1 /*
2 * Copyright (c) 2012 Clément Bœsch
3 *
4 * This file is part of FFmpeg.
5 *
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.
10 *
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.
15 *
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
19 */
20
21 /**
22 * @file
23 * EBU R.128 implementation
24 * @see http://tech.ebu.ch/loudness
25 * @see https://www.youtube.com/watch?v=iuEtQqC-Sqo "EBU R128 Introduction - Florian Camerer"
26 * @todo implement start/stop/reset through filter command injection
27 */
28
29 #include <math.h>
30
31 #include "libavutil/avassert.h"
32 #include "libavutil/avstring.h"
33 #include "libavutil/channel_layout.h"
34 #include "libavutil/dict.h"
35 #include "libavutil/ffmath.h"
36 #include "libavutil/xga_font_data.h"
37 #include "libavutil/opt.h"
38 #include "libavutil/timestamp.h"
39 #include "libswresample/swresample.h"
40 #include "audio.h"
41 #include "avfilter.h"
42 #include "filters.h"
43 #include "formats.h"
44 #include "internal.h"
45
46 #define ABS_THRES -70 ///< silence gate: we discard anything below this absolute (LUFS) threshold
47 #define ABS_UP_THRES 10 ///< upper loud limit to consider (ABS_THRES being the minimum)
48 #define HIST_GRAIN 100 ///< defines histogram precision
49 #define HIST_SIZE ((ABS_UP_THRES - ABS_THRES) * HIST_GRAIN + 1)
50
51 /**
52 * A histogram is an array of HIST_SIZE hist_entry storing all the energies
53 * recorded (with an accuracy of 1/HIST_GRAIN) of the loudnesses from ABS_THRES
54 * (at 0) to ABS_UP_THRES (at HIST_SIZE-1).
55 * This fixed-size system avoids the need of a list of energies growing
56 * infinitely over the time and is thus more scalable.
57 */
58 struct hist_entry {
59 unsigned count; ///< how many times the corresponding value occurred
60 double energy; ///< E = 10^((L + 0.691) / 10)
61 double loudness; ///< L = -0.691 + 10 * log10(E)
62 };
63
64 struct integrator {
65 double **cache; ///< window of filtered samples (N ms)
66 int cache_pos; ///< focus on the last added bin in the cache array
67 int cache_size;
68 double *sum; ///< sum of the last N ms filtered samples (cache content)
69 int filled; ///< 1 if the cache is completely filled, 0 otherwise
70 double rel_threshold; ///< relative threshold
71 double sum_kept_powers; ///< sum of the powers (weighted sums) above absolute threshold
72 int nb_kept_powers; ///< number of sum above absolute threshold
73 struct hist_entry *histogram; ///< histogram of the powers, used to compute LRA and I
74 };
75
76 struct rect { int x, y, w, h; };
77
78 typedef struct EBUR128Context {
79 const AVClass *class; ///< AVClass context for log and options purpose
80
81 /* peak metering */
82 int peak_mode; ///< enabled peak modes
83 double *true_peaks; ///< true peaks per channel
84 double *sample_peaks; ///< sample peaks per channel
85 double *true_peaks_per_frame; ///< true peaks in a frame per channel
86 #if CONFIG_SWRESAMPLE
87 SwrContext *swr_ctx; ///< over-sampling context for true peak metering
88 double *swr_buf; ///< resampled audio data for true peak metering
89 int swr_linesize;
90 #endif
91
92 /* video */
93 int do_video; ///< 1 if video output enabled, 0 otherwise
94 int w, h; ///< size of the video output
95 struct rect text; ///< rectangle for the LU legend on the left
96 struct rect graph; ///< rectangle for the main graph in the center
97 struct rect gauge; ///< rectangle for the gauge on the right
98 AVFrame *outpicref; ///< output picture reference, updated regularly
99 int meter; ///< select a EBU mode between +9 and +18
100 int scale_range; ///< the range of LU values according to the meter
101 int y_zero_lu; ///< the y value (pixel position) for 0 LU
102 int y_opt_max; ///< the y value (pixel position) for 1 LU
103 int y_opt_min; ///< the y value (pixel position) for -1 LU
104 int *y_line_ref; ///< y reference values for drawing the LU lines in the graph and the gauge
105
106 /* audio */
107 int nb_channels; ///< number of channels in the input
108 double *ch_weighting; ///< channel weighting mapping
109 int sample_count; ///< sample count used for refresh frequency, reset at refresh
110 int nb_samples; ///< number of samples to consume per single input frame
111 int idx_insample; ///< current sample position of processed samples in single input frame
112 AVFrame *insamples; ///< input samples reference, updated regularly
113
114 /* Filter caches.
115 * The mult by 3 in the following is for X[i], X[i-1] and X[i-2] */
116 double *x; ///< 3 input samples cache for each channel
117 double *y; ///< 3 pre-filter samples cache for each channel
118 double *z; ///< 3 RLB-filter samples cache for each channel
119 double pre_b[3]; ///< pre-filter numerator coefficients
120 double pre_a[3]; ///< pre-filter denominator coefficients
121 double rlb_b[3]; ///< rlb-filter numerator coefficients
122 double rlb_a[3]; ///< rlb-filter denominator coefficients
123
124 struct integrator i400; ///< 400ms integrator, used for Momentary loudness (M), and Integrated loudness (I)
125 struct integrator i3000; ///< 3s integrator, used for Short term loudness (S), and Loudness Range (LRA)
126
127 /* I and LRA specific */
128 double integrated_loudness; ///< integrated loudness in LUFS (I)
129 double loudness_range; ///< loudness range in LU (LRA)
130 double lra_low, lra_high; ///< low and high LRA values
131
132 /* misc */
133 int loglevel; ///< log level for frame logging
134 int metadata; ///< whether or not to inject loudness results in frames
135 int dual_mono; ///< whether or not to treat single channel input files as dual-mono
136 double pan_law; ///< pan law value used to calculate dual-mono measurements
137 int target; ///< target level in LUFS used to set relative zero LU in visualization
