lavfi: add aecho filter
authorPaul B Mahol <onemda@gmail.com>
Mon, 8 Jul 2013 13:44:35 +0000 (13:44 +0000)
committerPaul B Mahol <onemda@gmail.com>
Wed, 10 Jul 2013 12:10:58 +0000 (12:10 +0000)
Signed-off-by: Paul B Mahol <onemda@gmail.com>
Changelog
doc/filters.texi
libavfilter/Makefile
libavfilter/af_aecho.c [new file with mode: 0644]
libavfilter/allfilters.c
libavfilter/version.h

index c439633..ecbb39d 100644 (file)
--- a/Changelog
+++ b/Changelog
@@ -3,6 +3,8 @@ releases are sorted from youngest to oldest.
 
 version <next>
 
+- aecho filter
+
 
 version 2.0:
 
index 33436ad..92f8612 100644 (file)
@@ -347,6 +347,66 @@ aconvert=u8:auto
 @end example
 @end itemize
 
+@section aecho
+
+Apply echoing to the input audio.
+
+Echoes are reflected sound and can occur naturally amongst mountains
+(and sometimes large buildings) when talking or shouting; digital echo
+effects emulate this behaviour and are often used to help fill out the
+sound of a single instrument or vocal. The time difference between the
+original signal and the reflection is the @code{delay}, and the
+loudness of the reflected signal is the @code{decay}.
+Multiple echoes can have different delays and decays.
+
+A description of the accepted parameters follows.
+
+@table @option
+@item in_gain
+Set input gain of reflected signal. Default is @code{0.6}.
+
+@item out_gain
+Set output gain of reflected signal. Default is @code{0.3}.
+
+@item delays
+Set list of time intervals in milliseconds between original signal and reflections
+separated by '|'. Allowed range for each @code{delay} is @code{(0 - 90000.0]}.
+Default is @code{1000}.
+
+@item decays
+Set list of loudnesses of reflected signals separated by '|'.
+Allowed range for each @code{decay} is @code{(0 - 1.0]}.
+Default is @code{0.5}.
+@end table
+
+@subsection Examples
+
+@itemize
+@item
+Make it sound as if there are twice as many instruments as are actually playing:
+@example
+aecho=0.8:0.88:60:0.4
+@end example
+
+@item
+If delay is very short, then it sound like a (metallic) robot playing music:
+@example
+aecho=0.8:0.88:6:0.4
+@end example
+
+@item
+A longer delay will sound like an open air concert in the mountains:
+@example
+aecho=0.8:0.9:1000:0.3
+@end example
+
+@item
+Same as above but with one more mountain:
+@example
+aecho=0.8:0.9:1000|1800:0.3|0.25
+@end example
+@end itemize
+
 @section afade
 
 Apply fade-in/out effect to input audio.
index cf76ee1..306b24c 100644 (file)
@@ -52,6 +52,7 @@ OBJS-$(CONFIG_AVFORMAT)                      += lavfutils.o
 OBJS-$(CONFIG_SWSCALE)                       += lswsutils.o
 
