]> git.sesse.net Git - vlc/blobdiff - modules/codec/avcodec/encoder.c
Simplify FLAC extradata (streaminfo) parsing
[vlc] / modules / codec / avcodec / encoder.c
index 56e64539c22c1befdbd4aee96a62b800a8100941..9717d099b8c3b14c2a4165b425b61b0caacfad53 100644 (file)
 #include "avcodec.h"
 #include "avcommon.h"
 
+#if LIBAVUTIL_VERSION_CHECK( 52,2,6,0,0 )
+# include <libavutil/channel_layout.h>
+#endif
+
 #define HURRY_UP_GUARD1 (450000)
 #define HURRY_UP_GUARD2 (300000)
 #define HURRY_UP_GUARD3 (100000)
@@ -98,6 +102,7 @@ struct encoder_sys_t
      */
     uint8_t *p_buffer;
     size_t i_buffer_out;
+    uint8_t *p_interleave_buf;
 
     /*
      * Video properties
@@ -110,10 +115,13 @@ struct encoder_sys_t
     /*
      * Audio properties
      */
-    int i_sample_bytes;
-    int i_frame_size;
-    int i_samples_delay;
+    size_t i_sample_bytes;
+    size_t i_frame_size;
+    size_t i_samples_delay; //How much samples in delay buffer
+    bool b_planar;
+    bool b_variable;    //Encoder can be fed with any size frames not just frame_size
     mtime_t i_pts;
+    date_t  buffer_date;
 
     /* Encoding settings */
     int        i_key_int;
@@ -150,7 +158,7 @@ static const char *const ppsz_enc_options[] = {
 #if (LIBAVCODEC_VERSION_MAJOR < 55)
     "luma-elim-threshold", "chroma-elim-threshold",
 #endif
-    "aac-profile",
+    "aac-profile", "options",
     NULL
 };
 
@@ -300,13 +308,16 @@ int OpenEncoder( vlc_object_t *p_this )
     p_enc->p_sys = p_sys;
     p_sys->i_samples_delay = 0;
     p_sys->p_codec = p_codec;
+    p_sys->b_planar = false;
 
     p_sys->p_buffer = NULL;
+    p_sys->p_interleave_buf = NULL;
     p_sys->i_buffer_out = 0;
 
     p_context = avcodec_alloc_context3(p_codec);
     p_sys->p_context = p_context;
     p_sys->p_context->codec_id = p_sys->p_codec->id;
+    p_context->thread_type = 0;
     p_context->debug = var_InheritInteger( p_enc, "avcodec-debug" );
     p_context->opaque = (void *)p_this;
 
@@ -338,8 +349,12 @@ int OpenEncoder( vlc_object_t *p_this )
     p_sys->b_mpeg4_matrix = var_GetBool( p_enc, ENC_CFG_PREFIX "mpeg4-matrix" );
 
     f_val = var_GetFloat( p_enc, ENC_CFG_PREFIX "qscale" );
-    if( f_val < 0.01 || f_val > 255.0 ) f_val = 0;
-    p_sys->i_quality = (int)(FF_QP2LAMBDA * f_val + 0.5);
+
+    p_sys->i_quality = 0;
+    if( f_val < 0.01 || f_val > 255.0 )
+        f_val = 0;
+    else
+        p_sys->i_quality = (int)(FF_QP2LAMBDA * f_val + 0.5);
 
     psz_val = var_GetString( p_enc, ENC_CFG_PREFIX "hq" );
     p_sys->i_hq = FF_MB_DECISION_RD;
@@ -598,27 +613,50 @@ int OpenEncoder( vlc_object_t *p_this )
                                     p_codec->sample_fmts[0] :
                                     AV_SAMPLE_FMT_S16;
 
+        /* Try to match avcodec input format to vlc format so we could avoid one
+           format conversion */
+        if( GetVlcAudioFormat( p_context->sample_fmt ) != p_enc->fmt_in.i_codec )
+        {
+            msg_Dbg( p_enc, "Trying to find more suitable sample format instead of %s", av_get_sample_fmt_name( p_context->sample_fmt ) );
+            for( unsigned int i=0; p_codec->sample_fmts[i] != -1; i++ )
+            {
+                if( GetVlcAudioFormat( p_codec->sample_fmts[i] ) == p_enc->fmt_in.i_codec )
+                {
+                    p_context->sample_fmt = p_codec->sample_fmts[i];
+                    msg_Dbg( p_enc, "Using %s as new sample format", av_get_sample_fmt_name( p_context->sample_fmt ) );
+                    break;
+                }
+            }
+        }
+        p_sys->b_planar = av_sample_fmt_is_planar( p_context->sample_fmt );
         // Try if we can use interleaved format for codec input as VLC doesn't really do planar audio yet
         // FIXME: Remove when planar/interleaved audio in vlc is equally supported
-        if( av_sample_fmt_is_planar( p_context->sample_fmt ) )
+        if( p_sys->b_planar )
         {
-            msg_Dbg( p_enc, "Trying to find packet sample format instead of %s", av_get_sample_fmt_name( p_context->sample_fmt ) );
+            msg_Dbg( p_enc, "Trying to find packet sample format instead of planar %s", av_get_sample_fmt_name( p_context->sample_fmt ) );
             for( unsigned int i=0; p_codec->sample_fmts[i] != -1; i++ )
             {
                 if( !av_sample_fmt_is_planar( p_codec->sample_fmts[i] ) )
                 {
                     p_context->sample_fmt = p_codec->sample_fmts[i];
-                    msg_Dbg( p_enc, "Using %s as new sample format", av_get_sample_fmt_name( p_context->sample_fmt ) );
+                    msg_Dbg( p_enc, "Changing to packet format %s as new sample format", av_get_sample_fmt_name( p_context->sample_fmt ) );
                     break;
                 }
             }
         }
+        msg_Dbg( p_enc, "Ended up using %s as sample format", av_get_sample_fmt_name( p_context->sample_fmt ) );
         p_enc->fmt_in.i_codec  = GetVlcAudioFormat( p_context->sample_fmt );
+        p_sys->b_planar = av_sample_fmt_is_planar( p_context->sample_fmt );
 
