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;
58 double adj_knee_start;
60 double compressed_knee_start;
61 double compressed_knee_stop;
68 } SidechainCompressContext;
70 #define OFFSET(x) offsetof(SidechainCompressContext, x)
71 #define A AV_OPT_FLAG_AUDIO_PARAM
72 #define F AV_OPT_FLAG_FILTERING_PARAM
73 #define R AV_OPT_FLAG_RUNTIME_PARAM
75 static const AVOption options[] = {
76 { "level_in", "set input gain", OFFSET(level_in), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A|F|R },
77 { "mode", "set mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, A|F|R, "mode" },
78 { "downward",0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A|F|R, "mode" },
79 { "upward", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A|F|R, "mode" },
80 { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_DOUBLE, {.dbl=0.125}, 0.000976563, 1, A|F|R },
81 { "ratio", "set ratio", OFFSET(ratio), AV_OPT_TYPE_DOUBLE, {.dbl=2}, 1, 20, A|F|R },
82 { "attack", "set attack", OFFSET(attack), AV_OPT_TYPE_DOUBLE, {.dbl=20}, 0.01, 2000, A|F|R },
83 { "release", "set release", OFFSET(release), AV_OPT_TYPE_DOUBLE, {.dbl=250}, 0.01, 9000, A|F|R },
84 { "makeup", "set make up gain", OFFSET(makeup), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 1, 64, A|F|R },
85 { "knee", "set knee", OFFSET(knee), AV_OPT_TYPE_DOUBLE, {.dbl=2.82843}, 1, 8, A|F|R },
86 { "link", "set link type", OFFSET(link), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, A|F|R, "link" },
87 { "average", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A|F|R, "link" },
88 { "maximum", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A|F|R, "link" },
89 { "detection", "set detection", OFFSET(detection), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, A|F|R, "detection" },
90 { "peak", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A|F|R, "detection" },
91 { "rms", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A|F|R, "detection" },
92 { "level_sc", "set sidechain gain", OFFSET(level_sc), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A|F|R },
93 { "mix", "set mix", OFFSET(mix), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0, 1, A|F|R },
97 #define sidechaincompress_options options
98 AVFILTER_DEFINE_CLASS(sidechaincompress);
100 // A fake infinity value (because real infinity may break some hosts)
101 #define FAKE_INFINITY (65536.0 * 65536.0)
103 // Check for infinity (with appropriate-ish tolerance)
104 #define IS_FAKE_INFINITY(value) (fabs(value-FAKE_INFINITY) < 1.0)
106 static double output_gain(double lin_slope, double ratio, double thres,
107 double knee, double knee_start, double knee_stop,
108 double compressed_knee_start,
109 double compressed_knee_stop,
110 int detection, int mode)
112 double slope = log(lin_slope);
119 if (IS_FAKE_INFINITY(ratio)) {
123 gain = (slope - thres) / ratio + thres;
128 if (knee > 1.0 && slope > knee_start)
129 gain = hermite_interpolation(slope, knee_stop, knee_start,
130 knee_stop, compressed_knee_start,
133 if (knee > 1.0 && slope < knee_stop)
134 gain = hermite_interpolation(slope, knee_start, knee_stop,
135 knee_start, compressed_knee_stop,
139 return exp(gain - slope);
142 static int compressor_config_output(AVFilterLink *outlink)
144 AVFilterContext *ctx = outlink->src;
145 SidechainCompressContext *s = ctx->priv;
147 s->thres = log(s->threshold);
148 s->lin_knee_start = s->threshold / sqrt(s->knee);
149 s->lin_knee_stop = s->threshold * sqrt(s->knee);
150 s->adj_knee_start = s->lin_knee_start * s->lin_knee_start;
151 s->adj_knee_stop = s->lin_knee_stop * s->lin_knee_stop;
152 s->knee_start = log(s->lin_knee_start);
153 s->knee_stop = log(s->lin_knee_stop);
154 s->compressed_knee_start = (s->knee_start - s->thres) / s->ratio + s->thres;
155 s->compressed_knee_stop = (s->knee_stop - s->thres) / s->ratio + s->thres;
157 s->attack_coeff = FFMIN(1., 1. / (s->attack * outlink->sample_rate / 4000.));
158 s->release_coeff = FFMIN(1., 1. / (s->release * outlink->sample_rate / 4000.));
163 static void compressor(SidechainCompressContext *s,