138 int gauge_type; ///< whether gauge shows momentary or short
139 int scale; ///< display scale type of statistics
140 } EBUR128Context;
141
142 enum {
143 PEAK_MODE_NONE = 0,
144 PEAK_MODE_SAMPLES_PEAKS = 1<<1,
145 PEAK_MODE_TRUE_PEAKS = 1<<2,
146 };
147
148 enum {
149 GAUGE_TYPE_MOMENTARY = 0,
150 GAUGE_TYPE_SHORTTERM = 1,
151 };
152
153 enum {
154 SCALE_TYPE_ABSOLUTE = 0,
155 SCALE_TYPE_RELATIVE = 1,
156 };
157
158 #define OFFSET(x) offsetof(EBUR128Context, x)
159 #define A AV_OPT_FLAG_AUDIO_PARAM
160 #define V AV_OPT_FLAG_VIDEO_PARAM
161 #define F AV_OPT_FLAG_FILTERING_PARAM
162 static const AVOption ebur128_options[] = {
163 { "video", "set video output", OFFSET(do_video), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, V|F },
164 { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x480"}, 0, 0, V|F },
165 { "meter", "set scale meter (+9 to +18)", OFFSET(meter), AV_OPT_TYPE_INT, {.i64 = 9}, 9, 18, V|F },
166 { "framelog", "force frame logging level", OFFSET(loglevel), AV_OPT_TYPE_INT, {.i64 = -1}, INT_MIN, INT_MAX, A|V|F, "level" },
167 { "info", "information logging level", 0, AV_OPT_TYPE_CONST, {.i64 = AV_LOG_INFO}, INT_MIN, INT_MAX, A|V|F, "level" },
168 { "verbose", "verbose logging level", 0, AV_OPT_TYPE_CONST, {.i64 = AV_LOG_VERBOSE}, INT_MIN, INT_MAX, A|V|F, "level" },
169 { "metadata", "inject metadata in the filtergraph", OFFSET(metadata), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, A|V|F },
170 { "peak", "set peak mode", OFFSET(peak_mode), AV_OPT_TYPE_FLAGS, {.i64 = PEAK_MODE_NONE}, 0, INT_MAX, A|F, "mode" },
171 { "none", "disable any peak mode", 0, AV_OPT_TYPE_CONST, {.i64 = PEAK_MODE_NONE}, INT_MIN, INT_MAX, A|F, "mode" },
172 { "sample", "enable peak-sample mode", 0, AV_OPT_TYPE_CONST, {.i64 = PEAK_MODE_SAMPLES_PEAKS}, INT_MIN, INT_MAX, A|F, "mode" },
173 { "true", "enable true-peak mode", 0, AV_OPT_TYPE_CONST, {.i64 = PEAK_MODE_TRUE_PEAKS}, INT_MIN, INT_MAX, A|F, "mode" },
174 { "dualmono", "treat mono input files as dual-mono", OFFSET(dual_mono), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, A|F },
175 { "panlaw", "set a specific pan law for dual-mono files", OFFSET(pan_law), AV_OPT_TYPE_DOUBLE, {.dbl = -3.01029995663978}, -10.0, 0.0, A|F },
176 { "target", "set a specific target level in LUFS (-23 to 0)", OFFSET(target), AV_OPT_TYPE_INT, {.i64 = -23}, -23, 0, V|F },
177 { "gauge", "set gauge display type", OFFSET(gauge_type), AV_OPT_TYPE_INT, {.i64 = 0 }, GAUGE_TYPE_MOMENTARY, GAUGE_TYPE_SHORTTERM, V|F, "gaugetype" },
178 { "momentary", "display momentary value", 0, AV_OPT_TYPE_CONST, {.i64 = GAUGE_TYPE_MOMENTARY}, INT_MIN, INT_MAX, V|F, "gaugetype" },
179 { "m", "display momentary value", 0, AV_OPT_TYPE_CONST, {.i64 = GAUGE_TYPE_MOMENTARY}, INT_MIN, INT_MAX, V|F, "gaugetype" },
180 { "shortterm", "display short-term value", 0, AV_OPT_TYPE_CONST, {.i64 = GAUGE_TYPE_SHORTTERM}, INT_MIN, INT_MAX, V|F, "gaugetype" },
181 { "s", "display short-term value", 0, AV_OPT_TYPE_CONST, {.i64 = GAUGE_TYPE_SHORTTERM}, INT_MIN, INT_MAX, V|F, "gaugetype" },
182 { "scale", "sets display method for the stats", OFFSET(scale), AV_OPT_TYPE_INT, {.i64 = 0}, SCALE_TYPE_ABSOLUTE, SCALE_TYPE_RELATIVE, V|F, "scaletype" },
183 { "absolute", "display absolute values (LUFS)", 0, AV_OPT_TYPE_CONST, {.i64 = SCALE_TYPE_ABSOLUTE}, INT_MIN, INT_MAX, V|F, "scaletype" },
184 { "LUFS", "display absolute values (LUFS)", 0, AV_OPT_TYPE_CONST, {.i64 = SCALE_TYPE_ABSOLUTE}, INT_MIN, INT_MAX, V|F, "scaletype" },
185 { "relative", "display values relative to target (LU)", 0, AV_OPT_TYPE_CONST, {.i64 = SCALE_TYPE_RELATIVE}, INT_MIN, INT_MAX, V|F, "scaletype" },
186 { "LU", "display values relative to target (LU)", 0, AV_OPT_TYPE_CONST, {.i64 = SCALE_TYPE_RELATIVE}, INT_MIN, INT_MAX, V|F, "scaletype" },
187 { NULL },
188 };
189
190 AVFILTER_DEFINE_CLASS(ebur128);
191
192 static const uint8_t graph_colors[] = {
193 0xdd, 0x66, 0x66, // value above 1LU non reached below -1LU (impossible)
194 0x66, 0x66, 0xdd, // value below 1LU non reached below -1LU
195 0x96, 0x33, 0x33, // value above 1LU reached below -1LU (impossible)
196 0x33, 0x33, 0x96, // value below 1LU reached below -1LU
197 0xdd, 0x96, 0x96, // value above 1LU line non reached below -1LU (impossible)
198 0x96, 0x96, 0xdd, // value below 1LU line non reached below -1LU
199 0xdd, 0x33, 0x33, // value above 1LU line reached below -1LU (impossible)
200 0x33, 0x33, 0xdd, // value below 1LU line reached below -1LU
201 0xdd, 0x66, 0x66, // value above 1LU non reached above -1LU
202 0x66, 0xdd, 0x66, // value below 1LU non reached above -1LU
203 0x96, 0x33, 0x33, // value above 1LU reached above -1LU
204 0x33, 0x96, 0x33, // value below 1LU reached above -1LU
205 0xdd, 0x96, 0x96, // value above 1LU line non reached above -1LU
206 0x96, 0xdd, 0x96, // value below 1LU line non reached above -1LU
207 0xdd, 0x33, 0x33, // value above 1LU line reached above -1LU
208 0x33, 0xdd, 0x33, // value below 1LU line reached above -1LU
209 };
210
get_graph_color(const EBUR128Context * ebur128,int v,int y)211 static const uint8_t *get_graph_color(const EBUR128Context *ebur128, int v, int y)
212 {
213 const int above_opt_max = y > ebur128->y_opt_max;
214 const int below_opt_min = y < ebur128->y_opt_min;
215 const int reached = y >= v;
216 const int line = ebur128->y_line_ref[y] || y == ebur128->y_zero_lu;
217 const int colorid = 8*below_opt_min+ 4*line + 2*reached + above_opt_max;
218 return graph_colors + 3*colorid;
219 }
220
lu_to_y(const EBUR128Context * ebur128,double v)221 static inline int lu_to_y(const EBUR128Context *ebur128, double v)
222 {
223 v += 2 * ebur128->meter; // make it in range [0;...]
