFREQ || s->endfreq != (double) ENDFREQ) {
472 av_log(s->ctx, AV_LOG_WARNING, "font axis rendering is not implemented in non-default frequency range,"
473 " please use axisfile option instead.\n");
474 return AVERROR(EINVAL);
477 if (s->cqt_len == 1920)
479 else if (!(freq = create_freq_table(s->basefreq, s->endfreq, 1920)))
480 return AVERROR(ENOMEM);
482 if ((ret = av_expr_parse(&expr, s->fontcolor, var_names, func_names, funcs, NULL, NULL, 0, s->ctx)) < 0) {
488 for (x = 0; x < 1920; x++) {
489 double vars[] = { s->timeclamp, s->timeclamp, freq[x], freq[x], freq[x] };
490 int color = (int) av_expr_eval(expr, vars, NULL);
491 uint8_t r = (color >> 16) & 0xFF, g = (color >> 8) & 0xFF, b = color & 0xFF;
492 uint8_t *data = tmp->data[0];
493 int linesize = tmp->linesize[0];
494 for (y = 0; y < 32; y++) {
495 data[linesize * y + 4 * x] = r;
496 data[linesize * y + 4 * x + 1] = g;
497 data[linesize * y + 4 * x + 2] = b;
498 data[linesize * y + 4 * x + 3] = 0;
508 static int render_freetype(ShowCQTContext *s, AVFrame *tmp)
510 #if CONFIG_LIBFREETYPE
511 const char *str = "EF G A BC D ";
512 uint8_t *data = tmp->data[0];
513 int linesize = tmp->linesize[0];
514 FT_Library lib = NULL;
516 int font_width = 16, font_height = 32;
517 int font_repeat = font_width * 12;
518 int linear_hori_advance = font_width * 65536;
519 int non_monospace_warning = 0;
523 return AVERROR(EINVAL);
525 if (FT_Init_FreeType(&lib))
528 if (FT_New_Face(lib, s->fontfile, 0, &face))
531 if (FT_Set_Char_Size(face, 16*64, 0, 0, 0))
534 if (FT_Load_Char(face, 'A', FT_LOAD_RENDER))
537 if (FT_Set_Char_Size(face, 16*64 * linear_hori_advance / face->glyph->linearHoriAdvance, 0, 0, 0))
540 for (x = 0; x < 12; x++) {
541 int sx, sy, rx, bx, by, dx, dy;
546 if (FT_Load_Char(face, str[x], FT_LOAD_RENDER))
549 if (face->glyph->advance.x != font_width*64 && !non_monospace_warning) {
550 av_log(s->ctx, AV_LOG_WARNING, "font is not monospace.\n");
551 non_monospace_warning = 1;
554 sy = font_height - 8 - face->glyph->bitmap_top;
555 for (rx = 0; rx < 10; rx++) {
556 sx = rx * font_repeat + x * font_width + face->glyph->bitmap_left;
557 for (by = 0; by < face->glyph->bitmap.rows; by++) {
561 if (dy >= font_height)
564 for (bx = 0; bx < face->glyph->bitmap.width; bx++) {
570 data[dy*linesize+4*dx+3] = face->glyph->bitmap.buffer[by*face->glyph->bitmap.width+bx];
577 FT_Done_FreeType(lib);
581 av_log(s->ctx, AV_LOG_WARNING, "error while loading freetype font, using default font instead.\n");
583 FT_Done_FreeType(lib);
584 return AVERROR(EINVAL);
587 av_log(s->ctx, AV_LOG_WARNING, "freetype is not available, ignoring fontfile option.\n");
588 return AVERROR(EINVAL);
592 static int render_default_font(AVFrame *tmp)
594 const char *str = "EF G A BC D ";
596 uint8_t *data = tmp->data[0];
597 int linesize = tmp->linesize[0];
599 for (x = 0; x < 1920; x += 192) {
