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 {
63 #define OFFSET(x) offsetof(AudioGateContext, x)
64 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
66 static const AVOption options[] = {
67 { "level_in", "set input level", OFFSET(level_in), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A },
68 { "range", "set max gain reduction", OFFSET(range), AV_OPT_TYPE_DOUBLE, {.dbl=0.06125}, 0, 1, A },
69 { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_DOUBLE, {.dbl=0.125}, 0, 1, A },
70 { "ratio", "set ratio", OFFSET(ratio), AV_OPT_TYPE_DOUBLE, {.dbl=2}, 1, 9000, A },
71 { "attack", "set attack", OFFSET(attack), AV_OPT_TYPE_DOUBLE, {.dbl=20}, 0.01, 9000, A },
72 { "release", "set release", OFFSET(release), AV_OPT_TYPE_DOUBLE, {.dbl=250}, 0.01, 9000, A },
73 { "makeup", "set makeup gain", OFFSET(makeup), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 1, 64, A },
74 { "knee", "set knee", OFFSET(knee), AV_OPT_TYPE_DOUBLE, {.dbl=2.828427125}, 1, 8, A },
75 { "detection", "set detection", OFFSET(detection), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, A, "detection" },
76 { "peak", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A, "detection" },
77 { "rms", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A, "detection" },
78 { "link", "set link", OFFSET(link), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, A, "link" },
79 { "average", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A, "link" },
80 { "maximum", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A, "link" },
81 { "level_sc", "set sidechain gain", OFFSET(level_sc), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A },
85 static int agate_config_input(AVFilterLink *inlink)
87 AVFilterContext *ctx = inlink->dst;
88 AudioGateContext *s = ctx->priv;
89 double lin_threshold = s->threshold;
90 double lin_knee_sqrt = sqrt(s->knee);
91 double lin_knee_start;
94 lin_threshold *= lin_threshold;
96 s->attack_coeff = FFMIN(1., 1. / (s->attack * inlink->sample_rate / 4000.));
97 s->release_coeff = FFMIN(1., 1. / (s->release * inlink->sample_rate / 4000.));
98 s->lin_knee_stop = lin_threshold * lin_knee_sqrt;
99 lin_knee_start = lin_threshold / lin_knee_sqrt;
100 s->thres = log(lin_threshold);
101 s->knee_start = log(lin_knee_start);
102 s->knee_stop = log(s->lin_knee_stop);
107 // A fake infinity value (because real infinity may break some hosts)
108 #define FAKE_INFINITY (65536.0 * 65536.0)
110 // Check for infinity (with appropriate-ish tolerance)
111 #define IS_FAKE_INFINITY(value) (fabs(value-FAKE_INFINITY) < 1.0)
113 static double output_gain(double lin_slope, double ratio, double thres,
114 double knee, double knee_start, double knee_stop,
115 double lin_knee_stop, double range)
117 if (lin_slope < lin_knee_stop) {
118 double slope = log(lin_slope);
119 double tratio = ratio;
123 if (IS_FAKE_INFINITY(ratio))
125 gain = (slope - thres) * tratio + thres;
128 if (knee > 1. && slope > knee_start) {
129 gain = hermite_interpolation(slope, knee_start, knee_stop, ((knee_start - thres) * tratio + thres), knee_stop, delta, 1.);
131 return FFMAX(range, exp(gain - slope));
137 static void gate(AudioGateContext *s,
138 const double *src, double *dst, const double *scsrc,
139 int nb_samples, double level_in, double level_sc,
140 AVFilterLink *inlink, AVFilterLink *sclink)
142 const double makeup = s->makeup;
143 const double attack_coeff = s->attack_coeff;
144 const double release_coeff = s->release_coeff;
147 for (n = 0; n < nb_samples; n++, src += inlink->channels, dst += inlink->channels, scsrc += sclink->channels) {
148 double abs_sample = fabs(scsrc[0] * level_sc), gain = 1.0;
151 for (c = 1; c < sclink->channels; c++)
152 abs_sample = FFMAX(fabs(scsrc[c] * level_sc), abs_sample);
154 for (c = 1; c < sclink->channels; c++)
155 abs_sample += fabs(scsrc[c] * level_sc);
157 abs_sample /= sclink->channels;
161 abs_sample *= abs_sample;
163 s->lin_slope += (abs_sample - s->lin_slope) * (abs_sample > s->lin_slope ? attack_coeff : release_coeff);
164 if (s->lin_slope > 0.0)
165 gain = output_gain(s->lin_slope, s->ratio, s->thres,
