avfilter: add sidechain compress audio filter

Signed-off-by: Paul B Mahol <onemda@gmail.com>

Changelog

@@ -24,6 +24,7 @@ version <next>:
 - Random filter
 - deband filter
 - AAC fixed-point decoding
+- sidechaincompress audio filter
 
 
 version 2.7:

doc/filters.texi

@@ -622,6 +622,7 @@ slope
 Specify the band-width of a filter in width_type units.
 @end table
 
+@anchor{amerge}
 @section amerge
 
 Merge two or more audio streams into a single multi-channel stream.
@@ -2020,6 +2021,7 @@ Applies only to double-pole filter.
 The default is 0.707q and gives a Butterworth response.
 @end table
 
+@anchor{pan}
 @section pan
 
 Mix channels with specific gain levels. The filter accepts the output
@@ -2121,6 +2123,66 @@ At end of filtering it displays @code{track_gain} and @code{track_peak}.
 Convert the audio sample format, sample rate and channel layout. It is
 not meant to be used directly.
 
+@section sidechaincompress
+
+This filter acts like normal compressor but has the ability to compress
+detected signal using second input signal.
+It needs two input streams and returns one output stream.
+First input stream will be processed depending on second stream signal.
+The filtered signal then can be filtered with other filters in later stages of
+processing. See @ref{pan} and @ref{amerge} filter.
+
+The filter accepts the following options:
+
+@table @option
+@item threshold
+If a signal of second stream raises above this level it will affect the gain
+reduction of first stream.
+By default is 0.125. Range is between 0.00097563 and 1.
+
+@item ratio
+Set a ratio about which the signal is reduced. 1:2 means that if the level
+raised 4dB above the threshold, it will be only 2dB above after the reduction.
+Default is 2. Range is between 1 and 20.
+
+@item attack
+Amount of milliseconds the signal has to rise above the threshold before gain
+reduction starts. Default is 20. Range is between 0.01 and 2000.
+
+@item release
+Amount of milliseconds the signal has to fall bellow the threshold before
+reduction is decreased again. Default is 250. Range is between 0.01 and 9000.
+
+@item makeup
+Set the amount by how much signal will be amplified after processing.
+Default is 2. Range is from 1 and 64.
+
+@item knee
+Curve the sharp knee around the threshold to enter gain reduction more softly.
+Default is 2.82843. Range is between 1 and 8.
+
+@item link
+Choose if the @code{average} level between all channels of side-chain stream
+or the louder(@code{maximum}) channel of side-chain stream affects the
+reduction. Default is @code{average}.
+
+@item detection
+Should the exact signal be taken in case of @code{peak} or an RMS one in case
+of @code{rms}. Default is @code{rms} which is mainly smoother.
+@end table
+
+@subsection Examples
+
+@itemize
+@item
+Full ffmpeg example taking 2 audio inputs, 1st input to be compressed
+depending on the signal of 2nd input and later compressed signal to be
+merged with 2nd input:
+@example
+ffmpeg -i main.flac -i sidechain.flac -filter_complex "[1:a]asplit=2[sc][mix];[0:a][sc]sidechaincompress[compr];[compr][mix]amerge"
+@end example
+@end itemize
+
 @section silencedetect
 
 Detect silence in an audio stream.

libavfilter/Makefile

@@ -80,6 +80,7 @@ OBJS-$(CONFIG_LOWPASS_FILTER)                += af_biquads.o
 OBJS-$(CONFIG_PAN_FILTER)                    += af_pan.o
 OBJS-$(CONFIG_REPLAYGAIN_FILTER)             += af_replaygain.o
 OBJS-$(CONFIG_RESAMPLE_FILTER)               += af_resample.o
+OBJS-$(CONFIG_SIDECHAINCOMPRESS_FILTER)      += af_sidechaincompress.o
 OBJS-$(CONFIG_SILENCEDETECT_FILTER)          += af_silencedetect.o
 OBJS-$(CONFIG_SILENCEREMOVE_FILTER)          += af_silenceremove.o
 OBJS-$(CONFIG_TREBLE_FILTER)                 += af_biquads.o

