2 * Copyright (C) 2001-2010 Krzysztof Foltman, Markus Schmidt, Thor Harald Johansen, Damien Zammit
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * Audio (Sidechain) Gate filter
26 #include "libavutil/audio_fifo.h"
27 #include "libavutil/avassert.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/opt.h"
35 typedef struct AudioGateContext {
62 #define OFFSET(x) offsetof(AudioGateContext, x)
63 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
65 static const AVOption options[] = {
66 { "level_in", "set input level", OFFSET(level_in), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A },
67 { "range", "set max gain reduction", OFFSET(range), AV_OPT_TYPE_DOUBLE, {.dbl=0.06125}, 0, 1, A },
68 { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_DOUBLE, {.dbl=0.125}, 0, 1, A },
69 { "ratio", "set ratio", OFFSET(ratio), AV_OPT_TYPE_DOUBLE, {.dbl=2}, 1, 9000, A },
70 { "attack", "set attack", OFFSET(attack), AV_OPT_TYPE_DOUBLE, {.dbl=20}, 0.01, 9000, A },
71 { "release", "set release", OFFSET(release), AV_OPT_TYPE_DOUBLE, {.dbl=250}, 0.01, 9000, A },
72 { "makeup", "set makeup gain", OFFSET(makeup), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 1, 64, A },
73 { "knee", "set knee", OFFSET(knee), AV_OPT_TYPE_DOUBLE, {.dbl=2.828427125}, 1, 8, A },
74 { "detection", "set detection", OFFSET(detection), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, A, "detection" },
75 { "peak", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A, "detection" },
76 { "rms", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A, "detection" },
77 { "link", "set link", OFFSET(link), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, A, "link" },
78 { "average", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A, "link" },
79 { "maximum", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A, "link" },
80 { "level_sc", "set sidechain gain", OFFSET(level_sc), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A },
84 static int agate_config_input(AVFilterLink *inlink)
86 AVFilterContext *ctx = inlink->dst;
87 AudioGateContext *s = ctx->priv;
88 double lin_threshold = s->threshold;
89 double lin_knee_sqrt = sqrt(s->knee);
90 double lin_knee_start;
93 lin_threshold *= lin_threshold;
95 s->attack_coeff = FFMIN(1., 1. / (s->attack * inlink->sample_rate / 4000.));
96 s->release_coeff = FFMIN(1., 1. / (s->release * inlink->sample_rate / 4000.));
97 s->lin_knee_stop = lin_threshold * lin_knee_sqrt;
98 lin_knee_start = lin_threshold / lin_knee_sqrt;
99 s->thres = log(lin_threshold);
100 s->knee_start = log(lin_knee_start);
101 s->knee_stop = log(s->lin_knee_stop);
106 // A fake infinity value (because real infinity may break some hosts)
107 #define FAKE_INFINITY (65536.0 * 65536.0)
109 // Check for infinity (with appropriate-ish tolerance)
110 #define IS_FAKE_INFINITY(value) (fabs(value-FAKE_INFINITY) < 1.0)
112 static double output_gain(double lin_slope, double ratio, double thres,
113 double knee, double knee_start, double knee_stop,
114 double lin_knee_stop, double range)
116 if (lin_slope < lin_knee_stop) {
117 double slope = log(lin_slope);
118 double tratio = ratio;
122 if (IS_FAKE_INFINITY(ratio))
124 gain = (slope - thres) * tratio + thres;
127 if (knee > 1. && slope > knee_start) {
128 gain = hermite_interpolation(slope, knee_start, knee_stop, ((knee_start - thres) * tratio + thres), knee_stop, delta, 1.);
130 return FFMAX(range, exp(gain - slope));
136 static void gate(AudioGateContext *s,