224 v = av_clipf(v, 0, ebur128->scale_range); // make sure it's in the graph scale
225 v = ebur128->scale_range - v; // invert value (y=0 is on top)
226 return v * ebur128->graph.h / ebur128->scale_range; // rescale from scale range to px height
227 }
228
229 #define FONT8 0
230 #define FONT16 1
231
232 static const uint8_t font_colors[] = {
233 0xdd, 0xdd, 0x00,
234 0x00, 0x96, 0x96,
235 };
236
drawtext(AVFrame * pic,int x,int y,int ftid,const uint8_t * color,const char * fmt,...)237 static void drawtext(AVFrame *pic, int x, int y, int ftid, const uint8_t *color, const char *fmt, ...)
238 {
239 int i;
240 char buf[128] = {0};
241 const uint8_t *font;
242 int font_height;
243 va_list vl;
244
245 if (ftid == FONT16) font = avpriv_vga16_font, font_height = 16;
246 else if (ftid == FONT8) font = avpriv_cga_font, font_height = 8;
247 else return;
248
249 va_start(vl, fmt);
250 vsnprintf(buf, sizeof(buf), fmt, vl);
251 va_end(vl);
252
253 for (i = 0; buf[i]; i++) {
254 int char_y, mask;
255 uint8_t *p = pic->data[0] + y*pic->linesize[0] + (x + i*8)*3;
256
257 for (char_y = 0; char_y < font_height; char_y++) {
258 for (mask = 0x80; mask; mask >>= 1) {
259 if (font[buf[i] * font_height + char_y] & mask)
260 memcpy(p, color, 3);
261 else
262 memcpy(p, "\x00\x00\x00", 3);
263 p += 3;
264 }
265 p += pic->linesize[0] - 8*3;
266 }
267 }
268 }
269
drawline(AVFrame * pic,int x,int y,int len,int step)270 static void drawline(AVFrame *pic, int x, int y, int len, int step)
271 {
272 int i;
273 uint8_t *p = pic->data[0] + y*pic->linesize[0] + x*3;
274
275 for (i = 0; i < len; i++) {
276 memcpy(p, "\x00\xff\x00", 3);
277 p += step;
278 }
279 }
280
config_video_output(AVFilterLink * outlink)281 static int config_video_output(AVFilterLink *outlink)
282 {
283 int i, x, y;
284 uint8_t *p;
285 AVFilterContext *ctx = outlink->src;
286 AVFilterLink *inlink = ctx->inputs[0];
287 EBUR128Context *ebur128 = ctx->priv;
288 AVFrame *outpicref;
289
290 /* check if there is enough space to represent everything decently */
291 if (ebur128->w < 640 || ebur128->h < 480) {
292 av_log(ctx, AV_LOG_ERROR, "Video size %dx%d is too small, "
293 "minimum size is 640x480\n", ebur128->w, ebur128->h);
294 return AVERROR(EINVAL);
295 }
296 outlink->w = ebur128->w;
297 outlink->h = ebur128->h;
298 outlink->sample_aspect_ratio = (AVRational){1,1};
299 outlink->time_base = inlink->time_base;
300 outlink->frame_rate = av_make_q(10, 1);
301
302 #define PAD 8
303
304 /* configure text area position and size */
305 ebur128->text.x = PAD;
306 ebur128->text.y = 40;
307 ebur128->text.w = 3 * 8; // 3 characters
308 ebur128->text.h = ebur128->h - PAD - ebur128->text.y;
309
310 /* configure gauge position and size */
311 ebur128->gauge.w = 20;
312 ebur128->gauge.h = ebur128->text.h;
313 ebur128->gauge.x = ebur128->w - PAD - ebur128->gauge.w;
314 ebur128->gauge.y = ebur128->text.y;
315
316 /* configure graph position and size */
317 ebur128->graph.x = ebur128->text.x + ebur128->text.w + PAD;
318 ebur128->graph.y = ebur128->gauge.y;
319 ebur128->graph.w = ebur128->gauge.x - ebur128->graph.x - PAD;
320 ebur128->graph.h = ebur128->gauge.h;
321
322 /* graph and gauge share the LU-to-pixel code */
323 av_assert0(ebur128->graph.h == ebur128->gauge.h);
324
325 /* prepare the initial picref buffer */
326 av_frame_free(&ebur128->outpicref);
327 ebur128->outpicref = outpicref =
328 ff_get_video_buffer(outlink, outlink->w, outlink->h);
329 if (!outpicref)
330 return AVERROR(ENOMEM);
331 outpicref->sample_aspect_ratio = (AVRational){1,1};
332
333 /* init y references values (to draw LU lines) */
334 ebur128->y_line_ref = av_calloc(ebur128->graph.h + 1, sizeof(*ebur128->y_line_ref));
335 if (!ebur128->y_line_ref)
336 return AVERROR(ENOMEM);
337
338 /* black background */
339 memset(outpicref->data[0], 0, ebur128->h * outpicref->linesize[0]);
340
341 /* draw LU legends */
342 drawtext(outpicref, PAD, PAD+16, FONT8, font_colors+3, " LU");
343 for (i = ebur128->meter; i >= -ebur128->meter * 2; i--) {
344 y = lu_to_y(ebur128, i);
345 x = PAD + (i < 10 && i > -10) * 8;
346 ebur128->y_line_ref[y] = i;
347 y -= 4; // -4 to center vertically
348 drawtext(outpicref, x, y + ebur128->graph.y, FONT8, font_colors+3,
349 "%c%d", i < 0 ? '-' : i > 0 ? '+' : ' ', FFABS(i));
350 }
351
352 /* draw graph */
353 ebur128->y_zero_lu = lu_to_y(ebur128, 0);
354 ebur128->y_opt_max = lu_to_y(ebur128, 1);
355 ebur128->y_opt_min = lu_to_y(ebur128, -1);
356 p = outpicref->data[0] + ebur128->graph.y * outpicref->linesize[0]
357 + ebur128->graph.x * 3;
358 for (y = 0; y < ebur128->graph.h; y++) {
359 const uint8_t *c = get_graph_color(ebur128, INT_MAX, y);
360
361 for (x = 0; x < ebur128->graph.w; x++)
362 memcpy(p + x*3, c, 3);
363 p += outpicref->linesize[0];
364 }
365
366 /* draw fancy rectangles around the graph and the gauge */
367 #define DRAW_RECT(r) do { \
368 drawline(outpicref, r.x, r.y - 1, r.w, 3); \
369 drawline(outpicref, r.x, r.y + r.h, r.w, 3); \
370 drawline(outpicref, r.x - 1, r.y, r.h, outpicref->linesize[0]); \
371 drawline(outpicref, r.x + r.w, r.y, r.h, outpicref->linesize[0]); \
372 } while (0)
373 DRAW_RECT(ebur128->graph);
374 DRAW_RECT(ebur128->gauge);
375
376 return 0;
377 }
378
config_audio_input(AVFilterLink * inlink)379 static int config_audio_input(AVFilterLink *inlink)
380 {
381 AVFilterContext *ctx = inlink->dst;
382 EBUR128Context *ebur128 = ctx->priv;
383
384 /* Unofficial reversed parametrization of PRE
385 * and RLB from 48kHz */
386
387 double f0 = 1681.974450955533;
388 double G = 3.999843853973347;
389 double Q = 0.7071752369554196;
390
391 double K = tan(M_PI * f0 / (double)inlink->sample_rate);
392 double Vh = pow(10.0, G / 20.0);
393 double Vb = pow(Vh, 0.4996667741545416);
394
395 double a0 = 1.0 + K / Q + K * K;
396
397 ebur128->pre_b[0] = (Vh + Vb * K / Q + K * K) / a0;
398 ebur128->pre_b[1] = 2.0 * (K * K - Vh) / a0;
399 ebur128->pre_b[2] = (Vh - Vb * K / Q + K * K) / a0;
400 ebur128->pre_a[1] = 2.0 * (K * K - 1.0) / a0;
401 ebur128->pre_a[2] = (1.0 - K / Q + K * K) / a0;
402
403 f0 = 38.13547087602444;
404 Q = 0.5003270373238773;
405 K = tan(M_PI * f0 / (double)inlink->sample_rate);
406
407 ebur128->rlb_b[0] = 1.0;
408 ebur128->rlb_b[1] = -2.0;
409 ebur128->rlb_b[2] = 1.0;
410 ebur128->rlb_a[1] = 2.0 * (K * K - 1.0) / (1.0 + K / Q + K * K);
411 ebur128->rlb_a[2] = (1.0 - K / Q + K * K) / (1.0 + K / Q + K * K);
412
413 /* Force 100ms framing in case of metadata injection: the frames must have
414 * a granularity of the window overlap to be accurately exploited.