libavfilter/af_sidechaincompress.c

+/*
+ * Copyright (C) 2001-2010 Krzysztof Foltman, Markus Schmidt, Thor Harald Johansen and others
+ * Copyright (c) 2015 Paul B Mahol
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * Sidechain compressor filter
+ */
+
+#include "libavutil/avassert.h"
+#include "libavutil/channel_layout.h"
+#include "libavutil/common.h"
+#include "libavutil/opt.h"
+
+#include "audio.h"
+#include "avfilter.h"
+#include "formats.h"
+#include "internal.h"
+
+typedef struct SidechainCompressContext {
+    const AVClass *class;
+
+    double attack, attack_coeff;
+    double release, release_coeff;
+    double lin_slope;
+    double ratio;
+    double threshold;
+    double makeup;
+    double thres;
+    double knee;
+    double knee_start;
+    double knee_stop;
+    double lin_knee_start;
+    double compressed_knee_stop;
+    int link;
+    int detection;
+
+    AVFrame *input_frame[2];
+} SidechainCompressContext;
+
+#define OFFSET(x) offsetof(SidechainCompressContext, x)
+#define A AV_OPT_FLAG_AUDIO_PARAM
+#define F AV_OPT_FLAG_FILTERING_PARAM
+
+static const AVOption sidechaincompress_options[] = {
+    { "threshold", "set threshold",    OFFSET(threshold), AV_OPT_TYPE_DOUBLE, {.dbl=0.125}, 0.000976563,    1, A|F },
+    { "ratio",     "set ratio",        OFFSET(ratio),     AV_OPT_TYPE_DOUBLE, {.dbl=2},               1,   20, A|F },
+    { "attack",    "set attack",       OFFSET(attack),    AV_OPT_TYPE_DOUBLE, {.dbl=20},           0.01, 2000, A|F },
+    { "release",   "set release",      OFFSET(release),   AV_OPT_TYPE_DOUBLE, {.dbl=250},          0.01, 9000, A|F },
+    { "makeup",    "set make up gain", OFFSET(makeup),    AV_OPT_TYPE_DOUBLE, {.dbl=2},               1,   64, A|F },
+    { "knee",      "set knee",         OFFSET(knee),      AV_OPT_TYPE_DOUBLE, {.dbl=2.82843},         1,    8, A|F },
+    { "link",      "set link type",    OFFSET(link),      AV_OPT_TYPE_INT,    {.i64=0},               0,    1, A|F, "link" },
+    {   "average", 0,                  0,                 AV_OPT_TYPE_CONST,  {.i64=0},               0,    0, A|F, "link" },
+    {   "maximum", 0,                  0,                 AV_OPT_TYPE_CONST,  {.i64=1},               0,    0, A|F, "link" },
+    { "detection", "set detection",    OFFSET(detection), AV_OPT_TYPE_INT,    {.i64=1},               0,    1, A|F, "detection" },
+    {   "peak",    0,                  0,                 AV_OPT_TYPE_CONST,  {.i64=0},               0,    0, A|F, "detection" },
+    {   "rms",     0,                  0,                 AV_OPT_TYPE_CONST,  {.i64=1},               0,    0, A|F, "detection" },
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(sidechaincompress);
+
+static av_cold int init(AVFilterContext *ctx)
+{
+    SidechainCompressContext *s = ctx->priv;
+
+    s->thres = log(s->threshold);
+    s->lin_knee_start = s->threshold / sqrt(s->knee);
+    s->knee_start = log(s->lin_knee_start);
+    s->knee_stop = log(s->threshold * sqrt(s->knee));
+    s->compressed_knee_stop = (s->knee_stop - s->thres) / s->ratio + s->thres;
+
+    return 0;
+}
+
+static inline float hermite_interpolation(float x, float x0, float x1,
+                                          float p0, float p1,
+                                          float m0, float m1)
+{
+    float width = x1 - x0;
+    float t = (x - x0) / width;
+    float t2, t3;
+    float ct0, ct1, ct2, ct3;
+
+    m0 *= width;
+    m1 *= width;
+
+    t2 = t*t;
+    t3 = t2*t;
+    ct0 = p0;
+    ct1 = m0;
+
+    ct2 = -3 * p0 - 2 * m0 + 3 * p1 - m1;
+    ct3 = 2 * p0 + m0  - 2 * p1 + m1;
+
+    return ct3 * t3 + ct2 * t2 + ct1 * t + ct0;
+}
+
+// A fake infinity value (because real infinity may break some hosts)
+#define FAKE_INFINITY (65536.0 * 65536.0)
+
