libhb: Allow control of audio mix normalization

Since switching to libavresample for audio mixing, our output
volume levels have been reduced because libavresample
does mix level normalization by default.  This change
applies a patch to libav to allow us to disable this behavior
and adds a new field to hb_audio_config_t to allow the
hb frontends to control this feature.

git-svn-id: svn://svn.handbrake.fr/HandBrake/trunk@4884 b64f7644-9d1e-0410-96f1-a4d463321fa5

diff --git a/contrib/ffmpeg/A06-mix-normalization.patch b/contrib/ffmpeg/A06-mix-normalization.patch
new file mode 100644 (file)
index 0000000..06fb275
--- /dev/null
+++ b/contrib/ffmpeg/A06-mix-normalization.patch
@@ -0,0 +1,39 @@
+diff --git a/libavresample/audio_mix.c b/libavresample/audio_mix.c
+index 2c2a356..25f9f98 100644
+--- a/libavresample/audio_mix.c
++++ b/libavresample/audio_mix.c
+@@ -327,7 +327,9 @@ int ff_audio_mix_init(AVAudioResampleContext *avr)
+                                       avr->out_channel_layout,
+                                       avr->center_mix_level,
+                                       avr->surround_mix_level,
+-                                      avr->lfe_mix_level, 1, matrix_dbl,
++                                      avr->lfe_mix_level,
++                                      avr->normalize_mix_level,
++                                      matrix_dbl,
+                                       avr->in_channels,
+                                       avr->matrix_encoding);
+         if (ret < 0) {
+diff --git a/libavresample/internal.h b/libavresample/internal.h
+index 7b7648f..006b6fd 100644
+--- a/libavresample/internal.h
++++ b/libavresample/internal.h
+@@ -45,6 +45,7 @@ struct AVAudioResampleContext {
+     double center_mix_level;                    /**< center mix level       */
+     double surround_mix_level;                  /**< surround mix level     */
+     double lfe_mix_level;                       /**< lfe mix level          */
++    int normalize_mix_level;                    /**< enable mix level normalization */
+     int force_resampling;                       /**< force resampling       */
+     int filter_size;                            /**< length of each FIR filter in the resampling filterbank relative to the cutoff frequency */
+     int phase_shift;                            /**< log2 of the number of entries in the resampling polyphase filterbank */
+diff --git a/libavresample/options.c b/libavresample/options.c
+index 02e1f86..e7e0c27 100644
+--- a/libavresample/options.c
++++ b/libavresample/options.c
+@@ -47,6 +47,7 @@ static const AVOption options[] = {
+     { "center_mix_level",       "Center Mix Level",         OFFSET(center_mix_level),       AV_OPT_TYPE_DOUBLE, { M_SQRT1_2             }, -32.0,                32.0,                   PARAM },
+     { "surround_mix_level",     "Surround Mix Level",       OFFSET(surround_mix_level),     AV_OPT_TYPE_DOUBLE, { M_SQRT1_2             }, -32.0,                32.0,                   PARAM },
+     { "lfe_mix_level",          "LFE Mix Level",            OFFSET(lfe_mix_level),          AV_OPT_TYPE_DOUBLE, { 0.0                   }, -32.0,                32.0,                   PARAM },
++    { "normalize_mix_level",    "Normalize Mix Level",      OFFSET(normalize_mix_level),    AV_OPT_TYPE_INT,    { 1                     }, 0,                    1,                      PARAM },
+     { "force_resampling",       "Force Resampling",         OFFSET(force_resampling),       AV_OPT_TYPE_INT,    { 0                     }, 0,                    1,                      PARAM },
+     { "filter_size",            "Resampling Filter Size",   OFFSET(filter_size),            AV_OPT_TYPE_INT,    { 16                    }, 0,                    32, /* ??? */           PARAM },
+     { "phase_shift",            "Resampling Phase Shift",   OFFSET(phase_shift),            AV_OPT_TYPE_INT,    { 10                    }, 0,                    30, /* ??? */           PARAM },