         p_context->sample_rate = p_enc->fmt_out.audio.i_rate;
+        date_Init( &p_sys->buffer_date, p_enc->fmt_out.audio.i_rate, 1 );
+        date_Set( &p_sys->buffer_date, 0 );
         p_context->time_base.num = 1;
         p_context->time_base.den = p_context->sample_rate;
         p_context->channels    = p_enc->fmt_out.audio.i_channels;
+#if LIBAVUTIL_VERSION_CHECK( 52, 2, 6, 0, 0)
+        p_context->channel_layout = av_get_default_channel_layout( p_context->channels );
+#endif
 
         if ( p_enc->fmt_out.i_codec == VLC_CODEC_MP4A )
         {
@@ -700,118 +738,100 @@ int OpenEncoder( vlc_object_t *p_this )
         p_context->thread_count = vlc_GetCPUCount();
 
     int ret;
+    char *psz_opts = var_InheritString(p_enc, ENC_CFG_PREFIX "options");
+    AVDictionary *options = NULL;
+    if (psz_opts && *psz_opts)
+        options = vlc_av_get_options(psz_opts);
+    free(psz_opts);
+
     vlc_avcodec_lock();
-    ret = avcodec_open2( p_context, p_codec, NULL /* options */ );
+    ret = avcodec_open2( p_context, p_codec, options ? &options : NULL );
     vlc_avcodec_unlock();
+
+    AVDictionaryEntry *t = NULL;
+    while ((t = av_dict_get(options, "", t, AV_DICT_IGNORE_SUFFIX))) {
+        msg_Err(p_enc, "Unknown option \"%s\"", t->key);
+    }
+
     if( ret )
     {
-        if( p_enc->fmt_in.i_cat == AUDIO_ES &&
-             (p_context->channels > 2 || i_codec_id == AV_CODEC_ID_MP2
-               || i_codec_id == AV_CODEC_ID_MP3) )
+        if( p_enc->fmt_in.i_cat != AUDIO_ES ||
+                (p_context->channels <= 2 && i_codec_id != AV_CODEC_ID_MP2
+                 && i_codec_id != AV_CODEC_ID_MP3) )
         {
-            if( p_context->channels > 2 )
-            {
-                p_context->channels = 2;
-                p_enc->fmt_in.audio.i_channels = 2; // FIXME
-                msg_Warn( p_enc, "stereo mode selected (codec limitation)" );
-            }
+            msg_Err( p_enc, "cannot open encoder" );
+            dialog_Fatal( p_enc, _("Streaming / Transcoding failed"),
+                    "%s", _("VLC could not open the encoder.") );
+            av_dict_free(&options);
+            goto error;
+        }
 
-            if( i_codec_id == AV_CODEC_ID_MP2 || i_codec_id == AV_CODEC_ID_MP3 )
-            {
-                int i_frequency, i;
+        if( p_context->channels > 2 )
+        {
+            p_context->channels = 2;
+            p_enc->fmt_in.audio.i_channels = 2; // FIXME
+            msg_Warn( p_enc, "stereo mode selected (codec limitation)" );
+        }
 
-                for ( i_frequency = 0; i_frequency < 6; i_frequency++ )
-                {
-                    if ( p_enc->fmt_out.audio.i_rate
-                            == mpa_freq_tab[i_frequency] )
-                        break;
-                }
-                if ( i_frequency == 6 )
-                {
-                    msg_Err( p_enc, "MPEG audio doesn't support frequency=%d",
-                             p_enc->fmt_out.audio.i_rate );
-                    free( p_sys );
-                    return VLC_EGENERIC;
-                }
+        if( i_codec_id == AV_CODEC_ID_MP2 || i_codec_id == AV_CODEC_ID_MP3 )
+        {
+            int i_frequency, i;
+            es_format_t *fmt = &p_enc->fmt_out;
 