600 uint8_t *startptr = data + 4 * x;
601 for (u = 0; u < 12; u++) {
602 for (v = 0; v < 16; v++) {
603 uint8_t *p = startptr + 2 * v * linesize + 16 * 4 * u;
604 for (mask = 0x80; mask; mask >>= 1, p += 8) {
605 if (mask & avpriv_vga16_font[str[u] * 16 + v]) {
619 static int init_axis_from_font(ShowCQTContext *s)
622 int ret = AVERROR(ENOMEM);
624 if (!(tmp = alloc_frame_empty(AV_PIX_FMT_RGBA, 1920, 32)))
627 if (!(s->axis_frame = av_frame_alloc()))
630 if ((ret = init_axis_color(s, tmp)) < 0)
633 if (render_freetype(s, tmp) < 0 && (ret = render_default_font(tmp)) < 0)
636 if ((ret = ff_scale_image(s->axis_frame->data, s->axis_frame->linesize, s->width, s->axis_h,
637 convert_axis_pixel_format(s->format), tmp->data, tmp->linesize,
638 1920, 32, AV_PIX_FMT_RGBA, s->ctx)) < 0)
642 s->axis_frame->width = s->width;
643 s->axis_frame->height = s->axis_h;
644 s->axis_frame->format = convert_axis_pixel_format(s->format);
649 av_frame_free(&s->axis_frame);
653 static float calculate_gamma(float v, float g)
662 return sqrtf(sqrtf(v));
663 return expf(logf(v) / g);
666 static void rgb_from_cqt(ColorFloat *c, const FFTComplex *v, float g, int len)
669 for (x = 0; x < len; x++) {
670 c[x].rgb.r = 255.0f * calculate_gamma(FFMIN(1.0f, v[x].re), g);
671 c[x].rgb.g = 255.0f * calculate_gamma(FFMIN(1.0f, 0.5f * (v[x].re + v[x].im)), g);
672 c[x].rgb.b = 255.0f * calculate_gamma(FFMIN(1.0f, v[x].im), g);
676 static void yuv_from_cqt(ColorFloat *c, const FFTComplex *v, float gamma, int len)
679 for (x = 0; x < len; x++) {
681 r = calculate_gamma(FFMIN(1.0f, v[x].re), gamma);
682 g = calculate_gamma(FFMIN(1.0f, 0.5f * (v[x].re + v[x].im)), gamma);
683 b = calculate_gamma(FFMIN(1.0f, v[x].im), gamma);
684 c[x].yuv.y = 65.481f * r + 128.553f * g + 24.966f * b;
685 c[x].yuv.u = -37.797f * r - 74.203f * g + 112.0f * b;
686 c[x].yuv.v = 112.0f * r - 93.786f * g - 18.214 * b;
690 static void draw_bar_rgb(AVFrame *out, const float *h, const float *rcp_h,
691 const ColorFloat *c, int bar_h)
693 int x, y, w = out->width;
694 float mul, ht, rcp_bar_h = 1.0f / bar_h;
695 uint8_t *v = out->data[0], *lp;
696 int ls = out->linesize[0];
698 for (y = 0; y < bar_h; y++) {
699 ht = (bar_h - y) * rcp_bar_h;
701 for (x = 0; x < w; x++) {
707 mul = (h[x] - ht) * rcp_h[x];
708 *lp++ = mul * c[x].rgb.r + 0.5f;
709 *lp++ = mul * c[x].rgb.g + 0.5f;
710 *lp++ = mul * c[x].rgb.b + 0.5f;
716 static void draw_bar_yuv(AVFrame *out, const float *h, const float *rcp_h,
717 const ColorFloat *c, int bar_h)
719 int x, y, yh, w = out->width;
720 float mul, ht, rcp_bar_h = 1.0f / bar_h;
721 uint8_t *vy = out->data[0], *vu = out->data[1], *vv = out->data[2];
722 uint8_t *lpy, *lpu, *lpv;
723 int lsy = out->linesize[0], lsu = out->linesize[1], lsv = out->linesize[2];
724 int fmt = out->format;
726 for (y = 0; y < bar_h; y += 2) {
727 yh = (fmt == AV_PIX_FMT_YUV420P) ? y / 2 : y;
728 ht = (bar_h - y) * rcp_bar_h;