415 * As for the true peaks mode, it just simplifies the resampling buffer
416 * allocation and the lookup in it (since sample buffers differ in size, it
417 * can be more complex to integrate in the one-sample loop of
418 * filter_frame()). */
419 if (ebur128->metadata || (ebur128->peak_mode & PEAK_MODE_TRUE_PEAKS))
420 ebur128->nb_samples = inlink->sample_rate / 10;
421 return 0;
422 }
423
config_audio_output(AVFilterLink * outlink)424 static int config_audio_output(AVFilterLink *outlink)
425 {
426 int i;
427 AVFilterContext *ctx = outlink->src;
428 EBUR128Context *ebur128 = ctx->priv;
429 const int nb_channels = outlink->ch_layout.nb_channels;
430
431 #define BACK_MASK (AV_CH_BACK_LEFT |AV_CH_BACK_CENTER |AV_CH_BACK_RIGHT| \
432 AV_CH_TOP_BACK_LEFT|AV_CH_TOP_BACK_CENTER|AV_CH_TOP_BACK_RIGHT| \
433 AV_CH_SIDE_LEFT |AV_CH_SIDE_RIGHT| \
434 AV_CH_SURROUND_DIRECT_LEFT |AV_CH_SURROUND_DIRECT_RIGHT)
435
436 ebur128->nb_channels = nb_channels;
437 ebur128->x = av_calloc(nb_channels, 3 * sizeof(*ebur128->x));
438 ebur128->y = av_calloc(nb_channels, 3 * sizeof(*ebur128->y));
439 ebur128->z = av_calloc(nb_channels, 3 * sizeof(*ebur128->z));
440 ebur128->ch_weighting = av_calloc(nb_channels, sizeof(*ebur128->ch_weighting));
441 if (!ebur128->ch_weighting || !ebur128->x || !ebur128->y || !ebur128->z)
442 return AVERROR(ENOMEM);
443
444 #define I400_BINS(x) ((x) * 4 / 10)
445 #define I3000_BINS(x) ((x) * 3)
446
447 ebur128->i400.sum = av_calloc(nb_channels, sizeof(*ebur128->i400.sum));
448 ebur128->i3000.sum = av_calloc(nb_channels, sizeof(*ebur128->i3000.sum));
449 ebur128->i400.cache = av_calloc(nb_channels, sizeof(*ebur128->i400.cache));
450 ebur128->i3000.cache = av_calloc(nb_channels, sizeof(*ebur128->i3000.cache));
451 if (!ebur128->i400.sum || !ebur128->i3000.sum ||
452 !ebur128->i400.cache || !ebur128->i3000.cache)
453 return AVERROR(ENOMEM);
454
455 for (i = 0; i < nb_channels; i++) {
456 /* channel weighting */
457 const enum AVChannel chl = av_channel_layout_channel_from_index(&outlink->ch_layout, i);
458 if (chl == AV_CHAN_LOW_FREQUENCY || chl == AV_CHAN_LOW_FREQUENCY_2) {
459 ebur128->ch_weighting[i] = 0;
460 } else if (chl < 64 && (1ULL << chl) & BACK_MASK) {
461 ebur128->ch_weighting[i] = 1.41;
462 } else {
463 ebur128->ch_weighting[i] = 1.0;
464 }
465
466 if (!ebur128->ch_weighting[i])
467 continue;
468
469 /* bins buffer for the two integration window (400ms and 3s) */
470 ebur128->i400.cache_size = I400_BINS(outlink->sample_rate);
471 ebur128->i3000.cache_size = I3000_BINS(outlink->sample_rate);
472 ebur128->i400.cache[i] = av_calloc(ebur128->i400.cache_size, sizeof(*ebur128->i400.cache[0]));
473 ebur128->i3000.cache[i] = av_calloc(ebur128->i3000.cache_size, sizeof(*ebur128->i3000.cache[0]));
474 if (!ebur128->i400.cache[i] || !ebur128->i3000.cache[i])
475 return AVERROR(ENOMEM);
476 }
477
478 #if CONFIG_SWRESAMPLE
479 if (ebur128->peak_mode & PEAK_MODE_TRUE_PEAKS) {
480 int ret;
481
482 ebur128->swr_buf = av_malloc_array(nb_channels, 19200 * sizeof(double));
483 ebur128->true_peaks = av_calloc(nb_channels, sizeof(*ebur128->true_peaks));
484 ebur128->true_peaks_per_frame = av_calloc(nb_channels, sizeof(*ebur128->true_peaks_per_frame));
485 ebur128->swr_ctx = swr_alloc();
486 if (!ebur128->swr_buf || !ebur128->true_peaks ||
487 !ebur128->true_peaks_per_frame || !ebur128->swr_ctx)
488 return AVERROR(ENOMEM);
489
490 av_opt_set_chlayout(ebur128->swr_ctx, "in_chlayout", &outlink->ch_layout, 0);
491 av_opt_set_int(ebur128->swr_ctx, "in_sample_rate", outlink->sample_rate, 0);
492 av_opt_set_sample_fmt(ebur128->swr_ctx, "in_sample_fmt", outlink->format, 0);
493
494 av_opt_set_chlayout(ebur128->swr_ctx, "out_chlayout", &outlink->ch_layout, 0);
495 av_opt_set_int(ebur128->swr_ctx, "out_sample_rate", 192000, 0);
496 av_opt_set_sample_fmt(ebur128->swr_ctx, "out_sample_fmt", outlink->format, 0);
497
498 ret = swr_init(ebur128->swr_ctx);
499 if (ret < 0)
500 return ret;
501 }
502 #endif
503
504 if (ebur128->peak_mode & PEAK_MODE_SAMPLES_PEAKS) {
505 ebur128->sample_peaks = av_calloc(nb_channels, sizeof(*ebur128->sample_peaks));
506 if (!ebur128->sample_peaks)
507 return AVERROR(ENOMEM);
508 }
509
510 return 0;
511 }
512
513 #define ENERGY(loudness) (ff_exp10(((loudness) + 0.691) / 10.))
