• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2001-2010 Krzysztof Foltman, Markus Schmidt, Thor Harald Johansen, Damien Zammit
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  * Audio (Sidechain) Gate filter
24  */
25 
26 #include "libavutil/audio_fifo.h"
27 #include "libavutil/avassert.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/opt.h"
30 #include "avfilter.h"
31 #include "audio.h"
32 #include "filters.h"
33 #include "formats.h"
34 #include "hermite.h"
35 
36 typedef struct AudioGateContext {
37     const AVClass *class;
38 
39     double level_in;
40     double level_sc;
41     double attack;
42     double release;
43     double threshold;
44     double ratio;
45     double knee;
46     double makeup;
47     double range;
48     int link;
49     int detection;
50     int mode;
51 
52     double thres;
53     double knee_start;
54     double knee_stop;
55     double lin_knee_start;
56     double lin_knee_stop;
57     double lin_slope;
58     double attack_coeff;
59     double release_coeff;
60 
61     AVAudioFifo *fifo[2];
62     int64_t pts;
63 } AudioGateContext;
64 
65 #define OFFSET(x) offsetof(AudioGateContext, x)
66 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
67 
68 static const AVOption options[] = {
69     { "level_in",  "set input level",        OFFSET(level_in),  AV_OPT_TYPE_DOUBLE, {.dbl=1},           0.015625,   64, A },
70     { "mode",      "set mode",               OFFSET(mode),      AV_OPT_TYPE_INT,    {.i64=0},           0, 1, A, "mode" },
71     {   "downward",0,                        0,                 AV_OPT_TYPE_CONST,  {.i64=0},           0, 0, A, "mode" },
72     {   "upward",  0,                        0,                 AV_OPT_TYPE_CONST,  {.i64=1},           0, 0, A, "mode" },
73     { "range",     "set max gain reduction", OFFSET(range),     AV_OPT_TYPE_DOUBLE, {.dbl=0.06125},     0, 1, A },
74     { "threshold", "set threshold",          OFFSET(threshold), AV_OPT_TYPE_DOUBLE, {.dbl=0.125},       0, 1, A },
75     { "ratio",     "set ratio",              OFFSET(ratio),     AV_OPT_TYPE_DOUBLE, {.dbl=2},           1,  9000, A },
76     { "attack",    "set attack",             OFFSET(attack),    AV_OPT_TYPE_DOUBLE, {.dbl=20},          0.01, 9000, A },
77     { "release",   "set release",            OFFSET(release),   AV_OPT_TYPE_DOUBLE, {.dbl=250},         0.01, 9000, A },
78     { "makeup",    "set makeup gain",        OFFSET(makeup),    AV_OPT_TYPE_DOUBLE, {.dbl=1},           1,   64, A },
79     { "knee",      "set knee",               OFFSET(knee),      AV_OPT_TYPE_DOUBLE, {.dbl=2.828427125}, 1,    8, A },
80     { "detection", "set detection",          OFFSET(detection), AV_OPT_TYPE_INT,    {.i64=1},           0,    1, A, "detection" },
81     {   "peak",    0,                        0,                 AV_OPT_TYPE_CONST,  {.i64=0},           0,    0, A, "detection" },
82     {   "rms",     0,                        0,                 AV_OPT_TYPE_CONST,  {.i64=1},           0,    0, A, "detection" },
83     { "link",      "set link",               OFFSET(link),      AV_OPT_TYPE_INT,    {.i64=0},           0,    1, A, "link" },
84     {   "average", 0,                        0,                 AV_OPT_TYPE_CONST,  {.i64=0},           0,    0, A, "link" },
85     {   "maximum", 0,                        0,                 AV_OPT_TYPE_CONST,  {.i64=1},           0,    0, A, "link" },
86     { "level_sc",  "set sidechain gain",     OFFSET(level_sc),  AV_OPT_TYPE_DOUBLE, {.dbl=1},           0.015625,   64, A },
87     { NULL }
88 };
89 
agate_config_input(AVFilterLink * inlink)90 static int agate_config_input(AVFilterLink *inlink)
91 {
92     AVFilterContext *ctx = inlink->dst;
93     AudioGateContext *s = ctx->priv;
94     double lin_threshold = s->threshold;
95     double lin_knee_sqrt = sqrt(s->knee);
96 
97     if (s->detection)
98         lin_threshold *= lin_threshold;
99 
100     s->attack_coeff  = FFMIN(1., 1. / (s->attack * inlink->sample_rate / 4000.));
101     s->release_coeff = FFMIN(1., 1. / (s->release * inlink->sample_rate / 4000.));
102     s->lin_knee_stop = lin_threshold * lin_knee_sqrt;
103     s->lin_knee_start = lin_threshold / lin_knee_sqrt;
104     s->thres = log(lin_threshold);