732 for (x = 0; x < w; x += 2) {
738 mul = (h[x] - ht) * rcp_h[x];
739 *lpy++ = mul * c[x].yuv.y + 16.5f;
740 *lpu++ = mul * c[x].yuv.u + 128.5f;
741 *lpv++ = mul * c[x].yuv.v + 128.5f;
743 /* u and v are skipped on yuv422p and yuv420p */
744 if (fmt == AV_PIX_FMT_YUV444P) {
750 mul = (h[x+1] - ht) * rcp_h[x+1];
751 *lpy++ = mul * c[x+1].yuv.y + 16.5f;
752 *lpu++ = mul * c[x+1].yuv.u + 128.5f;
753 *lpv++ = mul * c[x+1].yuv.v + 128.5f;
759 mul = (h[x+1] - ht) * rcp_h[x+1];
760 *lpy++ = mul * c[x+1].yuv.y + 16.5f;
765 ht = (bar_h - (y+1)) * rcp_bar_h;
766 lpy = vy + (y+1) * lsy;
767 lpu = vu + (y+1) * lsu;
768 lpv = vv + (y+1) * lsv;
769 for (x = 0; x < w; x += 2) {
770 /* u and v are skipped on yuv420p */
771 if (fmt != AV_PIX_FMT_YUV420P) {
777 mul = (h[x] - ht) * rcp_h[x];
778 *lpy++ = mul * c[x].yuv.y + 16.5f;
779 *lpu++ = mul * c[x].yuv.u + 128.5f;
780 *lpv++ = mul * c[x].yuv.v + 128.5f;
786 mul = (h[x] - ht) * rcp_h[x];
787 *lpy++ = mul * c[x].yuv.y + 16.5f;
790 /* u and v are skipped on yuv422p and yuv420p */
791 if (out->format == AV_PIX_FMT_YUV444P) {
797 mul = (h[x+1] - ht) * rcp_h[x+1];
798 *lpy++ = mul * c[x+1].yuv.y + 16.5f;
799 *lpu++ = mul * c[x+1].yuv.u + 128.5f;
800 *lpv++ = mul * c[x+1].yuv.v + 128.5f;
806 mul = (h[x+1] - ht) * rcp_h[x+1];
807 *lpy++ = mul * c[x+1].yuv.y + 16.5f;
814 static void draw_axis_rgb(AVFrame *out, AVFrame *axis, const ColorFloat *c, int off)
816 int x, y, w = axis->width, h = axis->height;
817 float a, rcp_255 = 1.0f / 255.0f;
820 for (y = 0; y < h; y++) {
821 lp = out->data[0] + (off + y) * out->linesize[0];
822 lpa = axis->data[0] + y * axis->linesize[0];
823 for (x = 0; x < w; x++) {
824 a = rcp_255 * lpa[3];
825 *lp++ = a * lpa[0] + (1.0f - a) * c[x].rgb.r + 0.5f;
826 *lp++ = a * lpa[1] + (1.0f - a) * c[x].rgb.g + 0.5f;
827 *lp++ = a * lpa[2] + (1.0f - a) * c[x].rgb.b + 0.5f;
833 static void draw_axis_yuv(AVFrame *out, AVFrame *axis, const ColorFloat *c, int off)
835 int fmt = out->format, x, y, yh, w = axis->width, h = axis->height;
836 int offh = (fmt == AV_PIX_FMT_YUV420P) ? off / 2 : off;
837 float a, rcp_255 = 1.0f / 255.0f;
838 uint8_t *vy = out->data[0], *vu = out->data[1], *vv = out->data[2];
839 uint8_t *vay = axis->data[0], *vau = axis->data[1], *vav = axis->data[2], *vaa = axis->data[3];
840 int lsy = out->linesize[0], lsu = out->linesize[1], lsv = out->linesize[2];
841 int lsay = axis->linesize[0], lsau = axis->linesize[1], lsav = axis->linesize[2], lsaa = axis->linesize[3];
842 uint8_t *lpy, *lpu, *lpv, *lpay, *lpau, *lpav, *lpaa;
844 for (y = 0; y < h; y += 2) {
845 yh = (fmt == AV_PIX_FMT_YUV420P) ? y / 2 : y;
846 lpy = vy + (off + y) * lsy;
847 lpu = vu + (offh + yh) * lsu;
848 lpv = vv + (offh + yh) * lsv;
849 lpay = vay + y * lsay;
850 lpau = vau + yh * lsau;
851 lpav = vav + yh * lsav;
852 lpaa = vaa + y * lsaa;
853 for (x = 0; x < w; x += 2) {