diff --git a/libhb/audio_resample.c b/libhb/audio_resample.c
index 3f67da87c476260addf601ca73b76bfe8b329ce1..ea09b38a236c16c94073d5b73d167f26e209cccb 100644 (file)
--- a/libhb/audio_resample.c
+++ b/libhb/audio_resample.c
@@ -13,20 +13,23 @@
 
 hb_audio_resample_t* hb_audio_resample_init(enum AVSampleFormat output_sample_fmt,
                                             uint64_t output_channel_layout,
-                                            enum AVMatrixEncoding matrix_encoding)
+                                            enum AVMatrixEncoding matrix_encoding,
+                                            int normalize_mix_level)
 {
     hb_audio_resample_t *resample = malloc(sizeof(hb_audio_resample_t));
     if (resample == NULL)
         return NULL;
 
-    resample->out.sample_fmt      = output_sample_fmt;
-    resample->out.sample_size     = av_get_bytes_per_sample(output_sample_fmt);
-    resample->out.channel_layout  = output_channel_layout;
-    resample->out.channels        =
+    resample->out.sample_fmt          = output_sample_fmt;
+    resample->out.sample_size         =
+        av_get_bytes_per_sample(output_sample_fmt);
+    resample->out.channel_layout      = output_channel_layout;
+    resample->out.channels            =
         av_get_channel_layout_nb_channels(output_channel_layout);
-    resample->out.matrix_encoding = matrix_encoding;
-    resample->resample_needed     = 0;
-    resample->avresample          = NULL;
+    resample->out.matrix_encoding     = matrix_encoding;
+    resample->out.normalize_mix_level = normalize_mix_level;
+    resample->resample_needed         = 0;
+    resample->avresample              = NULL;
 
     return resample;
 }
@@ -72,6 +75,8 @@ int hb_audio_resample_update(hb_audio_resample_t *resample,
                            resample->out.channel_layout, 0);
             av_opt_set_int(resample->avresample, "matrix_encoding",
                            resample->out.matrix_encoding, 0);
+            av_opt_set_int(resample->avresample, "normalize_mix_level",
+                           resample->out.normalize_mix_level, 0);
         }
         else if (resample_changed)
         {

diff --git a/libhb/audio_resample.h b/libhb/audio_resample.h
index 5982905e619412ae5d00827cbc3483dca1b3c474..4b1586e73b92f287f3fa4061a3f0742408589a24 100644 (file)
--- a/libhb/audio_resample.h
+++ b/libhb/audio_resample.h
@@ -39,6 +39,7 @@ typedef struct
         int channels;
         int linesize;
         int sample_size;
+        int normalize_mix_level;
         uint64_t channel_layout;
         enum AVSampleFormat sample_fmt;
         enum AVMatrixEncoding matrix_encoding;
@@ -52,7 +53,8 @@ typedef struct
  */
 hb_audio_resample_t* hb_audio_resample_init(enum AVSampleFormat output_sample_fmt,
                                             uint64_t output_channel_layout,
-                                            enum AVMatrixEncoding matrix_encoding);
+                                            enum AVMatrixEncoding matrix_encoding,
+                                            int normalize_mix_level);
 
 /* Update an hb_audio_resample_t, setting the input sample characteristics.
  *

diff --git a/libhb/common.c b/libhb/common.c
index 13a2e1d9536796cab0a7f554d3386f5f3ac98813..9ff9f44a3c2c78d156352dbcc18687f3d502b266 100644 (file)
--- a/libhb/common.c
+++ b/libhb/common.c
@@ -1637,6 +1637,7 @@ void hb_audio_config_init(hb_audio_config_t * audiocfg)
     audiocfg->out.mixdown = HB_INVALID_AMIXDOWN;
     audiocfg->out.dynamic_range_compression = 0;
     audiocfg->out.gain = 0;
+    audiocfg->out.normalize_mix_level = 0;
     audiocfg->out.name = NULL;
 }
 
@@ -1684,6 +1685,7 @@ int hb_audio_add(const hb_job_t * job, const hb_audio_config_t * audiocfg)
         audio->config.out.mixdown = HB_AMIXDOWN_NONE;
         audio->config.out.dynamic_range_compression = 0;
         audio->config.out.gain = 0;
+        audio->config.out.normalize_mix_level = 0;
         audio->config.out.compression_level = -1;
         audio->config.out.quality = HB_INVALID_AUDIO_QUALITY;
     }
@@ -1698,6 +1700,7 @@ int hb_audio_add(const hb_job_t * job, const hb_audio_config_t * audiocfg)
         audio->config.out.dynamic_range_compression = audiocfg->out.dynamic_range_compression;
         audio->config.out.mixdown = audiocfg->out.mixdown;
         audio->config.out.gain = audiocfg->out.gain;
+        audio->config.out.normalize_mix_level = audiocfg->out.normalize_mix_level;
     }
     if (audiocfg->out.name && *audiocfg->out.name)
     {

diff --git a/libhb/common.h b/libhb/common.h
index 5b2fc08fa43fee109d3161902fd4d44089e91533..1650f8002a96b9dd6f854b3f9609de22c7343964 100644 (file)
--- a/libhb/common.h
+++ b/libhb/common.h
@@ -452,6 +452,7 @@ struct hb_audio_config_s
         float    compression_level;  /* Output compression level (encoder-specific) */
         double   dynamic_range_compression; /* Amount of DRC applied to this track */
         double   gain; /* Gain (in dB), negative is quieter */
+        int      normalize_mix_level; /* mix level normalization (boolean) */
         char *   name; /* Output track name */
     } out;
 

diff --git a/libhb/decavcodec.c b/libhb/decavcodec.c
index 17f814d619217b72618aeffc3a534c7f8872f550..934f7611e453ae462162c9505d3c10522c8b3c43 100644 (file)
--- a/libhb/decavcodec.c
+++ b/libhb/decavcodec.c
@@ -192,7 +192,8 @@ static int decavcodecaInit( hb_work_object_t * w, hb_job_t * job )
         int mode;
         uint64_t layout = hb_ff_mixdown_xlat(w->audio->config.out.mixdown,
                                              &mode);
-        pv->resample = hb_audio_resample_init(AV_SAMPLE_FMT_FLT, layout, mode);
+        pv->resample = hb_audio_resample_init(AV_SAMPLE_FMT_FLT, layout, mode,
+                                              w->audio->config.out.normalize_mix_level);
         if (pv->resample == NULL)
         {
             hb_error("decavcodecaInit: hb_audio_resample_init() failed");

diff --git a/libhb/declpcm.c b/libhb/declpcm.c
index e48bf5db7ad188b5aabe26b7400e280eae9efbf6..0d4e449d11c7b63b46f654a41f55be8fdcc3c026 100644 (file)
--- a/libhb/declpcm.c
+++ b/libhb/declpcm.c
@@ -164,7 +164,8 @@ static int declpcmInit( hb_work_object_t * w, hb_job_t * job )
 
     int mode;
     uint64_t layout = hb_ff_mixdown_xlat(w->audio->config.out.mixdown, &mode);
-    pv->resample = hb_audio_resample_init(AV_SAMPLE_FMT_FLT, layout, mode);
+    pv->resample = hb_audio_resample_init(AV_SAMPLE_FMT_FLT, layout, mode,
+                                          w->audio->config.out.normalize_mix_level);
     if (pv->resample == NULL)
     {
         hb_error("declpcmInit: hb_audio_resample_init() failed");

diff --git a/libhb/encavcodecaudio.c b/libhb/encavcodecaudio.c
index 28eb0d9ae93429f452f77ecc141818ad3dad2ee1..ef5297ccab22e3e2be68f8eb9606ad4414acc271 100644 (file)
--- a/libhb/encavcodecaudio.c
+++ b/libhb/encavcodecaudio.c
@@ -119,7 +119,7 @@ static int encavcodecaInit(hb_work_object_t *w, hb_job_t *job)
     // sample_fmt conversion
     pv->resample = hb_audio_resample_init(context->sample_fmt,
                                           context->channel_layout,
-                                          AV_MATRIX_ENCODING_NONE);
+                                          AV_MATRIX_ENCODING_NONE, 0);
     if (hb_audio_resample_update(pv->resample, AV_SAMPLE_FMT_FLT,
                                  context->channel_layout, context->channels))
     {

diff --git a/test/test.c b/test/test.c
index 5d61b579c34c017bfb0e42f5052a065d7d0628eb..b59534d448b00189c62591e8f0b912a11e0d927b 100644 (file)
--- a/test/test.c
+++ b/test/test.c
@@ -71,6 +71,7 @@ static int    allowed_audio_copy = -1;
 static char * mixdowns    = NULL;
 static char * dynamic_range_compression = NULL;
 static char * audio_gain  = NULL;
+static char ** normalize_mix_level  = NULL;
 static char * atracks     = NULL;
 static char * arates      = NULL;
 static char ** abitrates  = NULL;
@@ -2096,6 +2097,41 @@ static int HandleEvents( hb_handle_t * h )
             }
             /* Audio Gain */
 
+            /* Audio Mix Normalization */
+            i = 0;
+            int norm = 0;
+            if( normalize_mix_level )
+            {
+                for ( i = 0; normalize_mix_level[i] != NULL && i < num_audio_tracks; i++ )
+                {
+                    char * token = normalize_mix_level[i];
+                    norm = atoi(token);
+                    audio = hb_list_audio_config_item(job->list_audio, i);
+
+                    if( audio != NULL )
+                    {
+                        audio->out.normalize_mix_level = norm;
+                    }
+                    else
+                    {
+                        fprintf(stderr, "Ignoring normalization %d, no audio tracks\n", norm);
+                    }
+                }
+            }
+            if (i < num_audio_tracks && i == 1)
+            {
+                /* We have fewer inputs than audio tracks,
+                 * and we only have one input, use
+                 * that for all tracks.
+                 */
+                for (; i < num_audio_tracks; i++)
+                {
+                    audio = hb_list_audio_config_item(job->list_audio, i);
+                    audio->out.normalize_mix_level = norm;
+                }
+            }
+            /* Audio Mix Normalization */
+
             /* Audio Track Names */
             if ( anames )
             {
@@ -2883,6 +2919,10 @@ static void ShowHelp()
     "                            Separated by commas for more than one audio track.\n"
     "                            (mono/stereo/dpl1/dpl2/6ch, default: up to 6ch for ac3,\n"
     "                            up to dpl2 for other encoders)\n"
+    "        --normalize-mix     Normalize audio mix levels to prevent clipping.\n"
+    "               <string>     Separated by commas for more than one audio track.\n"
+    "                            0 = Disable Normalization (default)\n"
+    "                            1 = Enable Normalization\n"
     "    -R, --arate             Set audio samplerate(s) (" );
     for( i = 0; i < hb_audio_rates_count; i++ )
     {
@@ -3186,6 +3226,7 @@ static int ParseOptions( int argc, char ** argv )
     #define X264_PRESET         284
     #define X264_TUNE           285
     #define H264_LEVEL          286
+    #define NORMALIZE_MIX       287
     
     for( ;; )
     {
@@ -3212,6 +3253,7 @@ static int ParseOptions( int argc, char ** argv )
             { "markers",     optional_argument, NULL,    'm' },
             { "audio",       required_argument, NULL,    'a' },
             { "mixdown",     required_argument, NULL,    '6' },
+            { "normalize-mix", required_argument, NULL,  NORMALIZE_MIX },
             { "drc",         required_argument, NULL,    'D' },
             { "gain",        required_argument, NULL,    AUDIO_GAIN },
             { "subtitle",    required_argument, NULL,    's' },
@@ -3430,6 +3472,12 @@ static int ParseOptions( int argc, char ** argv )
                     audio_gain = strdup( optarg );
                 }
                 break;
+            case NORMALIZE_MIX:
+                if( optarg != NULL )
+                {
+                    normalize_mix_level = str_split( optarg, ',' );
+                }
+                break;
             case 's':
                 subtracks = str_split( optarg, ',' );
                 break;