2 * Copyright (c) 2012 Clément Bœsch <ubitux@gmail.com>
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 silence detector
26 #include <float.h> /* DBL_MAX */
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/opt.h"
30 #include "libavutil/timestamp.h"
38 double noise; ///< noise amplitude ratio
39 double duration; ///< minimum duration of silence until notification
40 int64_t nb_null_samples; ///< current number of continuous zero samples
41 int64_t start; ///< if silence is detected, this value contains the time of the first zero sample
42 int last_sample_rate; ///< last sample rate to check for sample rate changes
43 } SilenceDetectContext;
45 #define OFFSET(x) offsetof(SilenceDetectContext, x)
46 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_AUDIO_PARAM
47 static const AVOption silencedetect_options[] = {
48 { "n", "set noise tolerance", OFFSET(noise), AV_OPT_TYPE_DOUBLE, {.dbl=0.001}, 0, DBL_MAX, FLAGS },
49 { "noise", "set noise tolerance", OFFSET(noise), AV_OPT_TYPE_DOUBLE, {.dbl=0.001}, 0, DBL_MAX, FLAGS },
50 { "d", "set minimum duration in seconds", OFFSET(duration), AV_OPT_TYPE_DOUBLE, {.dbl=2.}, 0, 24*60*60, FLAGS },
51 { "duration", "set minimum duration in seconds", OFFSET(duration), AV_OPT_TYPE_DOUBLE, {.dbl=2.}, 0, 24*60*60, FLAGS },
55 AVFILTER_DEFINE_CLASS(silencedetect);
57 static av_cold int init(AVFilterContext *ctx, const char *args)
60 SilenceDetectContext *silence = ctx->priv;
62 silence->class = &silencedetect_class;
63 av_opt_set_defaults(silence);
65 if ((ret = av_set_options_string(silence, args, "=", ":")) < 0)
73 static char *get_metadata_val(AVFrame *insamples, const char *key)
75 AVDictionaryEntry *e = av_dict_get(insamples->metadata, key, NULL, 0);
76 return e && e->value ? e->value : NULL;
79 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
82 SilenceDetectContext *silence = inlink->dst->priv;
83 const int nb_channels = av_get_channel_layout_nb_channels(inlink->channel_layout);
84 const int srate = inlink->sample_rate;
85 const int nb_samples = insamples->nb_samples * nb_channels;
86 const int64_t nb_samples_notify = srate * silence->duration * nb_channels;
88 // scale number of null samples to the new sample rate
89 if (silence->last_sample_rate && silence->last_sample_rate != srate)
90 silence->nb_null_samples =
91 srate * silence->nb_null_samples / silence->last_sample_rate;
92 silence->last_sample_rate = srate;
94 // TODO: support more sample formats
95 // TODO: document metadata
96 if (insamples->format == AV_SAMPLE_FMT_DBL) {
97 double *p = (double *)insamples->data[0];
99 for (i = 0; i < nb_samples; i++, p++) {
100 if (*p < silence->noise && *p > -silence->noise) {
101 if (!silence->start) {
102 silence->nb_null_samples++;
103 if (silence->nb_null_samples >= nb_samples_notify) {
104 silence->start = insamples->pts - (int64_t)(silence->duration / av_q2d(inlink->time_base) + .5);
105 av_dict_set(&insamples->metadata, "lavfi.silence_start",
106 av_ts2timestr(silence->start, &inlink->time_base), 0);
107 av_log(silence, AV_LOG_INFO, "silence_start: %s\n",
108 get_metadata_val(insamples, "lavfi.silence_start"));
112 if (silence->start) {
113 av_dict_set(&insamples->metadata, "lavfi.silence_end",
114 av_ts2timestr(insamples->pts, &inlink->time_base), 0);
115 av_dict_set(&insamples->metadata, "lavfi.silence_duration",
116 av_ts2timestr(insamples->pts - silence->start, &inlink->time_base), 0);
117 av_log(silence, AV_LOG_INFO,
118 "silence_end: %s | silence_duration: %s\n",
119 get_metadata_val(insamples, "lavfi.silence_end"),
120 get_metadata_val(insamples, "lavfi.silence_duration"));
122 silence->nb_null_samples = silence->start = 0;
127 return ff_filter_frame(inlink->dst->outputs[0], insamples);
130 static int query_formats(AVFilterContext *ctx)
132 AVFilterFormats *formats = NULL;
133 AVFilterChannelLayouts *layouts = NULL;
134 static const enum AVSampleFormat sample_fmts[] = {
139 layouts = ff_all_channel_layouts();
141 return AVERROR(ENOMEM);
142 ff_set_common_channel_layouts(ctx, layouts);
144 formats = ff_make_format_list(sample_fmts);
146 return AVERROR(ENOMEM);
147 ff_set_common_formats(ctx, formats);
149 formats = ff_all_samplerates();
151 return AVERROR(ENOMEM);
152 ff_set_common_samplerates(ctx, formats);
157 static const AVFilterPad silencedetect_inputs[] = {
160 .type = AVMEDIA_TYPE_AUDIO,
161 .get_audio_buffer = ff_null_get_audio_buffer,
162 .filter_frame = filter_frame,
167 static const AVFilterPad silencedetect_outputs[] = {
170 .type = AVMEDIA_TYPE_AUDIO,
175 AVFilter avfilter_af_silencedetect = {
176 .name = "silencedetect",
177 .description = NULL_IF_CONFIG_SMALL("Detect silence."),
178 .priv_size = sizeof(SilenceDetectContext),
180 .query_formats = query_formats,
181 .inputs = silencedetect_inputs,
182 .outputs = silencedetect_outputs,
183 .priv_class = &silencedetect_class,