164 const double *src, double *dst, const double *scsrc, int nb_samples,
165 double level_in, double level_sc,
166 AVFilterLink *inlink, AVFilterLink *sclink)
168 const double makeup = s->makeup;
169 const double mix = s->mix;
172 for (i = 0; i < nb_samples; i++) {
173 double abs_sample, gain = 1.0;
177 abs_sample = fabs(scsrc[0] * level_sc);
180 for (c = 1; c < sclink->channels; c++)
181 abs_sample = FFMAX(fabs(scsrc[c] * level_sc), abs_sample);
183 for (c = 1; c < sclink->channels; c++)
184 abs_sample += fabs(scsrc[c] * level_sc);
186 abs_sample /= sclink->channels;
190 abs_sample *= abs_sample;
192 s->lin_slope += (abs_sample - s->lin_slope) * (abs_sample > s->lin_slope ? s->attack_coeff : s->release_coeff);
195 detector = (s->detection ? s->adj_knee_stop : s->lin_knee_stop);
196 detected = s->lin_slope < detector;
198 detector = (s->detection ? s->adj_knee_start : s->lin_knee_start);
199 detected = s->lin_slope > detector;
202 if (s->lin_slope > 0.0 && detected)
203 gain = output_gain(s->lin_slope, s->ratio, s->thres, s->knee,
204 s->knee_start, s->knee_stop,
205 s->compressed_knee_start,
206 s->compressed_knee_stop,
207 s->detection, s->mode);
209 for (c = 0; c < inlink->channels; c++)
210 dst[c] = src[c] * level_in * (gain * makeup * mix + (1. - mix));
212 src += inlink->channels;
213 dst += inlink->channels;
214 scsrc += sclink->channels;
218 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
219 char *res, int res_len, int flags)
223 ret = ff_filter_process_command(ctx, cmd, args, res, res_len, flags);
227 compressor_config_output(ctx->outputs[0]);
232 #if CONFIG_SIDECHAINCOMPRESS_FILTER
233 static int activate(AVFilterContext *ctx)
235 SidechainCompressContext *s = ctx->priv;
236 AVFrame *out = NULL, *in[2] = { NULL };
237 int ret, i, nb_samples;
240 FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx);
241 if ((ret = ff_inlink_consume_frame(ctx->inputs[0], &in[0])) > 0) {
242 av_audio_fifo_write(s->fifo[0], (void **)in[0]->extended_data,
244 av_frame_free(&in[0]);
248 if ((ret = ff_inlink_consume_frame(ctx->inputs[1], &in[1])) > 0) {
249 av_audio_fifo_write(s->fifo[1], (void **)in[1]->extended_data,
251 av_frame_free(&in[1]);
256 nb_samples = FFMIN(av_audio_fifo_size(s->fifo[0]), av_audio_fifo_size(s->fifo[1]));
258 out = ff_get_audio_buffer(ctx->outputs[0], nb_samples);
260 return AVERROR(ENOMEM);
261 for (i = 0; i < 2; i++) {
262 in[i] = ff_get_audio_buffer(ctx->inputs[i], nb_samples);
264 av_frame_free(&in[0]);
265 av_frame_free(&in[1]);
267 return AVERROR(ENOMEM);
269 av_audio_fifo_read(s->fifo[i], (void **)in[i]->data, nb_samples);
272 dst = (double *)out->data[0];
274 s->pts += av_rescale_q(nb_samples, (AVRational){1, ctx->outputs[0]->sample_rate}, ctx->outputs[0]->time_base);
276 compressor(s, (double *)in[0]->data[0], dst,
277 (double *)in[1]->data[0], nb_samples,
278 s->level_in, s->level_sc,
279 ctx->inputs[0], ctx->inputs[1]);
281 av_frame_free(&in[0]);
282 av_frame_free(&in[1]);
284 ret = ff_filter_frame(ctx->outputs[0], out);
288 FF_FILTER_FORWARD_STATUS(ctx->inputs[0], ctx->outputs[0]);
289 FF_FILTER_FORWARD_STATUS(ctx->inputs[1], ctx->outputs[0]);
290 if (ff_outlink_frame_wanted(ctx->outputs[0])) {
291 if (!av_audio_fifo_size(s->fifo[0]))
292 ff_inlink_request_frame(ctx->inputs[0]);
293 if (!av_audio_fifo_size(s->fifo[1]))
294 ff_inlink_request_frame(ctx->inputs[1]);
299 static int query_formats(AVFilterContext *ctx)
301 AVFilterFormats *formats;
302 AVFilterChannelLayouts *layouts = NULL;
303 static const enum AVSampleFormat sample_fmts[] = {
309 if (!ctx->inputs[0]->incfg.channel_layouts ||
310 !ctx->inputs[0]->incfg.channel_layouts->nb_channel_layouts) {
311 av_log(ctx, AV_LOG_WARNING,
312 "No channel layout for input 1\n");
313 return AVERROR(EAGAIN);
316 if ((ret = ff_add_channel_layout(&layouts, ctx->inputs[0]->incfg.channel_layouts->channel_layouts[0])) < 0 ||
317 (ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->incfg.channel_layouts)) < 0)
320 for (i = 0; i < 2; i++) {
