yading@10: /* yading@10: * Copyright (c) 2012 Clément Bœsch yading@10: * yading@10: * This file is part of FFmpeg. yading@10: * yading@10: * FFmpeg is free software; you can redistribute it and/or yading@10: * modify it under the terms of the GNU Lesser General Public yading@10: * License as published by the Free Software Foundation; either yading@10: * version 2.1 of the License, or (at your option) any later version. yading@10: * yading@10: * FFmpeg is distributed in the hope that it will be useful, yading@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of yading@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU yading@10: * Lesser General Public License for more details. yading@10: * yading@10: * You should have received a copy of the GNU Lesser General Public yading@10: * License along with FFmpeg; if not, write to the Free Software yading@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA yading@10: */ yading@10: yading@10: /** yading@10: * @file yading@10: * Audio silence detector yading@10: */ yading@10: yading@10: #include /* DBL_MAX */ yading@10: yading@10: #include "libavutil/channel_layout.h" yading@10: #include "libavutil/opt.h" yading@10: #include "libavutil/timestamp.h" yading@10: #include "audio.h" yading@10: #include "formats.h" yading@10: #include "avfilter.h" yading@10: #include "internal.h" yading@10: yading@10: typedef struct { yading@10: const AVClass *class; yading@10: double noise; ///< noise amplitude ratio yading@10: double duration; ///< minimum duration of silence until notification yading@10: int64_t nb_null_samples; ///< current number of continuous zero samples yading@10: int64_t start; ///< if silence is detected, this value contains the time of the first zero sample yading@10: int last_sample_rate; ///< last sample rate to check for sample rate changes yading@10: } SilenceDetectContext; yading@10: yading@10: #define OFFSET(x) offsetof(SilenceDetectContext, x) yading@10: #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_AUDIO_PARAM yading@10: static const AVOption silencedetect_options[] = { yading@10: { "n", "set noise tolerance", OFFSET(noise), AV_OPT_TYPE_DOUBLE, {.dbl=0.001}, 0, DBL_MAX, FLAGS }, yading@10: { "noise", "set noise tolerance", OFFSET(noise), AV_OPT_TYPE_DOUBLE, {.dbl=0.001}, 0, DBL_MAX, FLAGS }, yading@10: { "d", "set minimum duration in seconds", OFFSET(duration), AV_OPT_TYPE_DOUBLE, {.dbl=2.}, 0, 24*60*60, FLAGS }, yading@10: { "duration", "set minimum duration in seconds", OFFSET(duration), AV_OPT_TYPE_DOUBLE, {.dbl=2.}, 0, 24*60*60, FLAGS }, yading@10: { NULL }, yading@10: }; yading@10: yading@10: AVFILTER_DEFINE_CLASS(silencedetect); yading@10: yading@10: static char *get_metadata_val(AVFrame *insamples, const char *key) yading@10: { yading@10: AVDictionaryEntry *e = av_dict_get(insamples->metadata, key, NULL, 0); yading@10: return e && e->value ? e->value : NULL; yading@10: } yading@10: yading@10: static int filter_frame(AVFilterLink *inlink, AVFrame *insamples) yading@10: { yading@10: int i; yading@10: SilenceDetectContext *silence = inlink->dst->priv; yading@10: const int nb_channels = av_get_channel_layout_nb_channels(inlink->channel_layout); yading@10: const int srate = inlink->sample_rate; yading@10: const int nb_samples = insamples->nb_samples * nb_channels; yading@10: const int64_t nb_samples_notify = srate * silence->duration * nb_channels; yading@10: yading@10: // scale number of null samples to the new sample rate yading@10: if (silence->last_sample_rate && silence->last_sample_rate != srate) yading@10: silence->nb_null_samples = yading@10: srate * silence->nb_null_samples / silence->last_sample_rate; yading@10: silence->last_sample_rate = srate; yading@10: yading@10: // TODO: support more sample formats yading@10: // TODO: document metadata yading@10: if (insamples->format == AV_SAMPLE_FMT_DBL) { yading@10: double *p = (double *)insamples->data[0]; yading@10: yading@10: for (i = 0; i < nb_samples; i++, p++) { yading@10: if (*p < silence->noise && *p > -silence->noise) { yading@10: if (!silence->start) { yading@10: silence->nb_null_samples++; yading@10: if (silence->nb_null_samples >= nb_samples_notify) { yading@10: silence->start = insamples->pts - (int64_t)(silence->duration / av_q2d(inlink->time_base) + .5); yading@10: av_dict_set(&insamples->metadata, "lavfi.silence_start", yading@10: av_ts2timestr(silence->start, &inlink->time_base), 0); yading@10: av_log(silence, AV_LOG_INFO, "silence_start: %s\n", yading@10: get_metadata_val(insamples, "lavfi.silence_start")); yading@10: } yading@10: } yading@10: } else { yading@10: if (silence->start) { yading@10: av_dict_set(&insamples->metadata, "lavfi.silence_end", yading@10: av_ts2timestr(insamples->pts, &inlink->time_base), 0); yading@10: av_dict_set(&insamples->metadata, "lavfi.silence_duration", yading@10: av_ts2timestr(insamples->pts - silence->start, &inlink->time_base), 0); yading@10: av_log(silence, AV_LOG_INFO, yading@10: "silence_end: %s | silence_duration: %s\n", yading@10: get_metadata_val(insamples, "lavfi.silence_end"), yading@10: get_metadata_val(insamples, "lavfi.silence_duration")); yading@10: } yading@10: silence->nb_null_samples = silence->start = 0; yading@10: } yading@10: } yading@10: } yading@10: yading@10: return ff_filter_frame(inlink->dst->outputs[0], insamples); yading@10: } yading@10: yading@10: static int query_formats(AVFilterContext *ctx) yading@10: { yading@10: AVFilterFormats *formats = NULL; yading@10: AVFilterChannelLayouts *layouts = NULL; yading@10: static const enum AVSampleFormat sample_fmts[] = { yading@10: AV_SAMPLE_FMT_DBL, yading@10: AV_SAMPLE_FMT_NONE yading@10: }; yading@10: yading@10: layouts = ff_all_channel_layouts(); yading@10: if (!layouts) yading@10: return AVERROR(ENOMEM); yading@10: ff_set_common_channel_layouts(ctx, layouts); yading@10: yading@10: formats = ff_make_format_list(sample_fmts); yading@10: if (!formats) yading@10: return AVERROR(ENOMEM); yading@10: ff_set_common_formats(ctx, formats); yading@10: yading@10: formats = ff_all_samplerates(); yading@10: if (!formats) yading@10: return AVERROR(ENOMEM); yading@10: ff_set_common_samplerates(ctx, formats); yading@10: yading@10: return 0; yading@10: } yading@10: yading@10: static const AVFilterPad silencedetect_inputs[] = { yading@10: { yading@10: .name = "default", yading@10: .type = AVMEDIA_TYPE_AUDIO, yading@10: .get_audio_buffer = ff_null_get_audio_buffer, yading@10: .filter_frame = filter_frame, yading@10: }, yading@10: { NULL } yading@10: }; yading@10: yading@10: static const AVFilterPad silencedetect_outputs[] = { yading@10: { yading@10: .name = "default", yading@10: .type = AVMEDIA_TYPE_AUDIO, yading@10: }, yading@10: { NULL } yading@10: }; yading@10: yading@10: AVFilter avfilter_af_silencedetect = { yading@10: .name = "silencedetect", yading@10: .description = NULL_IF_CONFIG_SMALL("Detect silence."), yading@10: .priv_size = sizeof(SilenceDetectContext), yading@10: .query_formats = query_formats, yading@10: .inputs = silencedetect_inputs, yading@10: .outputs = silencedetect_outputs, yading@10: .priv_class = &silencedetect_class, yading@10: };