2 * Copyright (C) 2001-2010 Krzysztof Foltman, Markus Schmidt, Thor Harald Johansen and others
3 * Copyright (c) 2015 Paul B Mahol
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Audio (Sidechain) Compressor filter
27 #include "libavutil/audio_fifo.h"
28 #include "libavutil/avassert.h"
29 #include "libavutil/channel_layout.h"
30 #include "libavutil/common.h"
31 #include "libavutil/opt.h"
40 typedef struct SidechainCompressContext {
45 double attack, attack_coeff;
46 double release, release_coeff;
56 double lin_knee_start;
57 double adj_knee_start;
58 double compressed_knee_stop;
64 } SidechainCompressContext;
66 #define OFFSET(x) offsetof(SidechainCompressContext, x)
67 #define A AV_OPT_FLAG_AUDIO_PARAM
68 #define F AV_OPT_FLAG_FILTERING_PARAM
70 static const AVOption options[] = {
71 { "level_in", "set input gain", OFFSET(level_in), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A|F },
72 { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_DOUBLE, {.dbl=0.125}, 0.000976563, 1, A|F },
73 { "ratio", "set ratio", OFFSET(ratio), AV_OPT_TYPE_DOUBLE, {.dbl=2}, 1, 20, A|F },
74 { "attack", "set attack", OFFSET(attack), AV_OPT_TYPE_DOUBLE, {.dbl=20}, 0.01, 2000, A|F },
75 { "release", "set release", OFFSET(release), AV_OPT_TYPE_DOUBLE, {.dbl=250}, 0.01, 9000, A|F },
76 { "makeup", "set make up gain", OFFSET(makeup), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 1, 64, A|F },
77 { "knee", "set knee", OFFSET(knee), AV_OPT_TYPE_DOUBLE, {.dbl=2.82843}, 1, 8, A|F },
78 { "link", "set link type", OFFSET(link), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, A|F, "link" },
79 { "average", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A|F, "link" },
80 { "maximum", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A|F, "link" },
81 { "detection", "set detection", OFFSET(detection), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, A|F, "detection" },
82 { "peak", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A|F, "detection" },
83 { "rms", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A|F, "detection" },
84 { "level_sc", "set sidechain gain", OFFSET(level_sc), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A|F },
85 { "mix", "set mix", OFFSET(mix), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0, 1, A|F },
89 #define sidechaincompress_options options
90 AVFILTER_DEFINE_CLASS(sidechaincompress);
92 // A fake infinity value (because real infinity may break some hosts)
93 #define FAKE_INFINITY (65536.0 * 65536.0)
95 // Check for infinity (with appropriate-ish tolerance)
96 #define IS_FAKE_INFINITY(value) (fabs(value-FAKE_INFINITY) < 1.0)
98 static double output_gain(double lin_slope, double ratio, double thres,
99 double knee, double knee_start, double knee_stop,
100 double compressed_knee_stop, int detection)
102 double slope = log(lin_slope);
109 if (IS_FAKE_INFINITY(ratio)) {
113 gain = (slope - thres) / ratio + thres;
117 if (knee > 1.0 && slope < knee_stop)
118 gain = hermite_interpolation(slope, knee_start, knee_stop,
119 knee_start, compressed_knee_stop,
122 return exp(gain - slope);
125 static int compressor_config_output(AVFilterLink *outlink)
127 AVFilterContext *ctx = outlink->src;
128 SidechainCompressContext *s = ctx->priv;
130 s->thres = log(s->threshold);
131 s->lin_knee_start = s->threshold / sqrt(s->knee);
132 s->adj_knee_start = s->lin_knee_start * s->lin_knee_start;
133 s->knee_start = log(s->lin_knee_start);
134 s->knee_stop = log(s->threshold * sqrt(s->knee));
135 s->compressed_knee_stop = (s->knee_stop - s->thres) / s->ratio + s->thres;
137 s->attack_coeff = FFMIN(1., 1. / (s->attack * outlink->sample_rate / 4000.));
138 s->release_coeff = FFMIN(1., 1. / (s->release * outlink->sample_rate / 4000.));