514 #define LOUDNESS(energy) (-0.691 + 10 * log10(energy))
515 #define DBFS(energy) (20 * log10(energy))
516
get_histogram(void)517 static struct hist_entry *get_histogram(void)
518 {
519 int i;
520 struct hist_entry *h = av_calloc(HIST_SIZE, sizeof(*h));
521
522 if (!h)
523 return NULL;
524 for (i = 0; i < HIST_SIZE; i++) {
525 h[i].loudness = i / (double)HIST_GRAIN + ABS_THRES;
526 h[i].energy = ENERGY(h[i].loudness);
527 }
528 return h;
529 }
530
init(AVFilterContext * ctx)531 static av_cold int init(AVFilterContext *ctx)
532 {
533 EBUR128Context *ebur128 = ctx->priv;
534 AVFilterPad pad;
535 int ret;
536
537 if (ebur128->loglevel != AV_LOG_INFO &&
538 ebur128->loglevel != AV_LOG_VERBOSE) {
539 if (ebur128->do_video || ebur128->metadata)
540 ebur128->loglevel = AV_LOG_VERBOSE;
541 else
542 ebur128->loglevel = AV_LOG_INFO;
543 }
544
545 if (!CONFIG_SWRESAMPLE && (ebur128->peak_mode & PEAK_MODE_TRUE_PEAKS)) {
546 av_log(ctx, AV_LOG_ERROR,
547 "True-peak mode requires libswresample to be performed\n");
548 return AVERROR(EINVAL);
549 }
550
551 // if meter is +9 scale, scale range is from -18 LU to +9 LU (or 3*9)
552 // if meter is +18 scale, scale range is from -36 LU to +18 LU (or 3*18)
553 ebur128->scale_range = 3 * ebur128->meter;
554
555 ebur128->i400.histogram = get_histogram();
556 ebur128->i3000.histogram = get_histogram();
557 if (!ebur128->i400.histogram || !ebur128->i3000.histogram)
558 return AVERROR(ENOMEM);
559
560 ebur128->integrated_loudness = ABS_THRES;
561 ebur128->loudness_range = 0;
562
563 /* insert output pads */
564 if (ebur128->do_video) {
565 pad = (AVFilterPad){
566 .name = "out0",
567 .type = AVMEDIA_TYPE_VIDEO,
568 .config_props = config_video_output,
569 };
570 ret = ff_append_outpad(ctx, &pad);
571 if (ret < 0)
572 return ret;
573 }
574 pad = (AVFilterPad){
575 .name = ebur128->do_video ? "out1" : "out0",
576 .type = AVMEDIA_TYPE_AUDIO,
577 .config_props = config_audio_output,
578 };
579 ret = ff_append_outpad(ctx, &pad);
580 if (ret < 0)
581 return ret;
582
583 /* summary */
584 av_log(ctx, AV_LOG_VERBOSE, "EBU +%d scale\n", ebur128->meter);
585
586 return 0;
587 }
588
589 #define HIST_POS(power) (int)(((power) - ABS_THRES) * HIST_GRAIN)
590
591 /* loudness and power should be set such as loudness = -0.691 +
592 * 10*log10(power), we just avoid doing that calculus two times */
gate_update(struct integrator * integ,double power,double loudness,int gate_thres)593 static int gate_update(struct integrator *integ, double power,
594 double loudness, int gate_thres)
595 {
596 int ipower;
597 double relative_threshold;
598 int gate_hist_pos;
599
600 /* update powers histograms by incrementing current power count */
601 ipower = av_clip(HIST_POS(loudness), 0, HIST_SIZE - 1);
602 integ->histogram[ipower].count++;
603
604 /* compute relative threshold and get its position in the histogram */
605 integ->sum_kept_powers += power;
606 integ->nb_kept_powers++;
607 relative_threshold = integ->sum_kept_powers / integ->nb_kept_powers;
608 if (!relative_threshold)
609 relative_threshold = 1e-12;
610 integ->rel_threshold = LOUDNESS(relative_threshold) + gate_thres;
611 gate_hist_pos = av_clip(HIST_POS(integ->rel_threshold), 0, HIST_SIZE - 1);
612
613 return gate_hist_pos;
614 }
615
filter_frame(AVFilterLink * inlink,AVFrame * insamples)616 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
617 {
618 int i, ch, idx_insample;
619 AVFilterContext *ctx = inlink->dst;
620 EBUR128Context *ebur128 = ctx->priv;
621 const int nb_channels = ebur128->nb_channels;
622 const int nb_samples = insamples->nb_samples;
623 const double *samples = (double *)insamples->data[0];
624 AVFrame *pic = ebur128->outpicref;
625
626 #if CONFIG_SWRESAMPLE
627 if (ebur128->peak_mode & PEAK_MODE_TRUE_PEAKS && ebur128->idx_insample == 0) {
628 const double *swr_samples = ebur128->swr_buf;
629 int ret = swr_convert(ebur128->swr_ctx, (uint8_t**)&ebur128->swr_buf, 19200,
630 (const uint8_t **)insamples->data, nb_samples);
631 if (ret < 0)
632 return ret;
633 for (ch = 0; ch < nb_channels; ch++)
634 ebur128->true_peaks_per_frame[ch] = 0.0;
635 for (idx_insample = 0; idx_insample < ret; idx_insample++) {
636 for (ch = 0; ch < nb_channels; ch++) {
637 ebur128->true_peaks[ch] = FFMAX(ebur128->true_peaks[ch], fabs(*swr_samples));
638 ebur128->true_peaks_per_frame[ch] = FFMAX(ebur128->true_peaks_per_frame[ch],
639 fabs(*swr_samples));
640 swr_samples++;
641 }
642 }
643 }
644 #endif
645
646 for (idx_insample = ebur128->idx_insample; idx_insample < nb_samples; idx_insample++) {
647 const int bin_id_400 = ebur128->i400.cache_pos;
648 const int bin_id_3000 = ebur128->i3000.cache_pos;
649
650 #define MOVE_TO_NEXT_CACHED_ENTRY(time) do { \
651 ebur128->i##time.cache_pos++; \
652 if (ebur128->i##time.cache_pos == \
653 ebur128->i##time.cache_size) { \
654 ebur128->i##time.filled = 1; \
655 ebur128->i##time.cache_pos = 0; \
656 } \
657 } while (0)
658
659 MOVE_TO_NEXT_CACHED_ENTRY(400);
660 MOVE_TO_NEXT_CACHED_ENTRY(3000);
661
662 for (ch = 0; ch < nb_channels; ch++) {
663 double bin;
664
665 if (ebur128->peak_mode & PEAK_MODE_SAMPLES_PEAKS)
666 ebur128->sample_peaks[ch] = FFMAX(ebur128->sample_peaks[ch], fabs(samples[idx_insample * nb_channels + ch]));
667
668 ebur128->x[ch * 3] = samples[idx_insample * nb_channels + ch]; // set X[i]
669
670 if (!ebur128->ch_weighting[ch])
671 continue;
672
673 /* Y[i] = X[i]*b0 + X[i-1]*b1 + X[i-2]*b2 - Y[i-1]*a1 - Y[i-2]*a2 */
674 #define FILTER(Y, X, NUM, DEN) do { \
675 double *dst = ebur128->Y + ch*3; \
676 double *src = ebur128->X + ch*3; \
677 dst[2] = dst[1]; \
678 dst[1] = dst[0]; \
679 dst[0] = src[0]*NUM[0] + src[1]*NUM[1] + src[2]*NUM[2] \
680 - dst[1]*DEN[1] - dst[2]*DEN[2]; \
681 } while (0)
682
683 // TODO: merge both filters in one?