 OBJS-$(CONFIG_ACONVERT_FILTER)               += af_aconvert.o
+OBJS-$(CONFIG_AECHO_FILTER)                  += af_aecho.o
 OBJS-$(CONFIG_AFADE_FILTER)                  += af_afade.o
 OBJS-$(CONFIG_AFORMAT_FILTER)                += af_aformat.o
 OBJS-$(CONFIG_AINTERLEAVE_FILTER)            += f_interleave.o
diff --git a/libavfilter/af_aecho.c b/libavfilter/af_aecho.c
new file mode 100644 (file)
index 0000000..09bb2f6
--- /dev/null
@@ -0,0 +1,357 @@
+/*
+ * Copyright (c) 2013 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
+ *
+ */
+
+#include "libavutil/avstring.h"
+#include "libavutil/opt.h"
+#include "libavutil/samplefmt.h"
+#include "libavutil/avassert.h"
+#include "avfilter.h"
+#include "audio.h"
+#include "internal.h"
+
+typedef struct AudioEchoContext {
+    const AVClass *class;
+    float in_gain, out_gain;
+    char *delays, *decays;
+    float *delay, *decay;
+    int nb_echoes;
+    int delay_index;
+    uint8_t **delayptrs;
+    int max_samples, fade_out;
+    int *samples;
+    int64_t next_pts;
+
+    void (*echo_samples)(struct AudioEchoContext *ctx, uint8_t **delayptrs,
+                         uint8_t * const *src, uint8_t **dst,
+                         int nb_samples, int channels);
+} AudioEchoContext;
+
+#define OFFSET(x) offsetof(AudioEchoContext, x)
+#define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
+
+static const AVOption aecho_options[] = {
+    { "in_gain",  "set signal input gain",  OFFSET(in_gain),  AV_OPT_TYPE_FLOAT,  {.dbl=0.6}, 0, 1, A },
+    { "out_gain", "set signal output gain", OFFSET(out_gain), AV_OPT_TYPE_FLOAT,  {.dbl=0.3}, 0, 1, A },
+    { "delays",   "set list of signal delays", OFFSET(delays), AV_OPT_TYPE_STRING, {.str="1000"}, 0, 0, A },
+    { "decays",   "set list of signal decays", OFFSET(decays), AV_OPT_TYPE_STRING, {.str="0.5"}, 0, 0, A },
+    { NULL },
+};
+
+AVFILTER_DEFINE_CLASS(aecho);
+
+static void count_items(char *item_str, int *nb_items)
+{
+    char *p;
+
+    *nb_items = 1;
+    for (p = item_str; *p; p++) {
+        if (*p == '|')
+            (*nb_items)++;
+    }
+
+}
+
+static void fill_items(char *item_str, int *nb_items, float *items)
+{
+    char *p, *saveptr = NULL;
+    int i, new_nb_items = 0;
+
+    p = item_str;
+    for (i = 0; i < *nb_items; i++) {
+        char *tstr = av_strtok(p, "|", &saveptr);
+        p = NULL;
+        new_nb_items += sscanf(tstr, "%f", &items[i]) == 1;
+    }
+
+    *nb_items = new_nb_items;
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    AudioEchoContext *s = ctx->priv;
+
+    av_freep(&s->delay);
+    av_freep(&s->decay);
+    av_freep(&s->samples);
+
+    if (s->delayptrs)
+        av_freep(s->delayptrs[0]);
+    av_freep(&s->delayptrs);
+}
+
+static av_cold int init(AVFilterContext *ctx)
+{
+    AudioEchoContext *s = ctx->priv;
+    int nb_delays, nb_decays, i;
+
+    if (!s->delays || !s->decays) {
+        av_log(ctx, AV_LOG_ERROR, "Missing delays and/or decays.\n");
+        return AVERROR(EINVAL);
+    }
+
+    count_items(s->delays, &nb_delays);
+    count_items(s->decays, &nb_decays);
+
+    s->delay = av_realloc_f(s->delay, nb_delays, sizeof(*s->delay));
+    s->decay = av_realloc_f(s->decay, nb_decays, sizeof(*s->decay));
+    if (!s->delay || !s->decay)
+        return AVERROR(ENOMEM);
+
+    fill_items(s->delays, &nb_delays, s->delay);
+    fill_items(s->decays, &nb_decays, s->decay);
+
+    if (nb_delays != nb_decays) {
+        av_log(ctx, AV_LOG_ERROR, "Number of delays %d differs from number of decays %d.\n", nb_delays, nb_decays);
+        return AVERROR(EINVAL);
+    }
+
+    s->nb_echoes = nb_delays;
+    if (!s->nb_echoes) {
+        av_log(ctx, AV_LOG_ERROR, "At least one decay & delay must be set.\n");
+        return AVERROR(EINVAL);
+    }
+
+    s->samples = av_realloc_f(s->samples, nb_delays, sizeof(*s->samples));
+    if (!s->samples)
+        return AVERROR(ENOMEM);
+
+    for (i = 0; i < nb_delays; i++) {
+        if (s->delay[i] <= 0 || s->delay[i] > 90000) {
+            av_log(ctx, AV_LOG_ERROR, "delay[%d]: %f is out of allowed range: (0, 90000]\n", i, s->delay[i]);
+            return AVERROR(EINVAL);
+        }
+        if (s->decay[i] <= 0 || s->decay[i] > 1) {
+            av_log(ctx, AV_LOG_ERROR, "decay[%d]: %f is out of allowed range: (0, 1]\n", i, s->decay[i]);
+            return AVERROR(EINVAL);
+        }
+    }
+
+    s->next_pts = AV_NOPTS_VALUE;
+
+    av_log(ctx, AV_LOG_DEBUG, "nb_echoes:%d\n", s->nb_echoes);
