avvidenc: timestamp output buffers

Same approach as x264enc, with the one hour offset

1 files changed, 11 insertions, 0 deletions

diff --git a/ext/libav/gstavvidenc.c b/ext/libav/gstavvidenc.c
index 5193a01..8211de2 100644
--- a/ext/libav/gstavvidenc.c
+++ b/ext/libav/gstavvidenc.c
@@ -682,6 +682,15 @@ gst_ffmpegvidenc_receive_packet (GstFFMpegVidEnc * ffmpegenc,
       GST_VIDEO_CODEC_FRAME_UNSET_SYNC_POINT (frame);
   }
 
+  frame->dts =
+      gst_ffmpeg_time_ff_to_gst (pkt->dts, ffmpegenc->context->time_base);
+  /* This will lose some precision compared to setting the PTS from the input
+   * buffer directly, but that way we're sure PTS and DTS are consistent, in
+   * particular DTS should always be <= PTS
+   */
+  frame->pts =
+      gst_ffmpeg_time_ff_to_gst (pkt->pts, ffmpegenc->context->time_base);
+
   ret = gst_video_encoder_finish_frame (GST_VIDEO_ENCODER (ffmpegenc), frame);
 
 done:
@@ -836,6 +845,8 @@ gst_ffmpegvidenc_start (GstVideoEncoder * encoder)
     return FALSE;
   }
 
+  gst_video_encoder_set_min_pts (encoder, GST_SECOND * 60 * 60 * 1000);
+
   return TRUE;
 }