684 FILTER(y, x, ebur128->pre_b, ebur128->pre_a); // apply pre-filter
685 ebur128->x[ch * 3 + 2] = ebur128->x[ch * 3 + 1];
686 ebur128->x[ch * 3 + 1] = ebur128->x[ch * 3 ];
687 FILTER(z, y, ebur128->rlb_b, ebur128->rlb_a); // apply RLB-filter
688
689 bin = ebur128->z[ch * 3] * ebur128->z[ch * 3];
690
691 /* add the new value, and limit the sum to the cache size (400ms or 3s)
692 * by removing the oldest one */
693 ebur128->i400.sum [ch] = ebur128->i400.sum [ch] + bin - ebur128->i400.cache [ch][bin_id_400];
694 ebur128->i3000.sum[ch] = ebur128->i3000.sum[ch] + bin - ebur128->i3000.cache[ch][bin_id_3000];
695
696 /* override old cache entry with the new value */
697 ebur128->i400.cache [ch][bin_id_400 ] = bin;
698 ebur128->i3000.cache[ch][bin_id_3000] = bin;
699 }
700
701 /* For integrated loudness, gating blocks are 400ms long with 75%
702 * overlap (see BS.1770-2 p5), so a re-computation is needed each 100ms
703 * (4800 samples at 48kHz). */
704 if (++ebur128->sample_count == inlink->sample_rate / 10) {
705 double loudness_400, loudness_3000;
706 double power_400 = 1e-12, power_3000 = 1e-12;
707 AVFilterLink *outlink = ctx->outputs[0];
708 const int64_t pts = insamples->pts +
709 av_rescale_q(idx_insample, (AVRational){ 1, inlink->sample_rate },
710 outlink->time_base);
711
712 ebur128->sample_count = 0;
713
714 #define COMPUTE_LOUDNESS(m, time) do { \
715 if (ebur128->i##time.filled) { \
716 /* weighting sum of the last <time> ms */ \
717 for (ch = 0; ch < nb_channels; ch++) \
718 power_##time += ebur128->ch_weighting[ch] * ebur128->i##time.sum[ch]; \
719 power_##time /= I##time##_BINS(inlink->sample_rate); \
720 } \
721 loudness_##time = LOUDNESS(power_##time); \
722 } while (0)
723
724 COMPUTE_LOUDNESS(M, 400);
725 COMPUTE_LOUDNESS(S, 3000);
726
727 /* Integrated loudness */
728 #define I_GATE_THRES -10 // initially defined to -8 LU in the first EBU standard
729
730 if (loudness_400 >= ABS_THRES) {
731 double integrated_sum = 0.0;
732 uint64_t nb_integrated = 0;
733 int gate_hist_pos = gate_update(&ebur128->i400, power_400,
734 loudness_400, I_GATE_THRES);
735
736 /* compute integrated loudness by summing the histogram values
737 * above the relative threshold */
738 for (i = gate_hist_pos; i < HIST_SIZE; i++) {
739 const unsigned nb_v = ebur128->i400.histogram[i].count;
740 nb_integrated += nb_v;
741 integrated_sum += nb_v * ebur128->i400.histogram[i].energy;
742 }
743 if (nb_integrated) {
744 ebur128->integrated_loudness = LOUDNESS(integrated_sum / nb_integrated);
745 /* dual-mono correction */
746 if (nb_channels == 1 && ebur128->dual_mono) {
747 ebur128->integrated_loudness -= ebur128->pan_law;
748 }
749 }
750 }
751
752 /* LRA */
753 #define LRA_GATE_THRES -20
754 #define LRA_LOWER_PRC 10
755 #define LRA_HIGHER_PRC 95
756
757 /* XXX: example code in EBU 3342 is ">=" but formula in BS.1770
758 * specs is ">" */
759 if (loudness_3000 >= ABS_THRES) {
760 uint64_t nb_powers = 0;
761 int gate_hist_pos = gate_update(&ebur128->i3000, power_3000,
762 loudness_3000, LRA_GATE_THRES);
763
764 for (i = gate_hist_pos; i < HIST_SIZE; i++)
765 nb_powers += ebur128->i3000.histogram[i].count;
766 if (nb_powers) {
767 uint64_t n, nb_pow;
768
769 /* get lower loudness to consider */
770 n = 0;
771 nb_pow = LRA_LOWER_PRC * nb_powers * 0.01 + 0.5;
772 for (i = gate_hist_pos; i < HIST_SIZE; i++) {
773 n += ebur128->i3000.histogram[i].count;
774 if (n >= nb_pow) {
775 ebur128->lra_low = ebur128->i3000.histogram[i].loudness;
776 break;
777 }
778 }
779
780 /* get higher loudness to consider */
781 n = nb_powers;
782 nb_pow = LRA_HIGHER_PRC * nb_powers * 0.01 + 0.5;
783 for (i = HIST_SIZE - 1; i >= 0; i--) {
784 n -= FFMIN(n, ebur128->i3000.histogram[i].count);
785 if (n < nb_pow) {
786 ebur128->lra_high = ebur128->i3000.histogram[i].loudness;
787 break;
788 }
789 }
790
791 // XXX: show low & high on the graph?