321 layouts = ff_all_channel_counts();
322 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->outcfg.channel_layouts)) < 0)
326 formats = ff_make_format_list(sample_fmts);
327 if ((ret = ff_set_common_formats(ctx, formats)) < 0)
330 formats = ff_all_samplerates();
331 return ff_set_common_samplerates(ctx, formats);
334 static int config_output(AVFilterLink *outlink)
336 AVFilterContext *ctx = outlink->src;
337 SidechainCompressContext *s = ctx->priv;
339 if (ctx->inputs[0]->sample_rate != ctx->inputs[1]->sample_rate) {
340 av_log(ctx, AV_LOG_ERROR,
341 "Inputs must have the same sample rate "
342 "%d for in0 vs %d for in1\n",
343 ctx->inputs[0]->sample_rate, ctx->inputs[1]->sample_rate);
344 return AVERROR(EINVAL);
347 outlink->sample_rate = ctx->inputs[0]->sample_rate;
348 outlink->time_base = ctx->inputs[0]->time_base;
349 outlink->channel_layout = ctx->inputs[0]->channel_layout;
350 outlink->channels = ctx->inputs[0]->channels;
352 s->fifo[0] = av_audio_fifo_alloc(ctx->inputs[0]->format, ctx->inputs[0]->channels, 1024);
353 s->fifo[1] = av_audio_fifo_alloc(ctx->inputs[1]->format, ctx->inputs[1]->channels, 1024);
354 if (!s->fifo[0] || !s->fifo[1])
355 return AVERROR(ENOMEM);
357 compressor_config_output(outlink);
362 static av_cold void uninit(AVFilterContext *ctx)
364 SidechainCompressContext *s = ctx->priv;
366 av_audio_fifo_free(s->fifo[0]);
367 av_audio_fifo_free(s->fifo[1]);
370 static const AVFilterPad sidechaincompress_inputs[] = {
373 .type = AVMEDIA_TYPE_AUDIO,
376 .type = AVMEDIA_TYPE_AUDIO,
381 static const AVFilterPad sidechaincompress_outputs[] = {
384 .type = AVMEDIA_TYPE_AUDIO,
385 .config_props = config_output,
390 AVFilter ff_af_sidechaincompress = {
391 .name = "sidechaincompress",
392 .description = NULL_IF_CONFIG_SMALL("Sidechain compressor."),
393 .priv_size = sizeof(SidechainCompressContext),
394 .priv_class = &sidechaincompress_class,
395 .query_formats = query_formats,
396 .activate = activate,
398 .inputs = sidechaincompress_inputs,
399 .outputs = sidechaincompress_outputs,
400 .process_command = process_command,
402 #endif /* CONFIG_SIDECHAINCOMPRESS_FILTER */
404 #if CONFIG_ACOMPRESSOR_FILTER
405 static int acompressor_filter_frame(AVFilterLink *inlink, AVFrame *in)
407 const double *src = (const double *)in->data[0];
408 AVFilterContext *ctx = inlink->dst;
409 SidechainCompressContext *s = ctx->priv;
410 AVFilterLink *outlink = ctx->outputs[0];
414 if (av_frame_is_writable(in)) {
417 out = ff_get_audio_buffer(outlink, in->nb_samples);
420 return AVERROR(ENOMEM);
422 av_frame_copy_props(out, in);
424 dst = (double *)out->data[0];
426 compressor(s, src, dst, src, in->nb_samples,
427 s->level_in, s->level_in,
432 return ff_filter_frame(outlink, out);
435 static int acompressor_query_formats(AVFilterContext *ctx)
437 AVFilterFormats *formats;
438 AVFilterChannelLayouts *layouts;
439 static const enum AVSampleFormat sample_fmts[] = {
445 layouts = ff_all_channel_counts();
447 return AVERROR(ENOMEM);
448 ret = ff_set_common_channel_layouts(ctx, layouts);
452 formats = ff_make_format_list(sample_fmts);
454 return AVERROR(ENOMEM);
455 ret = ff_set_common_formats(ctx, formats);
459 formats = ff_all_samplerates();
461 return AVERROR(ENOMEM);
462 return ff_set_common_samplerates(ctx, formats);
465 #define acompressor_options options
466 AVFILTER_DEFINE_CLASS(acompressor);
468 static const AVFilterPad acompressor_inputs[] = {
471 .type = AVMEDIA_TYPE_AUDIO,
472 .filter_frame = acompressor_filter_frame,
477 static const AVFilterPad acompressor_outputs[] = {
480 .type = AVMEDIA_TYPE_AUDIO,
481 .config_props = compressor_config_output,
486 AVFilter ff_af_acompressor = {
487 .name = "acompressor",
488 .description = NULL_IF_CONFIG_SMALL("Audio compressor."),
489 .priv_size = sizeof(SidechainCompressContext),
490 .priv_class = &acompressor_class,
491 .query_formats = acompressor_query_formats,
492 .inputs = acompressor_inputs,
493 .outputs = acompressor_outputs,
494 .process_command = process_command,
496 #endif /* CONFIG_ACOMPRESSOR_FILTER */