854 a = rcp_255 * (*lpaa++);
855 *lpy++ = a * (*lpay++) + (1.0f - a) * (c[x].yuv.y + 16.0f) + 0.5f;
856 *lpu++ = a * (*lpau++) + (1.0f - a) * (c[x].yuv.u + 128.0f) + 0.5f;
857 *lpv++ = a * (*lpav++) + (1.0f - a) * (c[x].yuv.v + 128.0f) + 0.5f;
858 /* u and v are skipped on yuv422p and yuv420p */
859 a = rcp_255 * (*lpaa++);
860 *lpy++ = a * (*lpay++) + (1.0f - a) * (c[x+1].yuv.y + 16.0f) + 0.5f;
861 if (fmt == AV_PIX_FMT_YUV444P) {
862 *lpu++ = a * (*lpau++) + (1.0f - a) * (c[x+1].yuv.u + 128.0f) + 0.5f;
863 *lpv++ = a * (*lpav++) + (1.0f - a) * (c[x+1].yuv.v + 128.0f) + 0.5f;
867 lpy = vy + (off + y + 1) * lsy;
868 lpu = vu + (off + y + 1) * lsu;
869 lpv = vv + (off + y + 1) * lsv;
870 lpay = vay + (y + 1) * lsay;
871 lpau = vau + (y + 1) * lsau;
872 lpav = vav + (y + 1) * lsav;
873 lpaa = vaa + (y + 1) * lsaa;
874 for (x = 0; x < out->width; x += 2) {
875 /* u and v are skipped on yuv420p */
876 a = rcp_255 * (*lpaa++);
877 *lpy++ = a * (*lpay++) + (1.0f - a) * (c[x].yuv.y + 16.0f) + 0.5f;
878 if (fmt != AV_PIX_FMT_YUV420P) {
879 *lpu++ = a * (*lpau++) + (1.0f - a) * (c[x].yuv.u + 128.0f) + 0.5f;
880 *lpv++ = a * (*lpav++) + (1.0f - a) * (c[x].yuv.v + 128.0f) + 0.5f;
882 /* u and v are skipped on yuv422p and yuv420p */
883 a = rcp_255 * (*lpaa++);
884 *lpy++ = a * (*lpay++) + (1.0f - a) * (c[x+1].yuv.y + 16.0f) + 0.5f;
885 if (fmt == AV_PIX_FMT_YUV444P) {
886 *lpu++ = a * (*lpau++) + (1.0f - a) * (c[x+1].yuv.u + 128.0f) + 0.5f;
887 *lpv++ = a * (*lpav++) + (1.0f - a) * (c[x+1].yuv.v + 128.0f) + 0.5f;
893 static void draw_sono(AVFrame *out, AVFrame *sono, int off, int idx)
895 int fmt = out->format, h = sono->height;
896 int nb_planes = (fmt == AV_PIX_FMT_RGB24) ? 1 : 3;
897 int offh = (fmt == AV_PIX_FMT_YUV420P) ? off / 2 : off;
898 int inc = (fmt == AV_PIX_FMT_YUV420P) ? 2 : 1;
901 ls = FFMIN(out->linesize[0], sono->linesize[0]);
902 for (y = 0; y < h; y++) {
903 memcpy(out->data[0] + (off + y) * out->linesize[0],
904 sono->data[0] + (idx + y) % h * sono->linesize[0], ls);
907 for (i = 1; i < nb_planes; i++) {
908 ls = FFMIN(out->linesize[i], sono->linesize[i]);
909 for (y = 0; y < h; y += inc) {
910 yh = (fmt == AV_PIX_FMT_YUV420P) ? y / 2 : y;
911 memcpy(out->data[i] + (offh + yh) * out->linesize[i],
912 sono->data[i] + (idx + y) % h * sono->linesize[i], ls);
917 static void update_sono_rgb(AVFrame *sono, const ColorFloat *c, int idx)
919 int x, w = sono->width;
920 uint8_t *lp = sono->data[0] + idx * sono->linesize[0];
922 for (x = 0; x < w; x++) {
923 *lp++ = c[x].rgb.r + 0.5f;
924 *lp++ = c[x].rgb.g + 0.5f;
925 *lp++ = c[x].rgb.b + 0.5f;
929 static void update_sono_yuv(AVFrame *sono, const ColorFloat *c, int idx)
931 int x, fmt = sono->format, w = sono->width;
932 uint8_t *lpy = sono->data[0] + idx * sono->linesize[0];
933 uint8_t *lpu = sono->data[1] + idx * sono->linesize[1];
934 uint8_t *lpv = sono->data[2] + idx * sono->linesize[2];