105     s->knee_start = log(s->lin_knee_start);
106     s->knee_stop = log(s->lin_knee_stop);
107 
108     return 0;
109 }
110 
111 // A fake infinity value (because real infinity may break some hosts)
112 #define FAKE_INFINITY (65536.0 * 65536.0)
113 
114 // Check for infinity (with appropriate-ish tolerance)
115 #define IS_FAKE_INFINITY(value) (fabs(value-FAKE_INFINITY) < 1.0)
116 
output_gain(double lin_slope,double ratio,double thres,double knee,double knee_start,double knee_stop,double range,int mode)117 static double output_gain(double lin_slope, double ratio, double thres,
118                           double knee, double knee_start, double knee_stop,
119                           double range, int mode)
120 {
121     double slope = log(lin_slope);
122     double tratio = ratio;
123     double gain = 0.;
124     double delta = 0.;
125 
126     if (IS_FAKE_INFINITY(ratio))
127         tratio = 1000.;
128     gain = (slope - thres) * tratio + thres;
129     delta = tratio;
130 
131     if (mode) {
132         if (knee > 1. && slope < knee_stop)
133             gain = hermite_interpolation(slope, knee_stop, knee_start, ((knee_stop - thres) * tratio  + thres), knee_start, delta, 1.);
134     } else {
135         if (knee > 1. && slope > knee_start)
136             gain = hermite_interpolation(slope, knee_start, knee_stop, ((knee_start - thres) * tratio  + thres), knee_stop, delta, 1.);
137     }
138     return FFMAX(range, exp(gain - slope));
139 }
140 
gate(AudioGateContext * s,const double * src,double * dst,const double * scsrc,int nb_samples,double level_in,double level_sc,AVFilterLink * inlink,AVFilterLink * sclink)141 static void gate(AudioGateContext *s,
142                  const double *src, double *dst, const double *scsrc,
143                  int nb_samples, double level_in, double level_sc,
144                  AVFilterLink *inlink, AVFilterLink *sclink)
145 {
146     const double makeup = s->makeup;
147     const double attack_coeff = s->attack_coeff;
148     const double release_coeff = s->release_coeff;
149     int n, c;
150 
151     for (n = 0; n < nb_samples; n++, src += inlink->channels, dst += inlink->channels, scsrc += sclink->channels) {
152         double abs_sample = fabs(scsrc[0] * level_sc), gain = 1.0;
153         int detected;
154 
155         if (s->link == 1) {
156             for (c = 1; c < sclink->channels; c++)
157                 abs_sample = FFMAX(fabs(scsrc[c] * level_sc), abs_sample);
158         } else {
159             for (c = 1; c < sclink->channels; c++)
160                 abs_sample += fabs(scsrc[c] * level_sc);
161 
162             abs_sample /= sclink->channels;
163         }
164 
165         if (s->detection)
166             abs_sample *= abs_sample;
167 
168         s->lin_slope += (abs_sample - s->lin_slope) * (abs_sample > s->lin_slope ? attack_coeff : release_coeff);
169 
170         if (s->mode)
171             detected = s->lin_slope > s->lin_knee_start;
172         else
173             detected = s->lin_slope < s->lin_knee_stop;
174 
175         if (s->lin_slope > 0.0 && detected)
176             gain = output_gain(s->lin_slope, s->ratio, s->thres,
177                                s->knee, s->knee_start, s->knee_stop,
178                                s->range, s->mode);
179 
180         for (c = 0; c < inlink->channels; c++)
181             dst[c] = src[c] * level_in * gain * makeup;
182     }
183 }
184 
185 #if CONFIG_AGATE_FILTER
186 
187 #define agate_options options
188 AVFILTER_DEFINE_CLASS(agate);
189 
query_formats(AVFilterContext * ctx)190 static int query_formats(AVFilterContext *ctx)
191 {
192     AVFilterFormats *formats = NULL;
193     AVFilterChannelLayouts *layouts;
194     int ret;
195 
196     if ((ret = ff_add_format(&formats, AV_SAMPLE_FMT_DBL)) < 0)
197         return ret;
198     ret = ff_set_common_formats(ctx, formats);
199     if (ret < 0)
200         return ret;
201 
202     layouts = ff_all_channel_counts();
203     if (!layouts)
204         return AVERROR(ENOMEM);
205     ret = ff_set_common_channel_layouts(ctx, layouts);
206     if (ret < 0)
207         return ret;
208 
209     formats = ff_all_samplerates();
210     if (!formats)
211         return AVERROR(ENOMEM);
212 
213     return ff_set_common_samplerates(ctx, formats);