792 ebur128->loudness_range = ebur128->lra_high - ebur128->lra_low;
793 }
794 }
795
796 /* dual-mono correction */
797 if (nb_channels == 1 && ebur128->dual_mono) {
798 loudness_400 -= ebur128->pan_law;
799 loudness_3000 -= ebur128->pan_law;
800 }
801
802 #define LOG_FMT "TARGET:%d LUFS M:%6.1f S:%6.1f I:%6.1f %s LRA:%6.1f LU"
803
804 /* push one video frame */
805 if (ebur128->do_video) {
806 AVFrame *clone;
807 int x, y;
808 uint8_t *p;
809 double gauge_value;
810 int y_loudness_lu_graph, y_loudness_lu_gauge;
811
812 if (ebur128->gauge_type == GAUGE_TYPE_MOMENTARY) {
813 gauge_value = loudness_400 - ebur128->target;
814 } else {
815 gauge_value = loudness_3000 - ebur128->target;
816 }
817
818 y_loudness_lu_graph = lu_to_y(ebur128, loudness_3000 - ebur128->target);
819 y_loudness_lu_gauge = lu_to_y(ebur128, gauge_value);
820
821 av_frame_make_writable(pic);
822 /* draw the graph using the short-term loudness */
823 p = pic->data[0] + ebur128->graph.y*pic->linesize[0] + ebur128->graph.x*3;
824 for (y = 0; y < ebur128->graph.h; y++) {
825 const uint8_t *c = get_graph_color(ebur128, y_loudness_lu_graph, y);
826
827 memmove(p, p + 3, (ebur128->graph.w - 1) * 3);
828 memcpy(p + (ebur128->graph.w - 1) * 3, c, 3);
829 p += pic->linesize[0];
830 }
831
832 /* draw the gauge using either momentary or short-term loudness */
833 p = pic->data[0] + ebur128->gauge.y*pic->linesize[0] + ebur128->gauge.x*3;
834 for (y = 0; y < ebur128->gauge.h; y++) {
835 const uint8_t *c = get_graph_color(ebur128, y_loudness_lu_gauge, y);
836
837 for (x = 0; x < ebur128->gauge.w; x++)
838 memcpy(p + x*3, c, 3);
839 p += pic->linesize[0];
840 }
841
842 /* draw textual info */
843 if (ebur128->scale == SCALE_TYPE_ABSOLUTE) {
844 drawtext(pic, PAD, PAD - PAD/2, FONT16, font_colors,
845 LOG_FMT " ", // padding to erase trailing characters
846 ebur128->target, loudness_400, loudness_3000,
847 ebur128->integrated_loudness, "LUFS", ebur128->loudness_range);
848 } else {
849 drawtext(pic, PAD, PAD - PAD/2, FONT16, font_colors,
850 LOG_FMT " ", // padding to erase trailing characters
851 ebur128->target, loudness_400-ebur128->target, loudness_3000-ebur128->target,
852 ebur128->integrated_loudness-ebur128->target, "LU", ebur128->loudness_range);
853 }
854
855 /* set pts and push frame */
856 pic->pts = pts;
857 clone = av_frame_clone(pic);
858 if (!clone)
859 return AVERROR(ENOMEM);
860 ebur128->idx_insample = idx_insample + 1;
861 ff_filter_set_ready(ctx, 100);
862 return ff_filter_frame(outlink, clone);
863 }
864
865 if (ebur128->metadata) { /* happens only once per filter_frame call */
866 char metabuf[128];
867 #define META_PREFIX "lavfi.r128."
868
869 #define SET_META(name, var) do { \
870 snprintf(metabuf, sizeof(metabuf), "%.3f", var); \
871 av_dict_set(&insamples->metadata, name, metabuf, 0); \
872 } while (0)
873
874 #define SET_META_PEAK(name, ptype) do { \
875 if (ebur128->peak_mode & PEAK_MODE_ ## ptype ## _PEAKS) { \
876 double max_peak = 0.0; \
877 char key[64]; \
878 for (ch = 0; ch < nb_channels; ch++) { \
879 snprintf(key, sizeof(key), \
880 META_PREFIX AV_STRINGIFY(name) "_peaks_ch%d", ch); \
881 max_peak = fmax(max_peak, ebur128->name##_peaks[ch]); \
882 SET_META(key, ebur128->name##_peaks[ch]); \
883 } \
884 snprintf(key, sizeof(key), \
885 META_PREFIX AV_STRINGIFY(name) "_peak"); \
886 SET_META(key, max_peak); \
887 } \
888 } while (0)
889
890 SET_META(META_PREFIX "M", loudness_400);
891 SET_META(META_PREFIX "S", loudness_3000);
892 SET_META(META_PREFIX "I", ebur128->integrated_loudness);
893 SET_META(META_PREFIX "LRA", ebur128->loudness_range);
894 SET_META(META_PREFIX "LRA.low", ebur128->lra_low);
895 SET_META(META_PREFIX "LRA.high", ebur128->lra_high);
896
897 SET_META_PEAK(sample, SAMPLES);
898 SET_META_PEAK(true, TRUE);
899 }
900
901 if (ebur128->scale == SCALE_TYPE_ABSOLUTE) {
902 av_log(ctx, ebur128->loglevel, "t: %-10s " LOG_FMT,
903 av_ts2timestr(pts, &outlink->time_base),
904 ebur128->target, loudness_400, loudness_3000,
905 ebur128->integrated_loudness, "LUFS", ebur128->loudness_range);
906 } else {
907 av_log(ctx, ebur128->loglevel, "t: %-10s " LOG_FMT,
908 av_ts2timestr(pts, &outlink->time_base),
909 ebur128->target, loudness_400-ebur128->target, loudness_3000-ebur128->target,
910 ebur128->integrated_loudness-ebur128->target, "LU", ebur128->loudness_range);
911 }
912
913 #define PRINT_PEAKS(str, sp, ptype) do { \
914 if (ebur128->peak_mode & PEAK_MODE_ ## ptype ## _PEAKS) { \
915 av_log(ctx, ebur128->loglevel, " " str ":"); \
916 for (ch = 0; ch < nb_channels; ch++) \
917 av_log(ctx, ebur128->loglevel, " %5.1f", DBFS(sp[ch])); \
918 av_log(ctx, ebur128->loglevel, " dBFS"); \
919 } \
920 } while (0)
921
922 PRINT_PEAKS("SPK", ebur128->sample_peaks, SAMPLES);
923 PRINT_PEAKS("FTPK", ebur128->true_peaks_per_frame, TRUE);
924 PRINT_PEAKS("TPK", ebur128->true_peaks, TRUE);
925 av_log(ctx, ebur128->loglevel, "\n");
926
927 }
928 }
929
930 ebur128->idx_insample = 0;
931 ebur128->insamples = NULL;
932
933 return ff_filter_frame(ctx->outputs[ebur128->do_video], insamples);
934 }
935
activate(AVFilterContext * ctx)936 static int activate(AVFilterContext *ctx)
937 {
938 AVFilterLink *inlink = ctx->inputs[0];
939 EBUR128Context *ebur128 = ctx->priv;
940 AVFilterLink *voutlink = ctx->outputs[0];
941 AVFilterLink *outlink = ctx->outputs[ebur128->do_video];