+// Check for infinity (with appropriate-ish tolerance)
+#define IS_FAKE_INFINITY(value) (fabs(value-FAKE_INFINITY) < 1.0)
+
+static double output_gain(double lin_slope, double ratio, double thres,
+                          double knee, double knee_start, double knee_stop,
+                          double compressed_knee_stop, int detection)
+{
+    double slope = log(lin_slope);
+    double gain = 0.0;
+    double delta = 0.0;
+
+    if (detection)
+        slope *= 0.5;
+
+    if (IS_FAKE_INFINITY(ratio)) {
+        gain = thres;
+        delta = 0.0;
+    } else {
+        gain = (slope - thres) / ratio + thres;
+        delta = 1.0 / ratio;
+    }
+
+    if (knee > 1.0 && slope < knee_stop)
+        gain = hermite_interpolation(slope, knee_start, knee_stop,
+                                     knee_start, compressed_knee_stop,
+                                     1.0, delta);
+
+    return exp(gain - slope);
+}
+
+static int filter_frame(AVFilterLink *link, AVFrame *frame)
+{
+    AVFilterContext *ctx = link->dst;
+    SidechainCompressContext *s = ctx->priv;
+    AVFilterLink *sclink = ctx->inputs[1];
+    AVFilterLink *outlink = ctx->outputs[0];
+    const double makeup = s->makeup;
+    const double *scsrc;
+    double *sample;
+    int nb_samples;
+    int ret, i, c;
+
+    for (i = 0; i < 2; i++)
+        if (link == ctx->inputs[i])
+            break;
+    av_assert0(!s->input_frame[i]);
+    s->input_frame[i] = frame;
+
+    if (!s->input_frame[0] || !s->input_frame[1])
+        return 0;
+
+    nb_samples = FFMIN(s->input_frame[0]->nb_samples,
+                       s->input_frame[1]->nb_samples);
+
+    sample = (double *)s->input_frame[0]->data[0];
+    scsrc = (const double *)s->input_frame[1]->data[0];
+
+    for (i = 0; i < nb_samples; i++) {
+        double abs_sample, gain = 1.0;
+
+        abs_sample = FFABS(scsrc[0]);
+
+        if (s->link == 1) {
+            for (c = 1; c < sclink->channels; c++)
+                abs_sample = FFMAX(FFABS(scsrc[c]), abs_sample);
+        } else {
+            for (c = 1; c < sclink->channels; c++)
+                abs_sample += FFABS(scsrc[c]);
+
+            abs_sample /= sclink->channels;
+        }
+
+        if (s->detection)
+            abs_sample *= abs_sample;
+
+        s->lin_slope += (abs_sample - s->lin_slope) * (abs_sample > s->lin_slope ? s->attack_coeff : s->release_coeff);
+
+        if (s->lin_slope > 0.0 && s->lin_slope > s->lin_knee_start)
+            gain = output_gain(s->lin_slope, s->ratio, s->thres, s->knee,
+                               s->knee_start, s->knee_stop,
+                               s->compressed_knee_stop, s->detection);
+
+        for (c = 0; c < outlink->channels; c++)
+            sample[c] *= gain * makeup;
+
+        sample += outlink->channels;
+        scsrc += sclink->channels;
+    }
+
+    ret = ff_filter_frame(outlink, s->input_frame[0]);
+
+    s->input_frame[0] = NULL;
+    av_frame_free(&s->input_frame[1]);
+
+    return ret;
+}
+
+static int request_frame(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    SidechainCompressContext *s = ctx->priv;
+    int i, ret;
+
+    /* get a frame on each input */
+    for (i = 0; i < 2; i++) {
+        AVFilterLink *inlink = ctx->inputs[i];
+        if (!s->input_frame[i] &&
+            (ret = ff_request_frame(inlink)) < 0)
+            return ret;
+
+        /* request the same number of samples on all inputs */
+        if (i == 0)
+            ctx->inputs[1]->request_samples = s->input_frame[0]->nb_samples;
+    }
+
+    return 0;
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+    AVFilterFormats *formats;
+    AVFilterChannelLayouts *layouts = NULL;
+    static const enum AVSampleFormat sample_fmts[] = {
+        AV_SAMPLE_FMT_DBL,