+    return 0;
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+    AVFilterChannelLayouts *layouts;
+    AVFilterFormats *formats;
+    static const enum AVSampleFormat sample_fmts[] = {
+        AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S32P,
+        AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_DBLP,
+        AV_SAMPLE_FMT_NONE
+    };
+
+    layouts = ff_all_channel_layouts();
+    if (!layouts)
+        return AVERROR(ENOMEM);
+    ff_set_common_channel_layouts(ctx, layouts);
+
+    formats = ff_make_format_list(sample_fmts);
+    if (!formats)
+        return AVERROR(ENOMEM);
+    ff_set_common_formats(ctx, formats);
+
+    formats = ff_all_samplerates();
+    if (!formats)
+        return AVERROR(ENOMEM);
+    ff_set_common_samplerates(ctx, formats);
+
+    return 0;
+}
+
+#define MOD(a, b) (((a) >= (b)) ? (a) - (b) : (a))
+
+#define ECHO(name, type, min, max)                                          \
+static void echo_samples_## name ##p(AudioEchoContext *ctx,                 \
+                                     uint8_t **delayptrs,                   \
+                                     uint8_t * const *src, uint8_t **dst,   \
+                                     int nb_samples, int channels)          \
+{                                                                           \
+    const double out_gain = ctx->out_gain;                                  \
+    const double in_gain = ctx->in_gain;                                    \
+    const int nb_echoes = ctx->nb_echoes;                                   \
+    const int max_samples = ctx->max_samples;                               \
+    int i, j, chan, index;                                                  \
+                                                                            \
+    for (chan = 0; chan < channels; chan++) {                               \
+        const type *s = (type *)src[chan];                                  \
+        type *d = (type *)dst[chan];                                        \
+        type *dbuf = (type *)delayptrs[chan];                               \
+                                                                            \
+        index = ctx->delay_index;                                           \
+        for (i = 0; i < nb_samples; i++, s++, d++) {                        \
+            double out, in;                                                 \
+                                                                            \
+            in = *s;                                                        \
+            out = in * in_gain;                                             \
+            for (j = 0; j < nb_echoes; j++) {                               \
+                int ix = index + max_samples - ctx->samples[j];             \
+                ix = MOD(ix, max_samples);                                  \
+                out += dbuf[ix] * ctx->decay[j];                            \
+            }                                                               \
+            out *= out_gain;                                                \
+                                                                            \
+            *d = av_clipd(out, min, max);                                   \
+            dbuf[index] = in;                                               \
+                                                                            \
+            index = MOD(index + 1, max_samples);                            \
+        }                                                                   \
+    }                                                                       \
+    ctx->delay_index = index;                                               \
+}
+
+ECHO(dbl, double,  -1.0,      1.0      )
+ECHO(flt, float,   -1.0,      1.0      )
+ECHO(s16, int16_t, INT16_MIN, INT16_MAX)
+ECHO(s32, int32_t, INT32_MIN, INT32_MAX)
+
+static int config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    AudioEchoContext *s = ctx->priv;
+    float volume = 1.0;
+    int i;
+
+    for (i = 0; i < s->nb_echoes; i++) {
+        s->samples[i] = s->delay[i] * outlink->sample_rate / 1000.0;
+        s->max_samples = FFMAX(s->max_samples, s->samples[i]);
+        volume += s->decay[i];
+    }
+
+    if (s->max_samples <= 0) {
+        av_log(ctx, AV_LOG_ERROR, "Nothing to echo - missing delay samples.\n");
+        return AVERROR(EINVAL);