143 static void compressor(SidechainCompressContext *s,
144 const double *src, double *dst, const double *scsrc, int nb_samples,
145 double level_in, double level_sc,
146 AVFilterLink *inlink, AVFilterLink *sclink)
148 const double makeup = s->makeup;
149 const double mix = s->mix;
152 for (i = 0; i < nb_samples; i++) {
153 double abs_sample, gain = 1.0;
155 abs_sample = fabs(scsrc[0] * level_sc);
158 for (c = 1; c < sclink->channels; c++)
159 abs_sample = FFMAX(fabs(scsrc[c] * level_sc), abs_sample);
161 for (c = 1; c < sclink->channels; c++)
162 abs_sample += fabs(scsrc[c] * level_sc);
164 abs_sample /= sclink->channels;
168 abs_sample *= abs_sample;
170 s->lin_slope += (abs_sample - s->lin_slope) * (abs_sample > s->lin_slope ? s->attack_coeff : s->release_coeff);
172 if (s->lin_slope > 0.0 && s->lin_slope > (s->detection ? s->adj_knee_start : s->lin_knee_start))
173 gain = output_gain(s->lin_slope, s->ratio, s->thres, s->knee,
174 s->knee_start, s->knee_stop,
175 s->compressed_knee_stop, s->detection);
177 for (c = 0; c < inlink->channels; c++)
178 dst[c] = src[c] * level_in * (gain * makeup * mix + (1. - mix));
180 src += inlink->channels;
181 dst += inlink->channels;
182 scsrc += sclink->channels;
186 #if CONFIG_SIDECHAINCOMPRESS_FILTER
187 static int activate(AVFilterContext *ctx)
189 SidechainCompressContext *s = ctx->priv;
190 AVFrame *out = NULL, *in[2] = { NULL };
191 int ret, i, nb_samples;
194 FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx);
195 if ((ret = ff_inlink_consume_frame(ctx->inputs[0], &in[0])) > 0) {
196 av_audio_fifo_write(s->fifo[0], (void **)in[0]->extended_data,
198 av_frame_free(&in[0]);
202 if ((ret = ff_inlink_consume_frame(ctx->inputs[1], &in[1])) > 0) {
203 av_audio_fifo_write(s->fifo[1], (void **)in[1]->extended_data,
205 av_frame_free(&in[1]);
210 nb_samples = FFMIN(av_audio_fifo_size(s->fifo[0]), av_audio_fifo_size(s->fifo[1]));
212 out = ff_get_audio_buffer(ctx->outputs[0], nb_samples);
214 return AVERROR(ENOMEM);
215 for (i = 0; i < 2; i++) {
216 in[i] = ff_get_audio_buffer(ctx->inputs[i], nb_samples);
218 av_frame_free(&in[0]);
219 av_frame_free(&in[1]);
221 return AVERROR(ENOMEM);
223 av_audio_fifo_read(s->fifo[i], (void **)in[i]->data, nb_samples);
226 dst = (double *)out->data[0];
228 s->pts += nb_samples;
230 compressor(s, (double *)in[0]->data[0], dst,
231 (double *)in[1]->data[0], nb_samples,
232 s->level_in, s->level_sc,
233 ctx->inputs[0], ctx->inputs[1]);
235 av_frame_free(&in[0]);
236 av_frame_free(&in[1]);
238 ret = ff_filter_frame(ctx->outputs[0], out);
242 FF_FILTER_FORWARD_STATUS(ctx->inputs[0], ctx->outputs[0]);
243 FF_FILTER_FORWARD_STATUS(ctx->inputs[1], ctx->outputs[0]);
244 if (ff_outlink_frame_wanted(ctx->outputs[0])) {
245 if (!av_audio_fifo_size(s->fifo[0]))
246 ff_inlink_request_frame(ctx->inputs[0]);
247 if (!av_audio_fifo_size(s->fifo[1]))
248 ff_inlink_request_frame(ctx->inputs[1]);
253 static int query_formats(AVFilterContext *ctx)
255 AVFilterFormats *formats;
256 AVFilterChannelLayouts *layouts = NULL;
257 static const enum AVSampleFormat sample_fmts[] = {
263 if (!ctx->inputs[0]->in_channel_layouts ||
264 !ctx->inputs[0]->in_channel_layouts->nb_channel_layouts) {
265 av_log(ctx, AV_LOG_WARNING,
266 "No channel layout for input 1\n");
267 return AVERROR(EAGAIN);
270 if ((ret = ff_add_channel_layout(&layouts, ctx->inputs[0]->in_channel_layouts->channel_layouts[0])) < 0 ||
271 (ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
274 for (i = 0; i < 2; i++) {
275 layouts = ff_all_channel_counts();
276 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0)
280 formats = ff_make_format_list(sample_fmts);
281 if ((ret = ff_set_common_formats(ctx, formats)) < 0)