936 for (x = 0; x < w; x += 2) {
937 *lpy++ = c[x].yuv.y + 16.5f;
938 *lpu++ = c[x].yuv.u + 128.5f;
939 *lpv++ = c[x].yuv.v + 128.5f;
940 *lpy++ = c[x+1].yuv.y + 16.5f;
941 if (fmt == AV_PIX_FMT_YUV444P) {
942 *lpu++ = c[x+1].yuv.u + 128.5f;
943 *lpv++ = c[x+1].yuv.v + 128.5f;
948 static void process_cqt(ShowCQTContext *s)
951 if (!s->sono_count) {
952 for (x = 0; x < s->cqt_len; x++) {
953 s->h_buf[x] = s->bar_v_buf[x] * 0.5f * (s->cqt_result[x].re + s->cqt_result[x].im);
956 float rcp_fcount = 1.0f / s->fcount;
957 for (x = 0; x < s->width; x++) {
959 for (i = 0; i < s->fcount; i++)
960 h += s->h_buf[s->fcount * x + i];
961 s->h_buf[x] = rcp_fcount * h;
964 for (x = 0; x < s->width; x++) {
965 s->h_buf[x] = calculate_gamma(s->h_buf[x], s->bar_g);
966 s->rcp_h_buf[x] = 1.0f / (s->h_buf[x] + 0.0001f);
970 for (x = 0; x < s->cqt_len; x++) {
971 s->cqt_result[x].re *= s->sono_v_buf[x];
972 s->cqt_result[x].im *= s->sono_v_buf[x];
976 float rcp_fcount = 1.0f / s->fcount;
977 for (x = 0; x < s->width; x++) {
978 FFTComplex result = {0.0f, 0.0f};
979 for (i = 0; i < s->fcount; i++) {
980 result.re += s->cqt_result[s->fcount * x + i].re;
981 result.im += s->cqt_result[s->fcount * x + i].im;
983 s->cqt_result[x].re = rcp_fcount * result.re;
984 s->cqt_result[x].im = rcp_fcount * result.im;
988 if (s->format == AV_PIX_FMT_RGB24)
989 rgb_from_cqt(s->c_buf, s->cqt_result, s->sono_g, s->width);
991 yuv_from_cqt(s->c_buf, s->cqt_result, s->sono_g, s->width);
994 static int plot_cqt(AVFilterContext *ctx, AVFrame **frameout)
996 AVFilterLink *outlink = ctx->outputs[0];
997 ShowCQTContext *s = ctx->priv;
999 memcpy(s->fft_result, s->fft_data, s->fft_len * sizeof(*s->fft_data));
1000 av_fft_permute(s->fft_ctx, s->fft_result);
1001 av_fft_calc(s->fft_ctx, s->fft_result);
1002 s->fft_result[s->fft_len] = s->fft_result[0];
1003 s->cqt_calc(s->cqt_result, s->fft_result, s->coeffs, s->cqt_len, s->fft_len);
1006 s->update_sono(s->sono_frame, s->c_buf, s->sono_idx);
1007 if (!s->sono_count) {
1008 AVFrame *out = *frameout = ff_get_video_buffer(outlink, outlink->w, outlink->h);
1010 return AVERROR(ENOMEM);
1012 s->draw_bar(out, s->h_buf, s->rcp_h_buf, s->c_buf, s->bar_h);
1014 s->draw_axis(out, s->axis_frame, s->c_buf, s->bar_h);
1016 s->draw_sono(out, s->sono_frame, s->bar_h + s->axis_h, s->sono_idx);
1017 out->pts = s->next_pts;
1018 s->next_pts += PTS_STEP;
1020 s->sono_count = (s->sono_count + 1) % s->count;
1022 s->sono_idx = (s->sono_idx + s->sono_h - 1) % s->sono_h;
1026 /* main filter control */
1027 static av_cold int init(AVFilterContext *ctx)
1029 ShowCQTContext *s = ctx->priv;
1033 av_log(ctx, AV_LOG_WARNING, "fullhd option is deprecated, use size/s option instead.\n");
1034 if (s->width != 1920 || s->height != 1080) {
1035 av_log(ctx, AV_LOG_ERROR, "fullhd set to 0 but with custom dimension.\n");
1036 return AVERROR(EINVAL);