+    }
+    s->fade_out = s->max_samples;
+
+    if (volume * s->in_gain * s->out_gain > 1.0)
+        av_log(ctx, AV_LOG_WARNING,
+               "out_gain %f can cause saturation of output\n", s->out_gain);
+
+    switch (outlink->format) {
+    case AV_SAMPLE_FMT_DBLP: s->echo_samples = echo_samples_dblp; break;
+    case AV_SAMPLE_FMT_FLTP: s->echo_samples = echo_samples_fltp; break;
+    case AV_SAMPLE_FMT_S16P: s->echo_samples = echo_samples_s16p; break;
+    case AV_SAMPLE_FMT_S32P: s->echo_samples = echo_samples_s32p; break;
+    }
+
+
+    if (s->delayptrs)
+        av_freep(s->delayptrs[0]);
+    av_freep(&s->delayptrs);
+
+    return av_samples_alloc_array_and_samples(&s->delayptrs, NULL,
+                                              outlink->channels,
+                                              s->max_samples,
+                                              outlink->format, 0);
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *frame)
+{
+    AVFilterContext *ctx = inlink->dst;
+    AudioEchoContext *s = ctx->priv;
+    AVFrame *out_frame;
+
+    if (av_frame_is_writable(frame)) {
+        out_frame = frame;
+    } else {
+        out_frame = ff_get_audio_buffer(inlink, frame->nb_samples);
+        if (!out_frame)
+            return AVERROR(ENOMEM);
+        av_frame_copy_props(out_frame, frame);
+    }
+
+    s->echo_samples(s, s->delayptrs, frame->data, out_frame->data,
+                    frame->nb_samples, inlink->channels);
+
+    if (frame != out_frame)
+        av_frame_free(&frame);
+
+    s->next_pts = frame->pts + av_rescale_q(frame->nb_samples, (AVRational){1, inlink->sample_rate}, inlink->time_base);
+    return ff_filter_frame(ctx->outputs[0], out_frame);
+}
+
+static int request_frame(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    AudioEchoContext *s = ctx->priv;
+    int ret;
+
+    ret = ff_request_frame(ctx->inputs[0]);
+
+    if (ret == AVERROR_EOF && !ctx->is_disabled && s->fade_out) {
+        int nb_samples = FFMIN(s->fade_out, 2048);
+        AVFrame *frame;
+
+        frame = ff_get_audio_buffer(outlink, nb_samples);
+        if (!frame)
+            return AVERROR(ENOMEM);
+        s->fade_out -= nb_samples;
+
+        av_samples_set_silence(frame->extended_data, 0,
+                               frame->nb_samples,
+                               outlink->channels,
+                               frame->format);
+
+        s->echo_samples(s, s->delayptrs, frame->data, frame->data,
+                        frame->nb_samples, outlink->channels);
+
+        frame->pts = s->next_pts;
+        if (s->next_pts != AV_NOPTS_VALUE)
+            s->next_pts += av_rescale_q(nb_samples, (AVRational){1, outlink->sample_rate}, outlink->time_base);
+
+        return ff_filter_frame(outlink, frame);
+    }
+
+    return ret;
+}
+
+static const AVFilterPad aecho_inputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_AUDIO,
+        .filter_frame = filter_frame,
+    },
+    { NULL },
+};
+
+static const AVFilterPad aecho_outputs[] = {
+    {
+        .name          = "default",
+        .request_frame = request_frame,
+        .config_props  = config_output,
+        .type          = AVMEDIA_TYPE_AUDIO,
+    },
+    { NULL },
+};
+
+AVFilter avfilter_af_aecho = {
+    .name          = "aecho",
+    .description   = NULL_IF_CONFIG_SMALL("Add echoing to the audio."),
+    .query_formats = query_formats,
+    .priv_size     = sizeof(AudioEchoContext),
+    .priv_class    = &aecho_class,
+    .init          = init,
+    .uninit        = uninit,
+    .inputs        = aecho_inputs,
+    .outputs       = aecho_outputs,
+};
index 9a11feb..26472f8 100644 (file)
@@ -48,6 +48,7 @@ void avfilter_register_all(void)
 #if FF_API_ACONVERT_FILTER
     REGISTER_FILTER(ACONVERT,       aconvert,       af);
 #endif
+    REGISTER_FILTER(AECHO,          aecho,          af);
     REGISTER_FILTER(AFADE,          afade,          af);
     REGISTER_FILTER(AFORMAT,        aformat,        af);
     REGISTER_FILTER(AINTERLEAVE,    ainterleave,    af);
index c24e129..40034c9 100644 (file)
@@ -30,8 +30,8 @@
 #include "libavutil/avutil.h"
 
 #define LIBAVFILTER_VERSION_MAJOR  3
-#define LIBAVFILTER_VERSION_MINOR  79
-#define LIBAVFILTER_VERSION_MICRO 101
+#define LIBAVFILTER_VERSION_MINOR  80
+#define LIBAVFILTER_VERSION_MICRO 100
 
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \
                                                LIBAVFILTER_VERSION_MINOR, \