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"
36 typedef struct AudioGateContext {
55 double lin_knee_start;
65 #define OFFSET(x) offsetof(AudioGateContext, x)
66 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
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 },
90 static int agate_config_input(AVFilterLink *inlink)
92 AVFilterContext *ctx = inlink->dst;
93 AudioGateContext *s = ctx->priv;
94 double lin_threshold = s->threshold;
95 double lin_knee_sqrt = sqrt(s->knee);
98 lin_threshold *= lin_threshold;
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);
111 // A fake infinity value (because real infinity may break some hosts)
112 #define FAKE_INFINITY (65536.0 * 65536.0)
114 // Check for infinity (with appropriate-ish tolerance)
115 #define IS_FAKE_INFINITY(value) (fabs(value-FAKE_INFINITY) < 1.0)
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)
121 double slope = log(lin_slope);
122 double tratio = ratio;
126 if (IS_FAKE_INFINITY(ratio))
128 gain = (slope - thres) * tratio + thres;
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.);
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.);
138 return FFMAX(range, exp(gain - slope));
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)
146 const double makeup = s->makeup;
147 const double attack_coeff = s->attack_coeff;
148 const double release_coeff = s->release_coeff;
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;
156 for (c = 1; c < sclink->channels; c++)
157 abs_sample = FFMAX(fabs(scsrc[c] * level_sc), abs_sample);
159 for (c = 1; c < sclink->channels; c++)
160 abs_sample += fabs(scsrc[c] * level_sc);
162 abs_sample /= sclink->channels;
166 abs_sample *= abs_sample;
168 s->lin_slope += (abs_sample - s->lin_slope) * (abs_sample > s->lin_slope ? attack_coeff : release_coeff);
171 detected = s->lin_slope > s->lin_knee_start;
173 detected = s->lin_slope < s->lin_knee_stop;
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,
180 for (c = 0; c < inlink->channels; c++)
181 dst[c] = src[c] * level_in * gain * makeup;
185 #if CONFIG_AGATE_FILTER
187 #define agate_options options
188 AVFILTER_DEFINE_CLASS(agate);
190 static int query_formats(AVFilterContext *ctx)
192 AVFilterFormats *formats = NULL;
193 AVFilterChannelLayouts *layouts;
196 if ((ret = ff_add_format(&formats, AV_SAMPLE_FMT_DBL)) < 0)
198 ret = ff_set_common_formats(ctx, formats);
202 layouts = ff_all_channel_counts();
204 return AVERROR(ENOMEM);
205 ret = ff_set_common_channel_layouts(ctx, layouts);
209 formats = ff_all_samplerates();
211 return AVERROR(ENOMEM);
213 return ff_set_common_samplerates(ctx, formats);
216 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
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;
225 if (av_frame_is_writable(in)) {
228 out = ff_get_audio_buffer(outlink, in->nb_samples);
231 return AVERROR(ENOMEM);
233 av_frame_copy_props(out, in);
235 dst = (double *)out->data[0];
237 gate(s, src, dst, src, in->nb_samples,
238 s->level_in, s->level_in, inlink, inlink);
242 return ff_filter_frame(outlink, out);
245 static const AVFilterPad inputs[] = {
248 .type = AVMEDIA_TYPE_AUDIO,
249 .filter_frame = filter_frame,
250 .config_props = agate_config_input,
255 static const AVFilterPad outputs[] = {
258 .type = AVMEDIA_TYPE_AUDIO,
263 AVFilter ff_af_agate = {
265 .description = NULL_IF_CONFIG_SMALL("Audio gate."),
266 .query_formats = query_formats,
267 .priv_size = sizeof(AudioGateContext),
268 .priv_class = &agate_class,
273 #endif /* CONFIG_AGATE_FILTER */
275 #if CONFIG_SIDECHAINGATE_FILTER
277 #define sidechaingate_options options
278 AVFILTER_DEFINE_CLASS(sidechaingate);
280 static int activate(AVFilterContext *ctx)
282 AudioGateContext *s = ctx->priv;
283 AVFrame *out = NULL, *in[2] = { NULL };
284 int ret, i, nb_samples;
287 FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx);
288 if ((ret = ff_inlink_consume_frame(ctx->inputs[0], &in[0])) > 0) {