-                for ( i = 1; i < 14; i++ )
-                {
-                    if ( p_enc->fmt_out.i_bitrate / 1000
-                          <= mpa_bitrate_tab[i_frequency / 3][i] )
-                        break;
-                }
-                if ( p_enc->fmt_out.i_bitrate / 1000
-                      != mpa_bitrate_tab[i_frequency / 3][i] )
-                {
-                    msg_Warn( p_enc,
-                              "MPEG audio doesn't support bitrate=%d, using %d",
-                              p_enc->fmt_out.i_bitrate,
-                              mpa_bitrate_tab[i_frequency / 3][i] * 1000 );
-                    p_enc->fmt_out.i_bitrate =
-                        mpa_bitrate_tab[i_frequency / 3][i] * 1000;
-                    p_context->bit_rate = p_enc->fmt_out.i_bitrate;
-                }
+            for ( i_frequency = 0; i_frequency < 6; i_frequency++ )
+                if ( fmt->audio.i_rate == mpa_freq_tab[i_frequency] )
+                    break;
+
+            if ( i_frequency == 6 )
+            {
+                msg_Err( p_enc, "MPEG audio doesn't support frequency=%d",
+                        fmt->audio.i_rate );
+                av_dict_free(&options);
+                goto error;
             }
 
-            p_context->codec = NULL;
-            vlc_avcodec_lock();
-            ret = avcodec_open2( p_context, p_codec, NULL /* options */ );
-            vlc_avcodec_unlock();
-            if( ret )
+            for ( i = 1; i < 14; i++ )
+                if (fmt->i_bitrate/1000 <= mpa_bitrate_tab[i_frequency / 3][i])
+                    break;
+
+            if (fmt->i_bitrate / 1000 != mpa_bitrate_tab[i_frequency / 3][i])
             {
-                msg_Err( p_enc, "cannot open encoder" );
-                dialog_Fatal( p_enc,
-                                _("Streaming / Transcoding failed"),
-                                "%s", _("VLC could not open the encoder.") );
-                free( p_sys );
-                return VLC_EGENERIC;
+                msg_Warn( p_enc,
+                        "MPEG audio doesn't support bitrate=%d, using %d",
+                        fmt->i_bitrate,
+                        mpa_bitrate_tab[i_frequency / 3][i] * 1000 );
+                fmt->i_bitrate = mpa_bitrate_tab[i_frequency / 3][i] * 1000;
+                p_context->bit_rate = fmt->i_bitrate;
             }
         }
-        else
+
+        p_context->codec = NULL;
+        vlc_avcodec_lock();
+        ret = avcodec_open2( p_context, p_codec, options ? &options : NULL );
+        vlc_avcodec_unlock();
+        if( ret )
         {
             msg_Err( p_enc, "cannot open encoder" );
-            dialog_Fatal( p_enc, _("Streaming / Transcoding failed"),
-                            "%s", _("VLC could not open the encoder.") );
-            free( p_sys );
-            return VLC_EGENERIC;
+            dialog_Fatal( p_enc,
+                    _("Streaming / Transcoding failed"),
+                    "%s", _("VLC could not open the encoder.") );
+            av_dict_free(&options);
+            goto error;
         }
     }
 
-    if( i_codec_id == AV_CODEC_ID_FLAC )
+    av_dict_free(&options);
+
+    p_enc->fmt_out.i_extra = p_context->extradata_size;
+    if( p_enc->fmt_out.i_extra )
     {
-        p_enc->fmt_out.i_extra = 4 + 1 + 3 + p_context->extradata_size;
         p_enc->fmt_out.p_extra = malloc( p_enc->fmt_out.i_extra );
-        if( p_enc->fmt_out.p_extra )
-        {
-            uint8_t *p = p_enc->fmt_out.p_extra;
-            p[0] = 0x66;    /* f */
-            p[1] = 0x4C;    /* L */
-            p[2] = 0x61;    /* a */
-            p[3] = 0x43;    /* C */
-            p[4] = 0x80;    /* streaminfo block, last block before audio */
-            p[5] = ( p_context->extradata_size >> 16 ) & 0xff;
-            p[6] = ( p_context->extradata_size >>  8 ) & 0xff;
-            p[7] = ( p_context->extradata_size       ) & 0xff;
-            memcpy( &p[8], p_context->extradata, p_context->extradata_size );
-        }
-        else
-        {
-            p_enc->fmt_out.i_extra = 0;
-        }
-    }
-    else
-    {
-        p_enc->fmt_out.i_extra = p_context->extradata_size;
-        if( p_enc->fmt_out.i_extra )
+        if ( p_enc->fmt_out.p_extra == NULL )
         {
-            p_enc->fmt_out.p_extra = malloc( p_enc->fmt_out.i_extra );
-            if ( p_enc->fmt_out.p_extra == NULL )
-            {
-                goto error;
-            }
-            memcpy( p_enc->fmt_out.p_extra, p_context->extradata,
-                    p_enc->fmt_out.i_extra );
+            goto error;
         }
+        memcpy( p_enc->fmt_out.p_extra, p_context->extradata,
+                p_enc->fmt_out.i_extra );
     }
 
     p_context->flags &= ~CODEC_FLAG_GLOBAL_HEADER;
@@ -821,23 +841,28 @@ int OpenEncoder( vlc_object_t *p_this )
         p_enc->fmt_in.i_codec = GetVlcAudioFormat( p_sys->p_context->sample_fmt );
         p_enc->fmt_in.audio.i_bitspersample = aout_BitsPerSample( p_enc->fmt_in.i_codec );
 