1043 if (s->axis_h < 0) {
1044 s->axis_h = s->width / 60;
1047 if (s->bar_h >= 0 && s->sono_h >= 0)
1048 s->axis_h = s->height - s->bar_h - s->sono_h;
1049 if (s->bar_h >= 0 && s->sono_h < 0)
1050 s->axis_h = FFMIN(s->axis_h, s->height - s->bar_h);
1051 if (s->bar_h < 0 && s->sono_h >= 0)
1052 s->axis_h = FFMIN(s->axis_h, s->height - s->sono_h);
1056 s->bar_h = (s->height - s->axis_h) / 2;
1060 s->bar_h = s->height - s->sono_h - s->axis_h;
1064 s->sono_h = s->height - s->axis_h - s->bar_h;
1066 if ((s->width & 1) || (s->height & 1) || (s->bar_h & 1) || (s->axis_h & 1) || (s->sono_h & 1) ||
1067 (s->bar_h < 0) || (s->axis_h < 0) || (s->sono_h < 0) || (s->bar_h > s->height) ||
1068 (s->axis_h > s->height) || (s->sono_h > s->height) || (s->bar_h + s->axis_h + s->sono_h != s->height)) {
1069 av_log(ctx, AV_LOG_ERROR, "invalid dimension.\n");
1070 return AVERROR(EINVAL);
1076 } while(s->fcount * s->width < 1920 && s->fcount < 10);
1082 static av_cold void uninit(AVFilterContext *ctx)
1084 common_uninit(ctx->priv);
1087 static int query_formats(AVFilterContext *ctx)
1089 AVFilterFormats *formats = NULL;
1090 AVFilterChannelLayouts *layouts = NULL;
1091 AVFilterLink *inlink = ctx->inputs[0];
1092 AVFilterLink *outlink = ctx->outputs[0];
1093 enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_NONE };
1094 enum AVPixelFormat pix_fmts[] = {
1095 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P,
1096 AV_PIX_FMT_YUV444P, AV_PIX_FMT_RGB24, AV_PIX_FMT_NONE
1098 int64_t channel_layouts[] = { AV_CH_LAYOUT_STEREO, AV_CH_LAYOUT_STEREO_DOWNMIX, -1 };
1101 /* set input audio formats */
1102 formats = ff_make_format_list(sample_fmts);
1103 if ((ret = ff_formats_ref(formats, &inlink->out_formats)) < 0)
1106 layouts = avfilter_make_format64_list(channel_layouts);
1107 if ((ret = ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts)) < 0)
1110 formats = ff_all_samplerates();
1111 if ((ret = ff_formats_ref(formats, &inlink->out_samplerates)) < 0)
1114 /* set output video format */
1115 formats = ff_make_format_list(pix_fmts);
1116 if ((ret = ff_formats_ref(formats, &outlink->in_formats)) < 0)
1122 static int config_output(AVFilterLink *outlink)
1124 AVFilterContext *ctx = outlink->src;
1125 AVFilterLink *inlink = ctx->inputs[0];
1126 ShowCQTContext *s = ctx->priv;
1131 outlink->w = s->width;
1132 outlink->h = s->height;
1133 s->format = outlink->format;
1134 outlink->sample_aspect_ratio = av_make_q(1, 1);
1135 outlink->frame_rate = s->rate;
1136 outlink->time_base = av_mul_q(av_inv_q(s->rate), av_make_q(1, PTS_STEP));
1137 av_log(ctx, AV_LOG_INFO, "video: %dx%d %s %d/%d fps, bar_h = %d, axis_h = %d, sono_h = %d.\n",
1138 s->width, s->height, av_get_pix_fmt_name(s->format), s->rate.num, s->rate.den,
1139 s->bar_h, s->axis_h, s->sono_h);
1141 s->cqt_len = s->width * s->fcount;
1142 if (!(s->freq = create_freq_table(s->basefreq, s->endfreq, s->cqt_len)))