214 }
215 
filter_frame(AVFilterLink * inlink,AVFrame * in)216 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
217 {
218     const double *src = (const double *)in->data[0];
219     AVFilterContext *ctx = inlink->dst;
220     AVFilterLink *outlink = ctx->outputs[0];
221     AudioGateContext *s = ctx->priv;
222     AVFrame *out;
223     double *dst;
224 
225     if (av_frame_is_writable(in)) {
226         out = in;
227     } else {
228         out = ff_get_audio_buffer(outlink, in->nb_samples);
229         if (!out) {
230             av_frame_free(&in);
231             return AVERROR(ENOMEM);
232         }
233         av_frame_copy_props(out, in);
234     }
235     dst = (double *)out->data[0];
236 
237     gate(s, src, dst, src, in->nb_samples,
238          s->level_in, s->level_in, inlink, inlink);
239 
240     if (out != in)
241         av_frame_free(&in);
242     return ff_filter_frame(outlink, out);
243 }
244 
245 static const AVFilterPad inputs[] = {
246     {
247         .name         = "default",
248         .type         = AVMEDIA_TYPE_AUDIO,
249         .filter_frame = filter_frame,
250         .config_props = agate_config_input,
251     },
252     { NULL }
253 };
254 
255 static const AVFilterPad outputs[] = {
256     {
257         .name = "default",
258         .type = AVMEDIA_TYPE_AUDIO,
259     },
260     { NULL }
261 };
262 
263 AVFilter ff_af_agate = {
264     .name           = "agate",
265     .description    = NULL_IF_CONFIG_SMALL("Audio gate."),
266     .query_formats  = query_formats,
267     .priv_size      = sizeof(AudioGateContext),
268     .priv_class     = &agate_class,
269     .inputs         = inputs,
270     .outputs        = outputs,
271     .process_command = ff_filter_process_command,
272     .flags          = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
273 };
274 
275 #endif /* CONFIG_AGATE_FILTER */
276 
277 #if CONFIG_SIDECHAINGATE_FILTER
278 
279 #define sidechaingate_options options
280 AVFILTER_DEFINE_CLASS(sidechaingate);
281 
activate(AVFilterContext * ctx)282 static int activate(AVFilterContext *ctx)
283 {
284     AudioGateContext *s = ctx->priv;
285     AVFrame *out = NULL, *in[2] = { NULL };
286     int ret, i, nb_samples;
287     double *dst;
288 
289     FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx);
290     if ((ret = ff_inlink_consume_frame(ctx->inputs[0], &in[0])) > 0) {
291         av_audio_fifo_write(s->fifo[0], (void **)in[0]->extended_data,
292                             in[0]->nb_samples);
293         av_frame_free(&in[0]);
294     }
295     if (ret < 0)
296         return ret;
297     if ((ret = ff_inlink_consume_frame(ctx->inputs[1], &in[1])) > 0) {
298         av_audio_fifo_write(s->fifo[1], (void **)in[1]->extended_data,
299                             in[1]->nb_samples);
300         av_frame_free(&in[1]);
301     }
302     if (ret < 0)
303         return ret;
304 
305     nb_samples = FFMIN(av_audio_fifo_size(s->fifo[0]), av_audio_fifo_size(s->fifo[1]));
306     if (nb_samples) {
307         out = ff_get_audio_buffer(ctx->outputs[0], nb_samples);
308         if (!out)
309             return AVERROR(ENOMEM);
310         for (i = 0; i < 2; i++) {
311             in[i] = ff_get_audio_buffer(ctx->inputs[i], nb_samples);
312             if (!in[i]) {
313                 av_frame_free(&in[0]);
314                 av_frame_free(&in[1]);
315                 av_frame_free(&out);
316                 return AVERROR(ENOMEM);
317             }
318             av_audio_fifo_read(s->fifo[i], (void **)in[i]->data, nb_samples);
319         }
320 
321         dst = (double *)out->data[0];
322         out->pts = s->pts;
323         s->pts += av_rescale_q(nb_samples, (AVRational){1, ctx->outputs[0]->sample_rate}, ctx->outputs[0]->time_base);
324 
325         gate(s, (double *)in[0]->data[0], dst,
326              (double *)in[1]->data[0], nb_samples,
327              s->level_in, s->level_sc,
328              ctx->inputs[0], ctx->inputs[1]);
329 
330         av_frame_free(&in[0]);
331         av_frame_free(&in[1]);
332 
333         ret = ff_filter_frame(ctx->outputs[0], out);
334         if (ret < 0)
335             return ret;
336     }
337     FF_FILTER_FORWARD_STATUS(ctx->inputs[0], ctx->outputs[0]);