-        p_sys->i_sample_bytes = (p_enc->fmt_in.audio.i_bitspersample / 8) *
-                                p_context->channels;
+        p_sys->i_sample_bytes = (p_enc->fmt_in.audio.i_bitspersample / 8);
         p_sys->i_frame_size = p_context->frame_size > 1 ?
                                     p_context->frame_size :
-                                    RAW_AUDIO_FRAME_SIZE;
-        p_sys->p_buffer = malloc( p_sys->i_frame_size * p_sys->i_sample_bytes );
-        if ( p_sys->p_buffer == NULL )
+                                    FF_MIN_BUFFER_SIZE;
+        p_sys->p_buffer = malloc( p_sys->i_frame_size * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels);
+        if ( unlikely( p_sys->p_buffer == NULL ) )
         {
             goto error;
         }
         p_enc->fmt_out.audio.i_blockalign = p_context->block_align;
         p_enc->fmt_out.audio.i_bitspersample = aout_BitsPerSample( p_enc->fmt_out.i_codec );
+        //b_variable tells if we can feed any size frames to encoder
+        p_sys->b_variable = p_context->frame_size ? false : true;
 
-        if( p_context->frame_size > 1 )
-            p_sys->i_buffer_out = 8 * AVCODEC_MAX_AUDIO_FRAME_SIZE;
-        else
-            p_sys->i_buffer_out = p_sys->i_frame_size * p_sys->i_sample_bytes;
+        p_sys->i_buffer_out = p_sys->i_frame_size * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels;
+
+        if( p_sys->b_planar )
+        {
+            p_sys->p_interleave_buf = malloc( p_sys->i_buffer_out );
+            if( unlikely( p_sys->p_interleave_buf == NULL ) )
+                goto error;
+        }
     }
 
     p_sys->frame = avcodec_alloc_frame();
@@ -855,6 +880,7 @@ int OpenEncoder( vlc_object_t *p_this )
 error:
     free( p_enc->fmt_out.p_extra );
     free( p_sys->p_buffer );
+    free( p_sys->p_interleave_buf );
     free( p_sys );
     return VLC_ENOMEM;
 }
@@ -865,8 +891,7 @@ error:
 static block_t *EncodeVideo( encoder_t *p_enc, picture_t *p_pict )
 {
     encoder_sys_t *p_sys = p_enc->p_sys;
-    int i_out, i_plane;
-
+    int i_plane;
     /* Initialize the video output buffer the first time.
      * This is done here instead of OpenEncoder() because we need the actual
      * bits_per_pixel value, without having to assume anything.
@@ -875,9 +900,13 @@ static block_t *EncodeVideo( encoder_t *p_enc, picture_t *p_pict )
                          p_enc->fmt_out.video.i_bits_per_pixel / 8 : 3;
     const int blocksize = __MAX( FF_MIN_BUFFER_SIZE,bytesPerPixel * p_sys->p_context->height * p_sys->p_context->width + 200 );
     block_t *p_block = block_Alloc( blocksize );
+    if( unlikely(p_block == NULL) )
+        return NULL;
 
+    AVFrame *frame = NULL;
     if( likely(p_pict) ) {
-        avcodec_get_frame_defaults( p_sys->frame );
+        frame = p_sys->frame;
+        avcodec_get_frame_defaults( frame );
         for( i_plane = 0; i_plane < p_pict->i_planes; i_plane++ )
         {
             p_sys->frame->data[i_plane] = p_pict->p[i_plane].p_pixels;
@@ -885,20 +914,20 @@ static block_t *EncodeVideo( encoder_t *p_enc, picture_t *p_pict )
         }
 
         /* Let libavcodec select the frame type */
-        p_sys->frame->pict_type = 0;
+        frame->pict_type = 0;
 
-        p_sys->frame->repeat_pict = p_pict->i_nb_fields - 2;
-        p_sys->frame->interlaced_frame = !p_pict->b_progressive;
-        p_sys->frame->top_field_first = !!p_pict->b_top_field_first;
+        frame->repeat_pict = p_pict->i_nb_fields - 2;
+        frame->interlaced_frame = !p_pict->b_progressive;
+        frame->top_field_first = !!p_pict->b_top_field_first;
 
         /* Set the pts of the frame being encoded */
-        p_sys->frame->pts = p_pict->date ? p_pict->date : (int64_t)AV_NOPTS_VALUE;
+        frame->pts = p_pict->date ? p_pict->date : (int64_t)AV_NOPTS_VALUE;
 