289 av_audio_fifo_write(s->fifo[0], (void **)in[0]->extended_data,
291 av_frame_free(&in[0]);
295 if ((ret = ff_inlink_consume_frame(ctx->inputs[1], &in[1])) > 0) {
296 av_audio_fifo_write(s->fifo[1], (void **)in[1]->extended_data,
298 av_frame_free(&in[1]);
303 nb_samples = FFMIN(av_audio_fifo_size(s->fifo[0]), av_audio_fifo_size(s->fifo[1]));
305 out = ff_get_audio_buffer(ctx->outputs[0], nb_samples);
307 return AVERROR(ENOMEM);
308 for (i = 0; i < 2; i++) {
309 in[i] = ff_get_audio_buffer(ctx->inputs[i], nb_samples);
311 av_frame_free(&in[0]);
312 av_frame_free(&in[1]);
314 return AVERROR(ENOMEM);
316 av_audio_fifo_read(s->fifo[i], (void **)in[i]->data, nb_samples);
319 dst = (double *)out->data[0];
321 s->pts += nb_samples;
323 gate(s, (double *)in[0]->data[0], dst,
324 (double *)in[1]->data[0], nb_samples,
325 s->level_in, s->level_sc,
326 ctx->inputs[0], ctx->inputs[1]);
328 av_frame_free(&in[0]);
329 av_frame_free(&in[1]);
331 ret = ff_filter_frame(ctx->outputs[0], out);
335 FF_FILTER_FORWARD_STATUS(ctx->inputs[0], ctx->outputs[0]);
336 FF_FILTER_FORWARD_STATUS(ctx->inputs[1], ctx->outputs[0]);
337 if (ff_outlink_frame_wanted(ctx->outputs[0])) {
338 if (!av_audio_fifo_size(s->fifo[0]))
339 ff_inlink_request_frame(ctx->inputs[0]);
340 if (!av_audio_fifo_size(s->fifo[1]))
341 ff_inlink_request_frame(ctx->inputs[1]);
346 static int scquery_formats(AVFilterContext *ctx)
348 AVFilterFormats *formats;
349 AVFilterChannelLayouts *layouts = NULL;
350 static const enum AVSampleFormat sample_fmts[] = {
356 if (!ctx->inputs[0]->in_channel_layouts ||
357 !ctx->inputs[0]->in_channel_layouts->nb_channel_layouts) {
358 av_log(ctx, AV_LOG_WARNING,
359 "No channel layout for input 1\n");
360 return AVERROR(EAGAIN);
363 if ((ret = ff_add_channel_layout(&layouts, ctx->inputs[0]->in_channel_layouts->channel_layouts[0])) < 0 ||
364 (ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
367 for (i = 0; i < 2; i++) {
368 layouts = ff_all_channel_counts();
369 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0)
373 formats = ff_make_format_list(sample_fmts);
374 if ((ret = ff_set_common_formats(ctx, formats)) < 0)
377 formats = ff_all_samplerates();
378 return ff_set_common_samplerates(ctx, formats);
381 static int scconfig_output(AVFilterLink *outlink)
383 AVFilterContext *ctx = outlink->src;
384 AudioGateContext *s = ctx->priv;
386 if (ctx->inputs[0]->sample_rate != ctx->inputs[1]->sample_rate) {
387 av_log(ctx, AV_LOG_ERROR,
388 "Inputs must have the same sample rate "
389 "%d for in0 vs %d for in1\n",
390 ctx->inputs[0]->sample_rate, ctx->inputs[1]->sample_rate);
391 return AVERROR(EINVAL);
394 outlink->sample_rate = ctx->inputs[0]->sample_rate;
395 outlink->time_base = ctx->inputs[0]->time_base;
396 outlink->channel_layout = ctx->inputs[0]->channel_layout;
397 outlink->channels = ctx->inputs[0]->channels;
399 s->fifo[0] = av_audio_fifo_alloc(ctx->inputs[0]->format, ctx->inputs[0]->channels, 1024);
400 s->fifo[1] = av_audio_fifo_alloc(ctx->inputs[1]->format, ctx->inputs[1]->channels, 1024);
401 if (!s->fifo[0] || !s->fifo[1])
402 return AVERROR(ENOMEM);
405 agate_config_input(ctx->inputs[0]);
410 static av_cold void uninit(AVFilterContext *ctx)
412 AudioGateContext *s = ctx->priv;
414 av_audio_fifo_free(s->fifo[0]);
415 av_audio_fifo_free(s->fifo[1]);
418 static const AVFilterPad sidechaingate_inputs[] = {
421 .type = AVMEDIA_TYPE_AUDIO,
424 .type = AVMEDIA_TYPE_AUDIO,
429 static const AVFilterPad sidechaingate_outputs[] = {
432 .type = AVMEDIA_TYPE_AUDIO,
433 .config_props = scconfig_output,
438 AVFilter ff_af_sidechaingate = {
439 .name = "sidechaingate",
440 .description = NULL_IF_CONFIG_SMALL("Audio sidechain gate."),
441 .priv_size = sizeof(AudioGateContext),
442 .priv_class = &sidechaingate_class,
443 .query_formats = scquery_formats,
444 .activate = activate,
446 .inputs = sidechaingate_inputs,
447 .outputs = sidechaingate_outputs,
449 #endif /* CONFIG_SIDECHAINGATE_FILTER */