1143 return AVERROR(ENOMEM);
1145 if ((ret = init_volume(s)) < 0)
1148 s->fft_bits = ceil(log2(inlink->sample_rate * s->timeclamp));
1149 s->fft_len = 1 << s->fft_bits;
1150 av_log(ctx, AV_LOG_INFO, "fft_len = %d, cqt_len = %d.\n", s->fft_len, s->cqt_len);
1152 s->fft_ctx = av_fft_init(s->fft_bits, 0);
1153 s->fft_data = av_calloc(s->fft_len, sizeof(*s->fft_data));
1154 s->fft_result = av_calloc(s->fft_len + 64, sizeof(*s->fft_result));
1155 s->cqt_result = av_malloc_array(s->cqt_len, sizeof(*s->cqt_result));
1156 if (!s->fft_ctx || !s->fft_data || !s->fft_result || !s->cqt_result)
1157 return AVERROR(ENOMEM);
1160 s->cqt_coeffs_type = COEFFS_TYPE_DEFAULT;
1161 s->cqt_calc = cqt_calc;
1162 s->draw_sono = draw_sono;
1163 if (s->format == AV_PIX_FMT_RGB24) {
1164 s->draw_bar = draw_bar_rgb;
1165 s->draw_axis = draw_axis_rgb;
1166 s->update_sono = update_sono_rgb;
1168 s->draw_bar = draw_bar_yuv;
1169 s->draw_axis = draw_axis_yuv;
1170 s->update_sono = update_sono_yuv;
1173 if ((ret = init_cqt(s)) < 0)
1178 if ((ret = init_axis_empty(s)) < 0)
1180 } else if (s->axisfile) {
1181 if (init_axis_from_file(s) < 0) {
1182 av_log(ctx, AV_LOG_WARNING, "loading axis image failed, fallback to font rendering.\n");
1183 if (init_axis_from_font(s) < 0) {
1184 av_log(ctx, AV_LOG_WARNING, "loading axis font failed, disable text drawing.\n");
1185 if ((ret = init_axis_empty(s)) < 0)
1190 if (init_axis_from_font(s) < 0) {
1191 av_log(ctx, AV_LOG_WARNING, "loading axis font failed, disable text drawing.\n");
1192 if ((ret = init_axis_empty(s)) < 0)
1199 s->sono_frame = alloc_frame_empty((outlink->format == AV_PIX_FMT_YUV420P) ?
1200 AV_PIX_FMT_YUV422P : outlink->format, s->width, s->sono_h);
1202 return AVERROR(ENOMEM);
1205 s->h_buf = av_malloc_array(s->cqt_len, sizeof (*s->h_buf));
1206 s->rcp_h_buf = av_malloc_array(s->width, sizeof(*s->rcp_h_buf));
1207 s->c_buf = av_malloc_array(s->width, sizeof(*s->c_buf));
1208 if (!s->h_buf || !s->rcp_h_buf || !s->c_buf)
1209 return AVERROR(ENOMEM);
1214 s->remaining_fill = s->fft_len / 2;
1215 s->remaining_frac = 0;
1216 s->step_frac = av_div_q(av_make_q(inlink->sample_rate, s->count) , s->rate);
1217 s->step = (int)(s->step_frac.num / s->step_frac.den);
1218 s->step_frac.num %= s->step_frac.den;
1219 if (s->step_frac.num) {
1220 av_log(ctx, AV_LOG_INFO, "audio: %d Hz, step = %d + %d/%d.\n",
1221 inlink->sample_rate, s->step, s->step_frac.num, s->step_frac.den);
1222 av_log(ctx, AV_LOG_WARNING, "fractional step.\n");
1224 av_log(ctx, AV_LOG_INFO, "audio: %d Hz, step = %d.\n",
1225 inlink->sample_rate, s->step);
1232 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
1234 AVFilterContext *ctx = inlink->dst;
1235 AVFilterLink *outlink = ctx->outputs[0];
1236 ShowCQTContext *s = ctx->priv;
1237 int remaining, step, ret, x, i, j, m;
1239 AVFrame *out = NULL;