166 s->knee, s->knee_start, s->knee_stop,
167 s->lin_knee_stop, s->range);
169 for (c = 0; c < inlink->channels; c++)
170 dst[c] = src[c] * level_in * gain * makeup;
174 #if CONFIG_AGATE_FILTER
176 #define agate_options options
177 AVFILTER_DEFINE_CLASS(agate);
179 static int query_formats(AVFilterContext *ctx)
181 AVFilterFormats *formats = NULL;
182 AVFilterChannelLayouts *layouts;
185 if ((ret = ff_add_format(&formats, AV_SAMPLE_FMT_DBL)) < 0)
187 ret = ff_set_common_formats(ctx, formats);
191 layouts = ff_all_channel_counts();
193 return AVERROR(ENOMEM);
194 ret = ff_set_common_channel_layouts(ctx, layouts);
198 formats = ff_all_samplerates();
200 return AVERROR(ENOMEM);
202 return ff_set_common_samplerates(ctx, formats);
205 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
207 const double *src = (const double *)in->data[0];
208 AVFilterContext *ctx = inlink->dst;
209 AVFilterLink *outlink = ctx->outputs[0];
210 AudioGateContext *s = ctx->priv;
214 if (av_frame_is_writable(in)) {
217 out = ff_get_audio_buffer(outlink, in->nb_samples);
220 return AVERROR(ENOMEM);
222 av_frame_copy_props(out, in);
224 dst = (double *)out->data[0];
226 gate(s, src, dst, src, in->nb_samples,
227 s->level_in, s->level_in, inlink, inlink);
231 return ff_filter_frame(outlink, out);
234 static const AVFilterPad inputs[] = {
237 .type = AVMEDIA_TYPE_AUDIO,
238 .filter_frame = filter_frame,
239 .config_props = agate_config_input,
244 static const AVFilterPad outputs[] = {
247 .type = AVMEDIA_TYPE_AUDIO,
252 AVFilter ff_af_agate = {
254 .description = NULL_IF_CONFIG_SMALL("Audio gate."),
255 .query_formats = query_formats,
256 .priv_size = sizeof(AudioGateContext),
257 .priv_class = &agate_class,
262 #endif /* CONFIG_AGATE_FILTER */
264 #if CONFIG_SIDECHAINGATE_FILTER
266 #define sidechaingate_options options
267 AVFILTER_DEFINE_CLASS(sidechaingate);
269 static int activate(AVFilterContext *ctx)
271 AudioGateContext *s = ctx->priv;
272 AVFrame *out = NULL, *in[2] = { NULL };
273 int ret, i, nb_samples;
276 FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx);
277 if ((ret = ff_inlink_consume_frame(ctx->inputs[0], &in[0])) > 0) {
278 av_audio_fifo_write(s->fifo[0], (void **)in[0]->extended_data,
280 av_frame_free(&in[0]);
284 if ((ret = ff_inlink_consume_frame(ctx->inputs[1], &in[1])) > 0) {
285 av_audio_fifo_write(s->fifo[1], (void **)in[1]->extended_data,
287 av_frame_free(&in[1]);
292 nb_samples = FFMIN(av_audio_fifo_size(s->fifo[0]), av_audio_fifo_size(s->fifo[1]));
294 out = ff_get_audio_buffer(ctx->outputs[0], nb_samples);
296 return AVERROR(ENOMEM);
297 for (i = 0; i < 2; i++) {
298 in[i] = ff_get_audio_buffer(ctx->inputs[i], nb_samples);
300 av_frame_free(&in[0]);
301 av_frame_free(&in[1]);
303 return AVERROR(ENOMEM);
305 av_audio_fifo_read(s->fifo[i], (void **)in[i]->data, nb_samples);
308 dst = (double *)out->data[0];
310 s->pts += nb_samples;
312 gate(s, (double *)in[0]->data[0], dst,
313 (double *)in[1]->data[0], nb_samples,
314 s->level_in, s->level_sc,
315 ctx->inputs[0], ctx->inputs[1]);
317 av_frame_free(&in[0]);
318 av_frame_free(&in[1]);
320 ret = ff_filter_frame(ctx->outputs[0], out);
324 FF_FILTER_FORWARD_STATUS(ctx->inputs[0], ctx->outputs[0]);
325 FF_FILTER_FORWARD_STATUS(ctx->inputs[1], ctx->outputs[0]);
326 if (ff_outlink_frame_wanted(ctx->outputs[0])) {
327 if (!av_audio_fifo_size(s->fifo[0]))
328 ff_inlink_request_frame(ctx->inputs[0]);
329 if (!av_audio_fifo_size(s->fifo[1]))
330 ff_inlink_request_frame(ctx->inputs[1]);
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,
413 .type = AVMEDIA_TYPE_AUDIO,
418 static const AVFilterPad sidechaingate_outputs[] = {
421 .type = AVMEDIA_TYPE_AUDIO,
422 .config_props = scconfig_output,
427 AVFilter ff_af_sidechaingate = {
428 .name = "sidechaingate",
429 .description = NULL_IF_CONFIG_SMALL("Audio sidechain gate."),
430 .priv_size = sizeof(AudioGateContext),
431 .priv_class = &sidechaingate_class,
432 .query_formats = scquery_formats,
433 .activate = activate,
435 .inputs = sidechaingate_inputs,
436 .outputs = sidechaingate_outputs,
438 #endif /* CONFIG_SIDECHAINGATE_FILTER */