GStreamer - converting ts to mkv. No audio in video file

Question

I'm using RPi2 with Raspbian. I'm trying to find best solution for transcoding .ts files from tvheadend. At the moment the best is using GStreamer because I can handle best performace. With GStreamer I can transcode 1 minute .ts FHD video file to .mkv 480x270 in 30 seconds which is in my opinion very good performance and I want to keep it.

But I have problem with Gstreamer to make conversion .ts to mkv (or .mp4) with audio. Important is that I have to have output video resized.

Here is example which is working but no audio:

gst-launch-1.0 -v filesrc location=1.ts ! tsdemux parse-private-sections=false name=demux ! queue ! ac3parse ! matroskamux name=stream streamable=true demux. ! queue ! h264parse ! omxh264dec ! omxh264enc target-bitrate=1572864 control-rate=variable ! video/x-h264,stream-format=byte-stream,profile=high,width=480,height=270,framerate=25/1 ! h264parse ! filesink location=1.mkv

Here is another example which working, this time with audio, but I can't resize video by changing omxh264enc parameters from example above:

gst-launch-1.0 filesrc location=1.ts ! decodebin name=demux demux. ! queue ! audioresample ! "audio/x-raw,rate=44100" ! audioconvert ! "audio/x-raw,format=F32LE" ! vorbisenc ! mux. matroskamux name=mux ! filesink location=1.mkv demux. ! queue ! videoconvert ! omxh264enc ! "video/x-h264,profile=high" ! h264parse ! mux.

How to fix these command above to have audio in video files from GStreamer?

Answer 1

Ok, finally I have two options for converting .ts file.

First option is with audio conversion:

gst-launch-1.0 filesrc location=1.ts \
! tsdemux parse-private-sections=false name=demux \
! queue max-size-buffers=0 max-size-time=0 \
! mpegaudioparse \
! mpg123audiodec \
! audioconvert dithering=0 \
! audio/x-raw,channels=2 \
! avenc_aac compliance=-2 bitrate=65536 \
! matroskamux name=stream \
! queue max-size-buffers=0 max-size-time=0 \
! filesink location=1.mkv demux. \
! queue max-size-buffers=0 max-size-time=0 \
! h264parse \
! omxh264dec \
! omxh264enc target-bitrate=900000 control-rate=variable \
! video/x-h264,stream-format=byte-stream,profile=high,width=576,height=324,framerate=25/1 \
! h264parse \
! stream.

Second option is with audio pass no conversion:

gst-launch-1.0 filesrc location=1.ts \
! tsdemux parse-private-sections=false name=demux \
! queue max-size-buffers=0 max-size-time=0 \
! ac3parse \
! matroskamux name=stream \
! queue max-size-buffers=0 max-size-time=0 \
! filesink location=1.mkv demux. \
! queue max-size-buffers=0 max-size-time=0 \
! h264parse \
! omxh264dec \
! omxh264enc target-bitrate=900000 control-rate=variable \
! video/x-h264,stream-format=byte-stream,profile=high,width=576,height=324,framerate=25/1 \
! h264parse \
! stream.

Better results are with first option because output file is smaller and conversion is about 10% faster. Also with this option overall CPU performance is little bigger compare to second option and equals about 25% on RPi2.

Answer 2

You build your pipeline wrong. This is not what you wanted to archive. In particular you don't do anything with with the ac3 muxed audio. And the video is written to disk as H.264 elementary stream. Just because you named it .mkv doesn't mean it is a MKV file.

You need to feed both media to the muxer and write that result to disk. Something like this:

gst-launch-1.0 -v filesrc location=1.ts ! tsdemux parse-private-sections=false name=demux ! queue ! ac3parse ! matroskamux name=stream streamable=true ! filesink location=1.mkv demux. ! queue ! h264parse ! omxh264dec ! omxh264enc target-bitrate=1572864 control-rate=variable ! video/x-h264,stream-format=byte-stream,profile=high,width=480,height=270,framerate=25/1 ! h264parse ! stream.