338     FF_FILTER_FORWARD_STATUS(ctx->inputs[1], ctx->outputs[0]);
339     if (ff_outlink_frame_wanted(ctx->outputs[0])) {
340         if (!av_audio_fifo_size(s->fifo[0]))
341             ff_inlink_request_frame(ctx->inputs[0]);
342         if (!av_audio_fifo_size(s->fifo[1]))
343             ff_inlink_request_frame(ctx->inputs[1]);
344     }
345     return 0;
346 }
347 
scquery_formats(AVFilterContext * ctx)348 static int scquery_formats(AVFilterContext *ctx)
349 {
350     AVFilterFormats *formats;
351     AVFilterChannelLayouts *layouts = NULL;
352     static const enum AVSampleFormat sample_fmts[] = {
353         AV_SAMPLE_FMT_DBL,
354         AV_SAMPLE_FMT_NONE
355     };
356     int ret, i;
357 
358     if (!ctx->inputs[0]->incfg.channel_layouts ||
359         !ctx->inputs[0]->incfg.channel_layouts->nb_channel_layouts) {
360         av_log(ctx, AV_LOG_WARNING,
361                "No channel layout for input 1\n");
362             return AVERROR(EAGAIN);
363     }
364 
365     if ((ret = ff_add_channel_layout(&layouts, ctx->inputs[0]->incfg.channel_layouts->channel_layouts[0])) < 0 ||
366         (ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->incfg.channel_layouts)) < 0)
367         return ret;
368 
369     for (i = 0; i < 2; i++) {
370         layouts = ff_all_channel_counts();
371         if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->outcfg.channel_layouts)) < 0)
372             return ret;
373     }
374 
375     formats = ff_make_format_list(sample_fmts);
376     if ((ret = ff_set_common_formats(ctx, formats)) < 0)
377         return ret;
378 
379     formats = ff_all_samplerates();
380     return ff_set_common_samplerates(ctx, formats);
381 }
382 
scconfig_output(AVFilterLink * outlink)383 static int scconfig_output(AVFilterLink *outlink)
384 {
385     AVFilterContext *ctx = outlink->src;
386     AudioGateContext *s = ctx->priv;
387 
388     if (ctx->inputs[0]->sample_rate != ctx->inputs[1]->sample_rate) {
389         av_log(ctx, AV_LOG_ERROR,
390                "Inputs must have the same sample rate "
391                "%d for in0 vs %d for in1\n",
392                ctx->inputs[0]->sample_rate, ctx->inputs[1]->sample_rate);
393         return AVERROR(EINVAL);
394     }
395 
396     outlink->sample_rate = ctx->inputs[0]->sample_rate;
397     outlink->time_base   = ctx->inputs[0]->time_base;
398     outlink->channel_layout = ctx->inputs[0]->channel_layout;
399     outlink->channels = ctx->inputs[0]->channels;
400 
401     s->fifo[0] = av_audio_fifo_alloc(ctx->inputs[0]->format, ctx->inputs[0]->channels, 1024);
402     s->fifo[1] = av_audio_fifo_alloc(ctx->inputs[1]->format, ctx->inputs[1]->channels, 1024);
403     if (!s->fifo[0] || !s->fifo[1])
404         return AVERROR(ENOMEM);
405 
406 
407     agate_config_input(ctx->inputs[0]);
408 
409     return 0;
410 }
411 
uninit(AVFilterContext * ctx)412 static av_cold void uninit(AVFilterContext *ctx)
413 {
414     AudioGateContext *s = ctx->priv;
415 
416     av_audio_fifo_free(s->fifo[0]);
417     av_audio_fifo_free(s->fifo[1]);
418 }
419 
420 static const AVFilterPad sidechaingate_inputs[] = {
421     {
422         .name           = "main",
423         .type           = AVMEDIA_TYPE_AUDIO,
424     },{
425         .name           = "sidechain",
426         .type           = AVMEDIA_TYPE_AUDIO,
427     },
428     { NULL }
429 };
430 
431 static const AVFilterPad sidechaingate_outputs[] = {
432     {
433         .name          = "default",
434         .type          = AVMEDIA_TYPE_AUDIO,
435         .config_props  = scconfig_output,
436     },
437     { NULL }
438 };
439 
440 AVFilter ff_af_sidechaingate = {
441     .name           = "sidechaingate",
442     .description    = NULL_IF_CONFIG_SMALL("Audio sidechain gate."),
443     .priv_size      = sizeof(AudioGateContext),
444     .priv_class     = &sidechaingate_class,
445     .query_formats  = scquery_formats,
446     .activate       = activate,
447     .uninit         = uninit,
448     .inputs         = sidechaingate_inputs,
449     .outputs        = sidechaingate_outputs,
450     .process_command = ff_filter_process_command,
451     .flags          = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
452 };
453 #endif  /* CONFIG_SIDECHAINGATE_FILTER */
454