-        if ( p_sys->b_hurry_up && p_sys->frame->pts != (int64_t)AV_NOPTS_VALUE )
+        if ( p_sys->b_hurry_up && frame->pts != (int64_t)AV_NOPTS_VALUE )
         {
             mtime_t current_date = mdate();
 
-            if ( current_date + HURRY_UP_GUARD3 > p_sys->frame->pts )
+            if ( current_date + HURRY_UP_GUARD3 > frame->pts )
             {
                 p_sys->p_context->mb_decision = FF_MB_DECISION_SIMPLE;
                 p_sys->p_context->trellis = 0;
@@ -908,11 +937,11 @@ static block_t *EncodeVideo( encoder_t *p_enc, picture_t *p_pict )
             {
                 p_sys->p_context->mb_decision = p_sys->i_hq;
 
-                if ( current_date + HURRY_UP_GUARD2 > p_sys->frame->pts )
+                if ( current_date + HURRY_UP_GUARD2 > frame->pts )
                 {
                     p_sys->p_context->trellis = 0;
                     p_sys->p_context->noise_reduction = p_sys->i_noise_reduction
-                        + (HURRY_UP_GUARD2 + current_date - p_sys->frame->pts) / 500;
+                        + (HURRY_UP_GUARD2 + current_date - frame->pts) / 500;
                     msg_Dbg( p_enc, "hurry up mode 2" );
                 }
                 else
@@ -924,45 +953,60 @@ static block_t *EncodeVideo( encoder_t *p_enc, picture_t *p_pict )
                 }
             }
 
-            if ( current_date + HURRY_UP_GUARD1 > p_sys->frame->pts )
+            if ( current_date + HURRY_UP_GUARD1 > frame->pts )
             {
-                p_sys->frame->pict_type = AV_PICTURE_TYPE_P;
+                frame->pict_type = AV_PICTURE_TYPE_P;
                 /* msg_Dbg( p_enc, "hurry up mode 1 %lld", current_date + HURRY_UP_GUARD1 - frame.pts ); */
             }
         }
 
-        if ( p_sys->frame->pts != (int64_t)AV_NOPTS_VALUE && p_sys->frame->pts != 0 )
+        if ( frame->pts != (int64_t)AV_NOPTS_VALUE && frame->pts != 0 )
         {
-            if ( p_sys->i_last_pts == p_sys->frame->pts )
+            if ( p_sys->i_last_pts == frame->pts )
             {
-                msg_Warn( p_enc, "almost fed libavcodec with two frames with the "
-                         "same PTS (%"PRId64 ")", p_sys->frame->pts );
+                msg_Warn( p_enc, "almost fed libavcodec with two frames with "
+                          "the same PTS (%"PRId64 ")", frame->pts );
                 return NULL;
             }
-            else if ( p_sys->i_last_pts > p_sys->frame->pts )
+            else if ( p_sys->i_last_pts > frame->pts )
             {
                 msg_Warn( p_enc, "almost fed libavcodec with a frame in the "
                          "past (current: %"PRId64 ", last: %"PRId64")",
-                         p_sys->frame->pts, p_sys->i_last_pts );
+                         frame->pts, p_sys->i_last_pts );
                 return NULL;
             }
             else
-            {
-                p_sys->i_last_pts = p_sys->frame->pts;
-            }
+                p_sys->i_last_pts = frame->pts;
         }
 
-        p_sys->frame->quality = p_sys->i_quality;
-
-        i_out = avcodec_encode_video( p_sys->p_context, p_block->p_buffer,
-                                     p_block->i_buffer, p_sys->frame );
+        frame->quality = p_sys->i_quality;
     }
-    else
+
+#if (LIBAVCODEC_VERSION_MAJOR >= 54)
+    AVPacket av_pkt;
+    int is_data;
+
+    av_init_packet( &av_pkt );
+    av_pkt.data = p_block->p_buffer;
+    av_pkt.size = p_block->i_buffer;
+
+    if( avcodec_encode_video2( p_sys->p_context, &av_pkt, frame, &is_data ) < 0
+     || is_data == 0 )
     {
-        i_out = avcodec_encode_video( p_sys->p_context, p_block->p_buffer,
-                                     p_block->i_buffer, NULL);
+        block_Release( p_block );
+        return NULL;
     }
 
+    p_block->i_buffer = av_pkt.size;
+    p_block->i_length = av_pkt.duration / p_sys->p_context->time_base.den;
+    p_block->i_pts = av_pkt.pts;
+    p_block->i_dts = av_pkt.dts;
+    if( unlikely( av_pkt.flags & AV_PKT_FLAG_CORRUPT ) )
+        p_block->i_flags |= BLOCK_FLAG_CORRUPTED;
+
+#else
+    int i_out = avcodec_encode_video( p_sys->p_context, p_block->p_buffer,
+                                      p_block->i_buffer, frame );
     if( i_out <= 0 )
     {
         block_Release( p_block );
@@ -1016,19 +1060,24 @@ static block_t *EncodeVideo( encoder_t *p_enc, picture_t *p_pict )
          * correctly */
         p_block->i_dts = p_block->i_pts = p_pict->date;
     }
+#endif
 
     switch ( p_sys->p_context->coded_frame->pict_type )
     {
     case AV_PICTURE_TYPE_I:
+    case AV_PICTURE_TYPE_SI:
         p_block->i_flags |= BLOCK_FLAG_TYPE_I;
         break;
     case AV_PICTURE_TYPE_P:
+    case AV_PICTURE_TYPE_SP:
         p_block->i_flags |= BLOCK_FLAG_TYPE_P;
         break;
     case AV_PICTURE_TYPE_B:
+    case AV_PICTURE_TYPE_BI:
         p_block->i_flags |= BLOCK_FLAG_TYPE_B;
         break;
-
+    default:
+        p_block->i_flags |= BLOCK_FLAG_TYPE_PB;
     }
 
     return p_block;
@@ -1042,49 +1091,76 @@ static block_t *EncodeAudio( encoder_t *p_enc, block_t *p_aout_buf )
     encoder_sys_t *p_sys = p_enc->p_sys;
 
     block_t *p_block, *p_chain = NULL;
-    int got_packet,i_out,i_samples_left=0,i_data_offset = 0;
+    int got_packet,i_out;
+    size_t buffer_delay = 0, i_samples_left = 0;
 
-    //i_samples_left is amount of samples we get
+    //i_bytes_left is amount of bytes we get
     i_samples_left = p_aout_buf ? p_aout_buf->i_nb_samples : 0;
-
-    if( p_sys->i_samples_delay > 0 )
+    buffer_delay = p_sys->i_samples_delay * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels;
+
+    //p_sys->i_buffer_out = p_sys->i_frame_size * chan * p_sys->i_sample_bytes
+    //Calculate how many bytes we would need from current buffer to fill frame
+    size_t leftover_samples = __MAX(0,__MIN((ssize_t)i_samples_left, (ssize_t)(p_sys->i_frame_size - p_sys->i_samples_delay)));
+
+    // Check if we have enough samples in delay_buffer and current p_aout_buf to fill frame
+    // Or if we are cleaning up
+    if( ( buffer_delay > 0 ) &&
+            ( ( p_aout_buf && ( leftover_samples <= p_aout_buf->i_nb_samples ) &&
+               ( (leftover_samples + p_sys->i_samples_delay ) >= p_sys->i_frame_size )
+              ) ||
+             ( !p_aout_buf ) 
+            )
+         )
     {
-        AVPacket packet;
         //How much we need to copy from new packet
-        const int leftover = __MAX(0,__MIN(i_samples_left, (p_sys->i_frame_size - p_sys->i_samples_delay)));
+        const int leftover = leftover_samples * p_enc->fmt_in.audio.i_channels * p_sys->i_sample_bytes;
+
+#if LIBAVUTIL_VERSION_CHECK( 51,27,2,46,100 )
+        const int align = 0;
+#else
+        const int align = 1;
+#endif
 
+        AVPacket packet = {0};
         avcodec_get_frame_defaults( p_sys->frame );
-        p_sys->frame->nb_samples = p_sys->i_samples_delay + leftover;
         p_sys->frame->format     = p_sys->p_context->sample_fmt;
+        p_sys->frame->pts        = date_Get( &p_sys->buffer_date );
+        p_sys->frame->nb_samples = leftover_samples + p_sys->i_samples_delay;
+        date_Increment( &p_sys->buffer_date, p_sys->i_frame_size );
 
-        //Copy samples from new packet to buffer to get frame size
-        if( likely( leftover ) )
+        if( likely( p_aout_buf ) )
         {
-            if( av_sample_fmt_is_planar( p_sys->p_context->sample_fmt ) )
-                aout_Deinterleave( p_sys->p_buffer+(p_sys->i_samples_delay*p_sys->i_sample_bytes*p_enc->fmt_in.audio.i_channels ),
-                                  p_aout_buf->p_buffer, leftover, p_enc->fmt_in.audio.i_channels, p_enc->fmt_in.i_codec );
+            p_aout_buf->i_nb_samples -= leftover_samples;
+            memcpy( p_sys->p_buffer+buffer_delay, p_aout_buf->p_buffer, leftover );
+
+            // We need to deinterleave from p_aout_buf to p_buffer the leftover bytes
+            if( p_sys->b_planar )
+                aout_Deinterleave( p_sys->p_interleave_buf, p_sys->p_buffer,
+                    p_sys->i_frame_size, p_enc->fmt_in.audio.i_channels, p_enc->fmt_in.i_codec );
             else
-                memcpy( p_sys->p_buffer+(p_sys->i_samples_delay*p_sys->i_sample_bytes),
-                        p_aout_buf->p_buffer,
-                        leftover * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels
-                       );
-        }
+                memcpy( p_sys->p_buffer + buffer_delay, p_aout_buf->p_buffer, leftover);
 
+            p_aout_buf->p_buffer     += leftover;
+            p_aout_buf->i_buffer     -= leftover;
+            p_aout_buf->i_pts         = date_Get( &p_sys->buffer_date );
+        }
+        if(unlikely( ( (leftover + buffer_delay) < p_sys->i_buffer_out ) &&
+                     !(p_sys->p_codec->capabilities & CODEC_CAP_SMALL_LAST_FRAME ))
+          )
+        {
+            msg_Dbg( p_enc, "No small last frame support, padding");
+            size_t padding_size = p_sys->i_buffer_out - (leftover+buffer_delay);
+            memset( p_sys->p_buffer + (leftover+buffer_delay), 0, padding_size );
+            buffer_delay += padding_size;
+        }
         if( avcodec_fill_audio_frame( p_sys->frame, p_enc->fmt_in.audio.i_channels,
-                              p_sys->p_context->sample_fmt,
-                              p_sys->p_buffer,
-                              (p_sys->i_samples_delay + leftover) * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels,
-                              0) < 0 )
-            msg_Err( p_enc, "Filling on leftovers error i_leftover %d i_samples_left %d samples_delay %d frame size %d", leftover, i_samples_left, p_sys->i_samples_delay, p_sys->i_frame_size );
-
-        if( likely( p_aout_buf ) )
-            p_sys->frame->pts = p_aout_buf->i_pts -
-                     (mtime_t)1000000 * (mtime_t)p_sys->i_samples_delay /
-                     (mtime_t)p_enc->fmt_in.audio.i_rate;
+                p_sys->p_context->sample_fmt, p_sys->b_planar ? p_sys->p_interleave_buf : p_sys->p_buffer,
+                leftover + buffer_delay,
+                align) < 0 )
+            msg_Err( p_enc, "filling error on fillup" );
 
+        buffer_delay = 0;
         p_sys->i_samples_delay = 0;
-        i_data_offset += leftover * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels;
-        i_samples_left -= leftover;
 
         p_block = block_Alloc( p_sys->i_buffer_out );
         av_init_packet( &packet );
@@ -1092,34 +1168,33 @@ static block_t *EncodeAudio( encoder_t *p_enc, block_t *p_aout_buf )
         packet.size = p_block->i_buffer;
 
         i_out = avcodec_encode_audio2( p_sys->p_context, &packet, p_sys->frame, &got_packet );
-        p_block->i_buffer = packet.size;
 
-
-        if( unlikely( !got_packet || ( i_out < 0 ) ) )
+        if( unlikely( !got_packet || ( i_out < 0 ) || !packet.size ) )
         {
             if( i_out < 0 )
             {
-                msg_Err( p_enc,"Encoding problem...");
+                msg_Err( p_enc,"Encoding problem..");
                 return p_chain;
             }
-        } else {
-            p_block->i_buffer = packet.size;
+            block_Release( p_block );
+            return NULL;
+        }
 
-            p_block->i_length = (mtime_t)1000000 *
-             (mtime_t)p_sys->i_frame_size /
-             (mtime_t)p_sys->p_context->sample_rate;
+        p_block->i_buffer = packet.size;
+        p_block->i_length = (mtime_t)1000000 *
+            (mtime_t)p_sys->frame->nb_samples /
+            (mtime_t)p_sys->p_context->sample_rate;
 
-            p_block->i_dts = p_block->i_pts = packet.pts;
+        p_block->i_dts = p_block->i_pts = packet.pts;
 
-            block_ChainAppend( &p_chain, p_block );
-        }
+        block_ChainAppend( &p_chain, p_block );
     }
 
     if( unlikely( !p_aout_buf ) )
     {
         msg_Dbg(p_enc,"Flushing..");
         do {
-            AVPacket packet;
+            AVPacket packet = {0};
             p_block = block_Alloc( p_sys->i_buffer_out );
             av_init_packet( &packet );
             packet.data = p_block->p_buffer;
@@ -1134,42 +1209,54 @@ static block_t *EncodeAudio( encoder_t *p_enc, block_t *p_aout_buf )
 
             p_block->i_dts = p_block->i_pts = packet.pts;
 
-            if( !i_out && got_packet )
+            if( i_out >= 0 && got_packet )
                 block_ChainAppend( &p_chain, p_block );
-        } while( got_packet && !i_out );
+        } while( got_packet && (i_out>=0) );
         return p_chain;
     }
 
-    while( i_samples_left >= p_sys->i_frame_size )
+
+    while( ( p_aout_buf->i_nb_samples >= p_sys->i_frame_size ) ||
+           ( p_sys->b_variable && p_aout_buf->i_nb_samples ) )
     {
         AVPacket packet = {0};
+#if LIBAVUTIL_VERSION_CHECK( 51,27,2,46,100 )
+        const int align = 0;
+#else
+        const int align = 1;
+#endif
+
+        if( unlikely( p_aout_buf->i_pts > VLC_TS_INVALID &&
+                      p_aout_buf->i_pts != date_Get( &p_sys->buffer_date ) ) )
+            date_Set( &p_sys->buffer_date, p_aout_buf->i_pts );
 