+        AV_SAMPLE_FMT_NONE
+    };
+    int ret, i;
+
+    if (!ctx->inputs[0]->in_channel_layouts ||
+        !ctx->inputs[0]->in_channel_layouts->nb_channel_layouts) {
+        av_log(ctx, AV_LOG_WARNING,
+               "No channel layout for input 1\n");
+            return AVERROR(EAGAIN);
+    }
+
+    ff_add_channel_layout(&layouts, ctx->inputs[0]->in_channel_layouts->channel_layouts[0]);
+    if (!layouts)
+        return AVERROR(ENOMEM);
+    ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts);
+
+    for (i = 0; i < 2; i++) {
+        layouts = ff_all_channel_layouts();
+        if (!layouts)
+            return AVERROR(ENOMEM);
+        ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts);
+    }
+
+    formats = ff_make_format_list(sample_fmts);
+    if (!formats)
+        return AVERROR(ENOMEM);
+    ret = ff_set_common_formats(ctx, formats);
+    if (ret < 0)
+        return ret;
+
+    formats = ff_all_samplerates();
+    if (!formats)
+        return AVERROR(ENOMEM);
+    return ff_set_common_samplerates(ctx, formats);
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    SidechainCompressContext *s = ctx->priv;
+
+    if (ctx->inputs[0]->sample_rate != ctx->inputs[1]->sample_rate) {
+        av_log(ctx, AV_LOG_ERROR,
+               "Inputs must have the same sample rate "
+               "%d for in0 vs %d for in1\n",
+               ctx->inputs[0]->sample_rate, ctx->inputs[1]->sample_rate);
+        return AVERROR(EINVAL);
+    }
+
+    outlink->sample_rate = ctx->inputs[0]->sample_rate;
+    outlink->time_base   = ctx->inputs[0]->time_base;
+    outlink->channel_layout = ctx->inputs[0]->channel_layout;
+    outlink->channels = ctx->inputs[0]->channels;
+
+    s->attack_coeff = FFMIN(1.f, 1.f / (s->attack * outlink->sample_rate / 4000.f));
+    s->release_coeff = FFMIN(1.f, 1.f / (s->release * outlink->sample_rate / 4000.f));
+
+    return 0;
+}
+
+static const AVFilterPad sidechaincompress_inputs[] = {
+    {
+        .name           = "main",
+        .type           = AVMEDIA_TYPE_AUDIO,
+        .filter_frame   = filter_frame,
+        .needs_writable = 1,
+        .needs_fifo     = 1,
+    },{
+        .name           = "sidechain",
+        .type           = AVMEDIA_TYPE_AUDIO,
+        .filter_frame   = filter_frame,
+        .needs_fifo     = 1,
+    },
+    { NULL }
+};
+
+static const AVFilterPad sidechaincompress_outputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_AUDIO,
+        .config_props  = config_output,
+        .request_frame = request_frame,
+    },
+    { NULL }
+};
+
+AVFilter ff_af_sidechaincompress = {
+    .name           = "sidechaincompress",
+    .description    = NULL_IF_CONFIG_SMALL("Sidechain compressor."),
+    .priv_size      = sizeof(SidechainCompressContext),
+    .priv_class     = &sidechaincompress_class,
+    .init           = init,
+    .query_formats  = query_formats,
+    .inputs         = sidechaincompress_inputs,
+    .outputs        = sidechaincompress_outputs,
+};

libavfilter/allfilters.c

@@ -96,6 +96,7 @@ void avfilter_register_all(void)
     REGISTER_FILTER(PAN,            pan,            af);
     REGISTER_FILTER(REPLAYGAIN,     replaygain,     af);
     REGISTER_FILTER(RESAMPLE,       resample,       af);
+    REGISTER_FILTER(SIDECHAINCOMPRESS, sidechaincompress, af);
     REGISTER_FILTER(SILENCEDETECT,  silencedetect,  af);
     REGISTER_FILTER(SILENCEREMOVE,  silenceremove,  af);
     REGISTER_FILTER(TREBLE,         treble,         af);

libavfilter/version.h

@@ -30,7 +30,7 @@
 #include "libavutil/version.h"
 
 #define LIBAVFILTER_VERSION_MAJOR  5
-#define LIBAVFILTER_VERSION_MINOR  28
+#define LIBAVFILTER_VERSION_MINOR  29
 #define LIBAVFILTER_VERSION_MICRO 100
 
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \