lavfi: use common VAAPI VPP infrastructure for vf_scale_vaapi.
[ffmpeg.git] / libavfilter / af_silencedetect.c
index b19ac2e..b048d63 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012 Clément Bœsch <ubitux@gmail.com>
+ * Copyright (c) 2012 Clément Bœsch <u pkh me>
  *
  * This file is part of FFmpeg.
  *
@@ -23,7 +23,8 @@
  * Audio silence detector
  */
 
-#include "libavutil/audioconvert.h"
+#include <float.h> /* DBL_MAX */
+
 #include "libavutil/opt.h"
 #include "libavutil/timestamp.h"
 #include "audio.h"
 #include "avfilter.h"
 #include "internal.h"
 
-typedef struct {
+typedef struct SilenceDetectContext {
     const AVClass *class;
-    char *noise_str;            ///< noise option string
     double noise;               ///< noise amplitude ratio
     double duration;            ///< minimum duration of silence until notification
     int64_t nb_null_samples;    ///< current number of continuous zero samples
     int64_t start;              ///< if silence is detected, this value contains the time of the first zero sample
     int last_sample_rate;       ///< last sample rate to check for sample rate changes
+
+    void (*silencedetect)(struct SilenceDetectContext *s, AVFrame *insamples,
+                          int nb_samples, int64_t nb_samples_notify,
+                          AVRational time_base);
 } SilenceDetectContext;
 
 #define OFFSET(x) offsetof(SilenceDetectContext, x)
 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_AUDIO_PARAM
 static const AVOption silencedetect_options[] = {
-    { "n",         "set noise tolerance",              OFFSET(noise_str), AV_OPT_TYPE_STRING, {.str="-60dB"}, CHAR_MIN, CHAR_MAX, FLAGS },
-    { "noise",     "set noise tolerance",              OFFSET(noise_str), AV_OPT_TYPE_STRING, {.str="-60dB"}, CHAR_MIN, CHAR_MAX, FLAGS },
+    { "n",         "set noise tolerance",              OFFSET(noise),     AV_OPT_TYPE_DOUBLE, {.dbl=0.001},          0, DBL_MAX,  FLAGS },
+    { "noise",     "set noise tolerance",              OFFSET(noise),     AV_OPT_TYPE_DOUBLE, {.dbl=0.001},          0, DBL_MAX,  FLAGS },
     { "d",         "set minimum duration in seconds",  OFFSET(duration),  AV_OPT_TYPE_DOUBLE, {.dbl=2.},             0, 24*60*60, FLAGS },
     { "duration",  "set minimum duration in seconds",  OFFSET(duration),  AV_OPT_TYPE_DOUBLE, {.dbl=2.},             0, 24*60*60, FLAGS },
-    { NULL },
+    { NULL }
 };
 
 AVFILTER_DEFINE_CLASS(silencedetect);
 
-static av_cold int init(AVFilterContext *ctx, const char *args)
+static char *get_metadata_val(AVFrame *insamples, const char *key)
 {
-    int ret;
-    char *tail;
-    SilenceDetectContext *silence = ctx->priv;
+    AVDictionaryEntry *e = av_dict_get(insamples->metadata, key, NULL, 0);
+    return e && e->value ? e->value : NULL;
+}
+
+static av_always_inline void update(SilenceDetectContext *s, AVFrame *insamples,
+                                    int is_silence, int64_t nb_samples_notify,
+                                    AVRational time_base)
+{
+    if (is_silence) {
+        if (!s->start) {
+            s->nb_null_samples++;
+            if (s->nb_null_samples >= nb_samples_notify) {
+                s->start = insamples->pts - (int64_t)(s->duration / av_q2d(time_base) + .5);
+                av_dict_set(&insamples->metadata, "lavfi.silence_start",
+                            av_ts2timestr(s->start, &time_base), 0);
+                av_log(s, AV_LOG_INFO, "silence_start: %s\n",
+                       get_metadata_val(insamples, "lavfi.silence_start"));
+            }
+        }
+    } else {
+        if (s->start) {
+            av_dict_set(&insamples->metadata, "lavfi.silence_end",
+                        av_ts2timestr(insamples->pts, &time_base), 0);
+            av_dict_set(&insamples->metadata, "lavfi.silence_duration",
+                        av_ts2timestr(insamples->pts - s->start, &time_base), 0);
+            av_log(s, AV_LOG_INFO,
+                   "silence_end: %s | silence_duration: %s\n",
+                   get_metadata_val(insamples, "lavfi.silence_end"),
+                   get_metadata_val(insamples, "lavfi.silence_duration"));
+        }
+        s->nb_null_samples = s->start = 0;
+    }
+}
 
-    silence->class = &silencedetect_class;
-    av_opt_set_defaults(silence);
+#define SILENCE_DETECT(name, type)                                               \
+static void silencedetect_##name(SilenceDetectContext *s, AVFrame *insamples,    \
+                                 int nb_samples, int64_t nb_samples_notify,      \
+                                 AVRational time_base)                           \
+{                                                                                \
+    const type *p = (const type *)insamples->data[0];                            \
+    const type noise = s->noise;                                                 \
+    int i;                                                                       \
+                                                                                 \
+    for (i = 0; i < nb_samples; i++, p++)                                        \
+        update(s, insamples, *p < noise && *p > -noise,                          \
+               nb_samples_notify, time_base);                                    \
+}
 
-    if ((ret = av_set_options_string(silence, args, "=", ":")) < 0)
-        return ret;
+SILENCE_DETECT(dbl, double)
+SILENCE_DETECT(flt, float)
+SILENCE_DETECT(s32, int32_t)
+SILENCE_DETECT(s16, int16_t)
 
-    silence->noise = strtod(silence->noise_str, &tail);
-    if (!strcmp(tail, "dB")) {
-        silence->noise = pow(10, silence->noise/20);
-    } else if (*tail) {
-        av_log(ctx, AV_LOG_ERROR, "Invalid value '%s' for noise parameter.\n",
-               silence->noise_str);
-        return AVERROR(EINVAL);
+static int config_input(AVFilterLink *inlink)
+{
+    AVFilterContext *ctx = inlink->dst;
+    SilenceDetectContext *s = ctx->priv;
+
+    switch (inlink->format) {
+    case AV_SAMPLE_FMT_DBL: s->silencedetect = silencedetect_dbl; break;
+    case AV_SAMPLE_FMT_FLT: s->silencedetect = silencedetect_flt; break;
+    case AV_SAMPLE_FMT_S32:
+        s->noise *= INT32_MAX;
+        s->silencedetect = silencedetect_s32;
+        break;
+    case AV_SAMPLE_FMT_S16:
+        s->noise *= INT16_MAX;
+        s->silencedetect = silencedetect_s16;
+        break;
     }
-    av_opt_free(silence);
 
     return 0;
 }
 
-static int filter_samples(AVFilterLink *inlink, AVFilterBufferRef *insamples)
+static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
 {
-    int i;
-    SilenceDetectContext *silence = inlink->dst->priv;
-    const int nb_channels           = av_get_channel_layout_nb_channels(inlink->channel_layout);
+    SilenceDetectContext *s         = inlink->dst->priv;
+    const int nb_channels           = inlink->channels;
     const int srate                 = inlink->sample_rate;
-    const int nb_samples            = insamples->audio->nb_samples * nb_channels;
-    const int64_t nb_samples_notify = srate * silence->duration    * nb_channels;
+    const int nb_samples            = insamples->nb_samples     * nb_channels;
+    const int64_t nb_samples_notify = srate * s->duration * nb_channels;
 
     // scale number of null samples to the new sample rate
-    if (silence->last_sample_rate && silence->last_sample_rate != srate)
-        silence->nb_null_samples =
-            srate * silence->nb_null_samples / silence->last_sample_rate;
-    silence->last_sample_rate = srate;
-
-    // TODO: support more sample formats
-    if (insamples->format == AV_SAMPLE_FMT_DBL) {
-        double *p = (double *)insamples->data[0];
-
-        for (i = 0; i < nb_samples; i++, p++) {
-            if (*p < silence->noise && *p > -silence->noise) {
-                if (!silence->start) {
-                    silence->nb_null_samples++;
-                    if (silence->nb_null_samples >= nb_samples_notify) {
-                        silence->start = insamples->pts - (int64_t)(silence->duration / av_q2d(inlink->time_base) + .5);
-                        av_log(silence, AV_LOG_INFO,
-                               "silence_start: %s\n", av_ts2timestr(silence->start, &inlink->time_base));
-                    }
-                }
-            } else {
-                if (silence->start)
-                    av_log(silence, AV_LOG_INFO,
-                           "silence_end: %s | silence_duration: %s\n",
-                           av_ts2timestr(insamples->pts,                  &inlink->time_base),
-                           av_ts2timestr(insamples->pts - silence->start, &inlink->time_base));
-                silence->nb_null_samples = silence->start = 0;
-            }
-        }
-    }
+    if (s->last_sample_rate && s->last_sample_rate != srate)
+        s->nb_null_samples = srate * s->nb_null_samples / s->last_sample_rate;
+    s->last_sample_rate = srate;
+
+    // TODO: document metadata
+    s->silencedetect(s, insamples, nb_samples, nb_samples_notify,
+                     inlink->time_base);
 
-    return ff_filter_samples(inlink->dst->outputs[0], insamples);
+    return ff_filter_frame(inlink->dst->outputs[0], insamples);
 }
 
 static int query_formats(AVFilterContext *ctx)
 {
     AVFilterFormats *formats = NULL;
     AVFilterChannelLayouts *layouts = NULL;
-    enum AVSampleFormat sample_fmts[] = {
+    static const enum AVSampleFormat sample_fmts[] = {
         AV_SAMPLE_FMT_DBL,
+        AV_SAMPLE_FMT_FLT,
+        AV_SAMPLE_FMT_S32,
+        AV_SAMPLE_FMT_S16,
         AV_SAMPLE_FMT_NONE
     };
+    int ret;
 
     layouts = ff_all_channel_layouts();
     if (!layouts)
         return AVERROR(ENOMEM);
-    ff_set_common_channel_layouts(ctx, layouts);
+    ret = ff_set_common_channel_layouts(ctx, layouts);
+    if (ret < 0)
+        return ret;
 
     formats = ff_make_format_list(sample_fmts);
     if (!formats)
         return AVERROR(ENOMEM);
-    ff_set_common_formats(ctx, formats);
+    ret = ff_set_common_formats(ctx, formats);
+    if (ret < 0)
+        return ret;
 
     formats = ff_all_samplerates();
     if (!formats)
         return AVERROR(ENOMEM);
-    ff_set_common_samplerates(ctx, formats);
-
-    return 0;
+    return ff_set_common_samplerates(ctx, formats);
 }
 
-AVFilter avfilter_af_silencedetect = {
+static const AVFilterPad silencedetect_inputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_AUDIO,
+        .config_props = config_input,
+        .filter_frame = filter_frame,
+    },
+    { NULL }
+};
+
+static const AVFilterPad silencedetect_outputs[] = {
+    {
+        .name = "default",
+        .type = AVMEDIA_TYPE_AUDIO,
+    },
+    { NULL }
+};
+
+AVFilter ff_af_silencedetect = {
     .name          = "silencedetect",
     .description   = NULL_IF_CONFIG_SMALL("Detect silence."),
     .priv_size     = sizeof(SilenceDetectContext),
-    .init          = init,
     .query_formats = query_formats,
-
-    .inputs = (const AVFilterPad[]) {
-        { .name             = "default",
-          .type             = AVMEDIA_TYPE_AUDIO,
-          .get_audio_buffer = ff_null_get_audio_buffer,
-          .filter_samples   = filter_samples, },
-        { .name = NULL }
-    },
-    .outputs = (const AVFilterPad[]) {
-        { .name = "default",
-          .type = AVMEDIA_TYPE_AUDIO, },
-        { .name = NULL }
-    },
-    .priv_class = &silencedetect_class,
+    .inputs        = silencedetect_inputs,
+    .outputs       = silencedetect_outputs,
+    .priv_class    = &silencedetect_class,
 };