-        p_sys->frame->nb_samples = p_sys->i_frame_size;
+        avcodec_get_frame_defaults( p_sys->frame );
+        if( p_sys->b_variable )
+            p_sys->frame->nb_samples = p_aout_buf->i_nb_samples;
+        else
+            p_sys->frame->nb_samples = p_sys->i_frame_size;
         p_sys->frame->format     = p_sys->p_context->sample_fmt;
+        p_sys->frame->pts        = date_Get( &p_sys->buffer_date );
 
-        if( av_sample_fmt_is_planar( p_sys->p_context->sample_fmt ) )
+        if( p_sys->b_planar )
         {
-            aout_Deinterleave( p_sys->p_buffer, p_aout_buf->p_buffer+i_data_offset,
-                              p_sys->i_frame_size, p_enc->fmt_in.audio.i_channels, p_enc->fmt_in.i_codec );
-            if( avcodec_fill_audio_frame( p_sys->frame, p_enc->fmt_in.audio.i_channels,
-                                    p_sys->p_context->sample_fmt,
-                                    p_sys->p_buffer,
-                                    p_sys->i_frame_size * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels,
-                                    0) < 0 )
-                 msg_Err( p_enc, "filling error on encode" );
-        } else {
-            if( avcodec_fill_audio_frame( p_sys->frame, p_enc->fmt_in.audio.i_channels,
+            aout_Deinterleave( p_sys->p_buffer, p_aout_buf->p_buffer,
+                               p_sys->frame->nb_samples, p_enc->fmt_in.audio.i_channels, p_enc->fmt_in.i_codec );
+
+        }
+
+        if( avcodec_fill_audio_frame( p_sys->frame, p_enc->fmt_in.audio.i_channels,
                                     p_sys->p_context->sample_fmt,
-                                    p_aout_buf->p_buffer+i_data_offset,
-                                    p_sys->i_frame_size * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels,
-                                    0) < 0 )
+                                    p_sys->b_planar ? p_sys->p_buffer : p_aout_buf->p_buffer,
+                                    __MIN(p_sys->i_buffer_out, p_aout_buf->i_buffer),
+                                    align) < 0 )
                  msg_Err( p_enc, "filling error on encode" );
-        }
 
-        i_samples_left -= p_sys->i_frame_size;
-        i_data_offset += p_sys->i_frame_size * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels;
+        p_aout_buf->p_buffer     += (p_sys->frame->nb_samples * p_enc->fmt_in.audio.i_channels * p_sys->i_sample_bytes);
+        p_aout_buf->i_buffer     -= (p_sys->frame->nb_samples * p_enc->fmt_in.audio.i_channels * p_sys->i_sample_bytes);
+        p_aout_buf->i_nb_samples -= p_sys->frame->nb_samples;
+        date_Increment( &p_sys->buffer_date, p_sys->frame->nb_samples );
 
-        p_sys->frame->pts = p_aout_buf->i_pts;
 
         p_block = block_Alloc( p_sys->i_buffer_out );
         av_init_packet( &packet );
@@ -1193,25 +1280,21 @@ static block_t *EncodeAudio( encoder_t *p_enc, block_t *p_aout_buf )
         p_block->i_buffer = packet.size;
 
         p_block->i_length = (mtime_t)1000000 *
-            (mtime_t)p_sys->i_frame_size /
+            (mtime_t)p_sys->frame->nb_samples /
             (mtime_t)p_sys->p_context->sample_rate;
 
         p_block->i_dts = p_block->i_pts = packet.pts;
 
         block_ChainAppend( &p_chain, p_block );
     }
-    if( i_samples_left < 0 )
-        msg_Err( p_enc, "I_data_left overflow");
 
     // We have leftover samples that don't fill frame_size, and libavcodec doesn't seem to like
     // that frame has more data than p_sys->i_frame_size most of the cases currently.
-    if( i_samples_left > 0 )
+    if( p_aout_buf->i_nb_samples > 0 )
     {
-        if( av_sample_fmt_is_planar( p_sys->p_context->sample_fmt ) )
-            aout_Deinterleave( p_sys->p_buffer, p_aout_buf->p_buffer+i_data_offset, i_samples_left, p_enc->fmt_in.audio.i_channels, p_enc->fmt_in.i_codec );
-        else
-            memcpy( p_sys->p_buffer, p_aout_buf->p_buffer+i_data_offset , i_samples_left*p_sys->i_sample_bytes*p_enc->fmt_in.audio.i_channels);
-        p_sys->i_samples_delay = i_samples_left;
+       memcpy( p_sys->p_buffer + buffer_delay, p_aout_buf->p_buffer,
+               p_aout_buf->i_nb_samples * p_sys->i_sample_bytes * p_enc->fmt_in.audio.i_channels);
+       p_sys->i_samples_delay += p_aout_buf->i_nb_samples;
     }
 
     return p_chain;
@@ -1234,6 +1317,7 @@ void CloseEncoder( vlc_object_t *p_this )
     av_free( p_sys->p_context );
 
 
+    free( p_sys->p_interleave_buf );
     free( p_sys->p_buffer );
 
     free( p_sys );