284 formats = ff_all_samplerates();
285 return ff_set_common_samplerates(ctx, formats);
288 static int config_output(AVFilterLink *outlink)
290 AVFilterContext *ctx = outlink->src;
291 SidechainCompressContext *s = ctx->priv;
293 if (ctx->inputs[0]->sample_rate != ctx->inputs[1]->sample_rate) {
294 av_log(ctx, AV_LOG_ERROR,
295 "Inputs must have the same sample rate "
296 "%d for in0 vs %d for in1\n",
297 ctx->inputs[0]->sample_rate, ctx->inputs[1]->sample_rate);
298 return AVERROR(EINVAL);
301 outlink->sample_rate = ctx->inputs[0]->sample_rate;
302 outlink->time_base = ctx->inputs[0]->time_base;
303 outlink->channel_layout = ctx->inputs[0]->channel_layout;
304 outlink->channels = ctx->inputs[0]->channels;
306 s->fifo[0] = av_audio_fifo_alloc(ctx->inputs[0]->format, ctx->inputs[0]->channels, 1024);
307 s->fifo[1] = av_audio_fifo_alloc(ctx->inputs[1]->format, ctx->inputs[1]->channels, 1024);
308 if (!s->fifo[0] || !s->fifo[1])
309 return AVERROR(ENOMEM);
311 compressor_config_output(outlink);
316 static av_cold void uninit(AVFilterContext *ctx)
318 SidechainCompressContext *s = ctx->priv;
320 av_audio_fifo_free(s->fifo[0]);
321 av_audio_fifo_free(s->fifo[1]);
324 static const AVFilterPad sidechaincompress_inputs[] = {
327 .type = AVMEDIA_TYPE_AUDIO,
330 .type = AVMEDIA_TYPE_AUDIO,
335 static const AVFilterPad sidechaincompress_outputs[] = {
338 .type = AVMEDIA_TYPE_AUDIO,
339 .config_props = config_output,
344 AVFilter ff_af_sidechaincompress = {
345 .name = "sidechaincompress",
346 .description = NULL_IF_CONFIG_SMALL("Sidechain compressor."),
347 .priv_size = sizeof(SidechainCompressContext),
348 .priv_class = &sidechaincompress_class,
349 .query_formats = query_formats,
350 .activate = activate,
352 .inputs = sidechaincompress_inputs,
353 .outputs = sidechaincompress_outputs,
355 #endif /* CONFIG_SIDECHAINCOMPRESS_FILTER */
357 #if CONFIG_ACOMPRESSOR_FILTER
358 static int acompressor_filter_frame(AVFilterLink *inlink, AVFrame *in)
360 const double *src = (const double *)in->data[0];
361 AVFilterContext *ctx = inlink->dst;
362 SidechainCompressContext *s = ctx->priv;
363 AVFilterLink *outlink = ctx->outputs[0];
367 if (av_frame_is_writable(in)) {
370 out = ff_get_audio_buffer(outlink, in->nb_samples);
373 return AVERROR(ENOMEM);
375 av_frame_copy_props(out, in);
377 dst = (double *)out->data[0];
379 compressor(s, src, dst, src, in->nb_samples,
380 s->level_in, s->level_in,
385 return ff_filter_frame(outlink, out);
388 static int acompressor_query_formats(AVFilterContext *ctx)
390 AVFilterFormats *formats;
391 AVFilterChannelLayouts *layouts;
392 static const enum AVSampleFormat sample_fmts[] = {
398 layouts = ff_all_channel_counts();
400 return AVERROR(ENOMEM);
401 ret = ff_set_common_channel_layouts(ctx, layouts);
405 formats = ff_make_format_list(sample_fmts);
407 return AVERROR(ENOMEM);
408 ret = ff_set_common_formats(ctx, formats);
412 formats = ff_all_samplerates();
414 return AVERROR(ENOMEM);
415 return ff_set_common_samplerates(ctx, formats);
418 #define acompressor_options options
419 AVFILTER_DEFINE_CLASS(acompressor);
421 static const AVFilterPad acompressor_inputs[] = {
424 .type = AVMEDIA_TYPE_AUDIO,
425 .filter_frame = acompressor_filter_frame,
430 static const AVFilterPad acompressor_outputs[] = {
433 .type = AVMEDIA_TYPE_AUDIO,
434 .config_props = compressor_config_output,
439 AVFilter ff_af_acompressor = {
440 .name = "acompressor",
441 .description = NULL_IF_CONFIG_SMALL("Audio compressor."),
442 .priv_size = sizeof(SidechainCompressContext),
443 .priv_class = &acompressor_class,
444 .query_formats = acompressor_query_formats,
445 .inputs = acompressor_inputs,
446 .outputs = acompressor_outputs,
448 #endif /* CONFIG_ACOMPRESSOR_FILTER */