942 int ret;
943
944 FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
945 if (ebur128->do_video)
946 FF_FILTER_FORWARD_STATUS_BACK(voutlink, inlink);
947
948 if (!ebur128->insamples) {
949 AVFrame *in;
950
951 if (ebur128->nb_samples > 0) {
952 ret = ff_inlink_consume_samples(inlink, ebur128->nb_samples, ebur128->nb_samples, &in);
953 } else {
954 ret = ff_inlink_consume_frame(inlink, &in);
955 }
956 if (ret < 0)
957 return ret;
958 if (ret > 0)
959 ebur128->insamples = in;
960 }
961
962 if (ebur128->insamples)
963 ret = filter_frame(inlink, ebur128->insamples);
964
965 FF_FILTER_FORWARD_STATUS_ALL(inlink, ctx);
966 FF_FILTER_FORWARD_WANTED(outlink, inlink);
967 if (ebur128->do_video)
968 FF_FILTER_FORWARD_WANTED(voutlink, inlink);
969
970 return ret;
971 }
972
query_formats(AVFilterContext * ctx)973 static int query_formats(AVFilterContext *ctx)
974 {
975 EBUR128Context *ebur128 = ctx->priv;
976 AVFilterFormats *formats;
977 AVFilterChannelLayouts *layouts;
978 AVFilterLink *inlink = ctx->inputs[0];
979 AVFilterLink *outlink = ctx->outputs[0];
980 int ret;
981
982 static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_NONE };
983 static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_RGB24, AV_PIX_FMT_NONE };
984
985 /* set optional output video format */
986 if (ebur128->do_video) {
987 formats = ff_make_format_list(pix_fmts);
988 if ((ret = ff_formats_ref(formats, &outlink->incfg.formats)) < 0)
989 return ret;
990 outlink = ctx->outputs[1];
991 }
992
993 /* set input and output audio formats
994 * Note: ff_set_common_* functions are not used because they affect all the
995 * links, and thus break the video format negotiation */
996 formats = ff_make_format_list(sample_fmts);
997 if ((ret = ff_formats_ref(formats, &inlink->outcfg.formats)) < 0 ||
998 (ret = ff_formats_ref(formats, &outlink->incfg.formats)) < 0)
999 return ret;
1000
1001 layouts = ff_all_channel_layouts();
1002 if ((ret = ff_channel_layouts_ref(layouts, &inlink->outcfg.channel_layouts)) < 0 ||
1003 (ret = ff_channel_layouts_ref(layouts, &outlink->incfg.channel_layouts)) < 0)
1004 return ret;
1005
1006 formats = ff_all_samplerates();
1007 if ((ret = ff_formats_ref(formats, &inlink->outcfg.samplerates)) < 0 ||
1008 (ret = ff_formats_ref(formats, &outlink->incfg.samplerates)) < 0)
1009 return ret;
1010
1011 return 0;
1012 }
1013
uninit(AVFilterContext * ctx)1014 static av_cold void uninit(AVFilterContext *ctx)
1015 {
1016 int i;
1017 EBUR128Context *ebur128 = ctx->priv;
1018
1019 /* dual-mono correction */
1020 if (ebur128->nb_channels == 1 && ebur128->dual_mono) {
1021 ebur128->i400.rel_threshold -= ebur128->pan_law;
1022 ebur128->i3000.rel_threshold -= ebur128->pan_law;
1023 ebur128->lra_low -= ebur128->pan_law;
1024 ebur128->lra_high -= ebur128->pan_law;
1025 }
1026
1027 av_log(ctx, AV_LOG_INFO, "Summary:\n\n"
1028 " Integrated loudness:\n"
1029 " I: %5.1f LUFS\n"
1030 " Threshold: %5.1f LUFS\n\n"
1031 " Loudness range:\n"
1032 " LRA: %5.1f LU\n"
1033 " Threshold: %5.1f LUFS\n"
1034 " LRA low: %5.1f LUFS\n"
1035 " LRA high: %5.1f LUFS",
1036 ebur128->integrated_loudness, ebur128->i400.rel_threshold,
1037 ebur128->loudness_range, ebur128->i3000.rel_threshold,
1038 ebur128->lra_low, ebur128->lra_high);
1039
1040 #define PRINT_PEAK_SUMMARY(str, sp, ptype) do { \
1041 int ch; \
1042 double maxpeak; \
1043 maxpeak = 0.0; \
1044 if (ebur128->peak_mode & PEAK_MODE_ ## ptype ## _PEAKS) { \
1045 for (ch = 0; ch < ebur128->nb_channels; ch++) \
1046 maxpeak = FFMAX(maxpeak, sp[ch]); \
1047 av_log(ctx, AV_LOG_INFO, "\n\n " str " peak:\n" \
1048 " Peak: %5.1f dBFS", \
1049 DBFS(maxpeak)); \
1050 } \
1051 } while (0)
1052
1053 PRINT_PEAK_SUMMARY("Sample", ebur128->sample_peaks, SAMPLES);
1054 PRINT_PEAK_SUMMARY("True", ebur128->true_peaks, TRUE);
1055 av_log(ctx, AV_LOG_INFO, "\n");
1056
1057 av_freep(&ebur128->y_line_ref);
1058 av_freep(&ebur128->x);
1059 av_freep(&ebur128->y);
1060 av_freep(&ebur128->z);
1061 av_freep(&ebur128->ch_weighting);
1062 av_freep(&ebur128->true_peaks);
1063 av_freep(&ebur128->sample_peaks);
1064 av_freep(&ebur128->true_peaks_per_frame);
1065 av_freep(&ebur128->i400.sum);
1066 av_freep(&ebur128->i3000.sum);
1067 av_freep(&ebur128->i400.histogram);
1068 av_freep(&ebur128->i3000.histogram);
1069 for (i = 0; i < ebur128->nb_channels; i++) {
1070 if (ebur128->i400.cache)
1071 av_freep(&ebur128->i400.cache[i]);
1072 if (ebur128->i3000.cache)
1073 av_freep(&ebur128->i3000.cache[i]);
1074 }
1075 av_freep(&ebur128->i400.cache);
1076 av_freep(&ebur128->i3000.cache);
1077 av_frame_free(&ebur128->outpicref);
1078 #if CONFIG_SWRESAMPLE
1079 av_freep(&ebur128->swr_buf);
1080 swr_free(&ebur128->swr_ctx);
1081 #endif
1082 }
1083
1084 static const AVFilterPad ebur128_inputs[] = {
1085 {
1086 .name = "default",
1087 .type = AVMEDIA_TYPE_AUDIO,
1088 .config_props = config_audio_input,
1089 },
1090 };
1091
1092 const AVFilter ff_af_ebur128 = {
1093 .name = "ebur128",
1094 .description = NULL_IF_CONFIG_SMALL("EBU R128 scanner."),
1095 .priv_size = sizeof(EBUR128Context),
1096 .init = init,
1097 .uninit = uninit,
1098 .activate = activate,
1099 FILTER_INPUTS(ebur128_inputs),
1100 .outputs = NULL,
1101 FILTER_QUERY_FUNC(query_formats),
1102 .priv_class = &ebur128_class,
1103 .flags = AVFILTER_FLAG_DYNAMIC_OUTPUTS,
1104 };
1105