[Home]Import Modules/Import Nvrec


This import module connects NVrec - a very good v4l1/v4l2 grabber - and transcode. NVrec is very good, because of its audio/video synch mechanism which maintains perfect A/V synch over a really long time. Read more about NVrec at http://www.ee.up.ac.za/~justin/

"Isn't this braindead? I though NVrec can write divx/divx4/... files as well?" You might ask. Yes, this is true, but transcode has much more features, such as better deinterlacing, video postprocessing, clipping, etc.

In order to accomplish the connection between NVrec and transcode I wrote an "export module" for NVrec called RAWrec (which is now divx4rec -o raw://) In this mode, divx4rec basically writes its video data to stdout (where it gets capured from transcode) and the audio data to a mp3 file in an AVI container.

Before you get completly confused, here are some usage examples

  $ transcode -V -i /dev/v4l2/capture0 -x nvrec,null -y xvidcvs,null \
  -o video.avi -g 512x384 -u 100 -m audio.avi -H0

Since import_nvrec only works in YUV mode '-V' is mandatory. Use your v4l1/2 device as the input file or transcode will assume /dev/video. Write encoded video stream to video.avi. Write audio stream to audio.avi. Framesize is 512x384. Use 100 buffers.

You now have two AVI files: The video only in video.avi and the audio only in audio.avi. Use avimerge to multiplex them.

  $ avimerge -i video.avi -p audio.avi -a 0 -o complete.avi

Easy, eh?


  /--------\                                  /-----------\
  |        |<----- recording parameters <-----|           |
  |divx4rec|                                  | transcode |
  |-o raw: |---------> video stream --------->|           |
  \--------/                                  \-----------/
      |                                             |
     \|/                                           \|/
      |                                             |
 /---------\                                   /---------\
 |audio.avi|                                   |video.avi|
 \---------/                                   \---------/


divx4rec gets its options from transcode meaning transcode's defaults apply.

Audio parameters can be specified through transcodes '-e' option, the mp3 bitrate is controlled with '-b' , eg

 $ transcode (..) -e 44100,16,2 -b 192

Everything should work like you expect it to work.

If you don't specify the '-m' option, audio will be written to audio.avi. If you need to pass special options to divx4rec, you can pass options by supplying a string after the module name, just write down the divx4rec options. This can be useful if you want to control NVrecs buffer usage

 $ transcode (..) -x nvrec="-H 90 -L 10 -M 50"

The divx4rec command to be executed will get printed out, look for something like

 [import_nvrec.so] divx4rec -o raw://audio.avi -w 512 -h 384 -s -b 16 -r 48000 -ab 128 -aq 5


RAWrec is now integrated into NVrec. Since it was heavily based on divx4rec anyway it became a special output mode of divx4rec. NVrec lives at http://nvrec.sf.net

If you only want divx4rec out of NVrec, configure with

  $ ./configure --without-avifile --without-rte  \
  --without-ffmpeg --without-quicktime --without-sdl --without-mad

and copy divx4rec to some place in your $PATH.

  $ install -m 755 divx4rec /usr/local/bin



  $ transcode -V -i /dev/video -x nvrec,null -y xvidcvs,null \
              -g 512x384 -u 100 -o video.avi -H0

  $ transcode -V -i /dev/video -x nvrec,null -y xvidcvs,null \
              -g 512x384 -u 100 -o video.avi  -b 192 -e44100,16,2 -H0

  $ transcode -V -i /dev/video -x nvrec,null -y xvidcvs,null \
              -g 512x384 -u 100 -o video.avi -m audio-only.avi -H0

  $ transcode -V -i /dev/v4l/capture0 -p /dev/sound/dsp \
              -x nvrec,null -y xvidcvs,null \
              -g 512x384 -u 100 -o video.avi -m audio-only.avi -H0

  $ avimerge -i video.avi -p audio.avi -a 0 -o complete.avi


To squeeze out even more fps you can tell divx4rec to not encode the audio to mp3 and write a WAV file instead. Sebastian <scut@nb.in-berlin.de> writes in http://www.theorie.physik.uni-goettingen.de/ pipermail/transcode-users/2002-June/002575.html (in one line) (adjusted by tibit to reflect the new syntax)

  "I use this command line to record from TV (after setting the correct channel
with xawtv):

transcode --duration 01:00:00 \

          -x nvrec="-wav",null 
          -i /dev/video0 -V -H0 -u 100 \
          -g 576x432 \
          -y divx5,null -w 1200 -o output-video.avi \
          -e 44100,16,2 \
          -J dilyuvmmx

This produces the files "output-video.avi" and "audio.avi", which is just a WAV file. By using this method instead of encoding directly to MP3, I can archieve a +3 or +4 frame rate and can encoder a higher solution. After recording, I merge both files:

 transcode -i output-video.avi -p audio.wav -P 1 -b 128,0 \
           -o output-complete.avi -y raw

The output-complete.avi file is the finished recording, but with ads. I use 'avidemux' to cut out the commercials, which is easy since avidemux has a "skip to next keyframe" button. Usually at the end of commercials there is a new keyframe. Use the markers begin/end and the delete option from the "mark" menu to cut the ads out. Then write a new file with A/V to disk. For me the smart mode sometimes segfaults, and I do not use it."

Another Usage scenario is described by Christoph Gaitzsch <christoph.gaitzsch@t-online.de> in http://www.theorie.physik.uni-goettingen.de/pipermail/transcode-users/2002-August/003946.html (adjusted by tibit to reflect the new syntax)

"I also use my PC as a video recorder on a Athlon 1G. After experimenting much with transcode, nvrec, lavrec, etc. I now capture with transcode via nvrec, save this as yuv4mpeg raw stream to a fast disk and postprocess later. I found that the quality is much better when encoding in a separate step than live, even when I have no framedrops.

Basically I record with the following commands:

Pass 1 (recording)

 transcode -V -i /dev/video0 -x nvrec="-N 32",null -y yuv4mpeg,null 
           -o out_video.yuv4mpeg -g 384x288 -u 100 -m out_audio.avi 
           -H0 -f 25

Pass 2 (encoding)

nice -19 transcode -i out_video.yuv4mpeg -p out_audio.avi \

                   -g 384x288 -f 25 -x yuv4mpeg,mp3 -y ffmpeg4 -w 3000 \
                   -o out.avi -u 100 -J dnr,yuvdenoise,32detect=force_mode=3

The advantage is, that the recording stage needs ~5% CPU time, so I have no framedrops even when the machine runs updatedb or other, CPU-consuming cron jobs. It is also possible to record while encoding (because it's niced). Since I schedule my recordings via at and the second pass lasts very long (~4fps), it's an issue to record at low CPU usage.

The filters used in this way, dnr first, then yuvdenoise give me very good picture quality. Thanks to Jason Lunz' for his work on filter comparison."


(newest last)

Transcode Wiki | Import Modules | Recent Changes | Preferences
Password required to edit | View other revisions
Last edited January 4, 2005 4:32 pm by Monroe (diff)