1242 while (s->remaining_fill < s->fft_len / 2) {
1243 memset(&s->fft_data[s->fft_len - s->remaining_fill], 0, sizeof(*s->fft_data) * s->remaining_fill);
1244 ret = plot_cqt(ctx, &out);
1248 step = s->step + (s->step_frac.num + s->remaining_frac) / s->step_frac.den;
1249 s->remaining_frac = (s->step_frac.num + s->remaining_frac) % s->step_frac.den;
1250 for (x = 0; x < (s->fft_len-step); x++)
1251 s->fft_data[x] = s->fft_data[x+step];
1252 s->remaining_fill += step;
1255 return ff_filter_frame(outlink, out);
1260 remaining = insamples->nb_samples;
1261 audio_data = (float*) insamples->data[0];
1264 i = insamples->nb_samples - remaining;
1265 j = s->fft_len - s->remaining_fill;
1266 if (remaining >= s->remaining_fill) {
1267 for (m = 0; m < s->remaining_fill; m++) {
1268 s->fft_data[j+m].re = audio_data[2*(i+m)];
1269 s->fft_data[j+m].im = audio_data[2*(i+m)+1];
1271 ret = plot_cqt(ctx, &out);
1273 av_frame_free(&insamples);
1276 remaining -= s->remaining_fill;
1278 int64_t pts = av_rescale_q(insamples->pts, inlink->time_base, av_make_q(1, inlink->sample_rate));
1279 pts += insamples->nb_samples - remaining - s->fft_len/2;
1280 pts = av_rescale_q(pts, av_make_q(1, inlink->sample_rate), outlink->time_base);
1281 if (FFABS(pts - out->pts) > PTS_TOLERANCE) {
1282 av_log(ctx, AV_LOG_DEBUG, "changing pts from %"PRId64" (%.3f) to %"PRId64" (%.3f).\n",
1283 out->pts, out->pts * av_q2d(outlink->time_base),
1284 pts, pts * av_q2d(outlink->time_base));
1286 s->next_pts = pts + PTS_STEP;
1288 ret = ff_filter_frame(outlink, out);
1290 av_frame_free(&insamples);
1295 step = s->step + (s->step_frac.num + s->remaining_frac) / s->step_frac.den;
1296 s->remaining_frac = (s->step_frac.num + s->remaining_frac) % s->step_frac.den;
1297 for (m = 0; m < s->fft_len-step; m++)
1298 s->fft_data[m] = s->fft_data[m+step];
1299 s->remaining_fill = step;
1301 for (m = 0; m < remaining; m++) {
1302 s->fft_data[j+m].re = audio_data[2*(i+m)];
1303 s->fft_data[j+m].im = audio_data[2*(i+m)+1];
1305 s->remaining_fill -= remaining;
1309 av_frame_free(&insamples);
1313 static int request_frame(AVFilterLink *outlink)
1315 AVFilterLink *inlink = outlink->src->inputs[0];
1318 ret = ff_request_frame(inlink);
1319 if (ret == AVERROR_EOF)
1320 ret = filter_frame(inlink, NULL);
1324 static const AVFilterPad showcqt_inputs[] = {
1327 .type = AVMEDIA_TYPE_AUDIO,
1328 .filter_frame = filter_frame,
1333 static const AVFilterPad showcqt_outputs[] = {
1336 .type = AVMEDIA_TYPE_VIDEO,
1337 .config_props = config_output,
1338 .request_frame = request_frame,
1343 AVFilter ff_avf_showcqt = {
1345 .description = NULL_IF_CONFIG_SMALL("Convert input audio to a CQT (Constant/Clamped Q Transform) spectrum video output."),
1348 .query_formats = query_formats,
1349 .priv_size = sizeof(ShowCQTContext),
1350 .inputs = showcqt_inputs,
1351 .outputs = showcqt_outputs,
1352 .priv_class = &showcqt_class,