137 const double *src, double *dst, const double *scsrc,
138 int nb_samples, double level_in, double level_sc,
139 AVFilterLink *inlink, AVFilterLink *sclink)
141 const double makeup = s->makeup;
142 const double attack_coeff = s->attack_coeff;
143 const double release_coeff = s->release_coeff;
146 for (n = 0; n < nb_samples; n++, src += inlink->channels, dst += inlink->channels, scsrc += sclink->channels) {
147 double abs_sample = fabs(scsrc[0] * level_sc), gain = 1.0;
150 for (c = 1; c < sclink->channels; c++)
151 abs_sample = FFMAX(fabs(scsrc[c] * level_sc), abs_sample);
153 for (c = 1; c < sclink->channels; c++)
154 abs_sample += fabs(scsrc[c] * level_sc);
156 abs_sample /= sclink->channels;
160 abs_sample *= abs_sample;
162 s->lin_slope += (abs_sample - s->lin_slope) * (abs_sample > s->lin_slope ? attack_coeff : release_coeff);
163 if (s->lin_slope > 0.0)
164 gain = output_gain(s->lin_slope, s->ratio, s->thres,
165 s->knee, s->knee_start, s->knee_stop,
166 s->lin_knee_stop, s->range);
168 for (c = 0; c < inlink->channels; c++)
169 dst[c] = src[c] * level_in * gain * makeup;
173 #if CONFIG_AGATE_FILTER
175 #define agate_options options
176 AVFILTER_DEFINE_CLASS(agate);
178 static int query_formats(AVFilterContext *ctx)
180 AVFilterFormats *formats = NULL;
181 AVFilterChannelLayouts *layouts;
184 if ((ret = ff_add_format(&formats, AV_SAMPLE_FMT_DBL)) < 0)
186 ret = ff_set_common_formats(ctx, formats);
190 layouts = ff_all_channel_counts();
192 return AVERROR(ENOMEM);
193 ret = ff_set_common_channel_layouts(ctx, layouts);
197 formats = ff_all_samplerates();
199 return AVERROR(ENOMEM);
201 return ff_set_common_samplerates(ctx, formats);
204 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
206 const double *src = (const double *)in->data[0];
207 AVFilterContext *ctx = inlink->dst;
208 AVFilterLink *outlink = ctx->outputs[0];
209 AudioGateContext *s = ctx->priv;
213 if (av_frame_is_writable(in)) {
216 out = ff_get_audio_buffer(inlink, in->nb_samples);
219 return AVERROR(ENOMEM);
221 av_frame_copy_props(out, in);
223 dst = (double *)out->data[0];
225 gate(s, src, dst, src, in->nb_samples,
226 s->level_in, s->level_in, inlink, inlink);
230 return ff_filter_frame(outlink, out);
233 static const AVFilterPad inputs[] = {
236 .type = AVMEDIA_TYPE_AUDIO,
237 .filter_frame = filter_frame,
238 .config_props = agate_config_input,
243 static const AVFilterPad outputs[] = {
246 .type = AVMEDIA_TYPE_AUDIO,
251 AVFilter ff_af_agate = {
253 .description = NULL_IF_CONFIG_SMALL("Audio gate."),
254 .query_formats = query_formats,
255 .priv_size = sizeof(AudioGateContext),
256 .priv_class = &agate_class,
261 #endif /* CONFIG_AGATE_FILTER */
263 #if CONFIG_SIDECHAINGATE_FILTER
265 #define sidechaingate_options options
266 AVFILTER_DEFINE_CLASS(sidechaingate);
268 static int scfilter_frame(AVFilterLink *link, AVFrame *frame)
270 AVFilterContext *ctx = link->dst;
271 AudioGateContext *s = ctx->priv;
272 AVFilterLink *outlink = ctx->outputs[0];
273 AVFrame *out, *in[2] = { NULL };
278 for (i = 0; i < 2; i++)
279 if (link == ctx->inputs[i])
282 av_audio_fifo_write(s->fifo[i], (void **)frame->extended_data,
284 av_frame_free(&frame);
286 nb_samples = FFMIN(av_audio_fifo_size(s->fifo[0]), av_audio_fifo_size(s->fifo[1]));
290 out = ff_get_audio_buffer(outlink, nb_samples);
292 return AVERROR(ENOMEM);
293 for (i = 0; i < 2; i++) {
294 in[i] = ff_get_audio_buffer(ctx->inputs[i], nb_samples);
296 av_frame_free(&in[0]);
297 av_frame_free(&in[1]);
299 return AVERROR(ENOMEM);
301 av_audio_fifo_read(s->fifo[i], (void **)in[i]->data, nb_samples);
304 dst = (double *)out->data[0];
306 s->pts += nb_samples;
308 gate(s, (double *)in[0]->data[0], dst,
309 (double *)in[1]->data[0], nb_samples,
310 s->level_in, s->level_sc,
311 ctx->inputs[0], ctx->inputs[1]);
313 av_frame_free(&in[0]);
314 av_frame_free(&in[1]);
316 return ff_filter_frame(outlink, out);
319 static int screquest_frame(AVFilterLink *outlink)
321 AVFilterContext *ctx = outlink->src;
322 AudioGateContext *s = ctx->priv;
325 /* get a frame on each input */
326 for (i = 0; i < 2; i++) {
327 AVFilterLink *inlink = ctx->inputs[i];
328 if (!av_audio_fifo_size(s->fifo[i]))
329 return ff_request_frame(inlink);
335 static int scquery_formats(AVFilterContext *ctx)
337 AVFilterFormats *formats;
338 AVFilterChannelLayouts *layouts = NULL;
339 static const enum AVSampleFormat sample_fmts[] = {
345 if (!ctx->inputs[0]->in_channel_layouts ||
346 !ctx->inputs[0]->in_channel_layouts->nb_channel_layouts) {
347 av_log(ctx, AV_LOG_WARNING,
348 "No channel layout for input 1\n");
349 return AVERROR(EAGAIN);
352 if ((ret = ff_add_channel_layout(&layouts, ctx->inputs[0]->in_channel_layouts->channel_layouts[0])) < 0 ||
353 (ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
356 for (i = 0; i < 2; i++) {
357 layouts = ff_all_channel_counts();
358 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0)
362 formats = ff_make_format_list(sample_fmts);
363 if ((ret = ff_set_common_formats(ctx, formats)) < 0)
366 formats = ff_all_samplerates();
367 return ff_set_common_samplerates(ctx, formats);
370 static int scconfig_output(AVFilterLink *outlink)
372 AVFilterContext *ctx = outlink->src;
373 AudioGateContext *s = ctx->priv;
375 if (ctx->inputs[0]->sample_rate != ctx->inputs[1]->sample_rate) {
376 av_log(ctx, AV_LOG_ERROR,
377 "Inputs must have the same sample rate "
378 "%d for in0 vs %d for in1\n",
379 ctx->inputs[0]->sample_rate, ctx->inputs[1]->sample_rate);
380 return AVERROR(EINVAL);
383 outlink->sample_rate = ctx->inputs[0]->sample_rate;
384 outlink->time_base = ctx->inputs[0]->time_base;
385 outlink->channel_layout = ctx->inputs[0]->channel_layout;
386 outlink->channels = ctx->inputs[0]->channels;
388 s->fifo[0] = av_audio_fifo_alloc(ctx->inputs[0]->format, ctx->inputs[0]->channels, 1024);
389 s->fifo[1] = av_audio_fifo_alloc(ctx->inputs[1]->format, ctx->inputs[1]->channels, 1024);
390 if (!s->fifo[0] || !s->fifo[1])
391 return AVERROR(ENOMEM);
394 agate_config_input(ctx->inputs[0]);
399 static av_cold void uninit(AVFilterContext *ctx)
401 AudioGateContext *s = ctx->priv;
403 av_audio_fifo_free(s->fifo[0]);
404 av_audio_fifo_free(s->fifo[1]);
407 static const AVFilterPad sidechaingate_inputs[] = {
410 .type = AVMEDIA_TYPE_AUDIO,
411 .filter_frame = scfilter_frame,
414 .type = AVMEDIA_TYPE_AUDIO,
415 .filter_frame = scfilter_frame,
420 static const AVFilterPad sidechaingate_outputs[] = {
423 .type = AVMEDIA_TYPE_AUDIO,
424 .config_props = scconfig_output,
425 .request_frame = screquest_frame,
430 AVFilter ff_af_sidechaingate = {
431 .name = "sidechaingate",
432 .description = NULL_IF_CONFIG_SMALL("Audio sidechain gate."),
433 .priv_size = sizeof(AudioGateContext),
434 .priv_class = &sidechaingate_class,
435 .query_formats = scquery_formats,
437 .inputs = sidechaingate_inputs,
438 .outputs = sidechaingate_outputs,
440 #endif /* CONFIG_SIDECHAINGATE_FILTER */