Install alsa-utils, v4l-utils and ffmpeg packages which contain tools to
find and setup Linux video and audio devices. I can also recommend install
vlc video player for checking results but you can use your favorite one. To
cut a random noise between scenes and split the resulting video interactively I
would recommend using avidemux.
Pinnacle Dazzle DVC100 on Linux is supported by em28xx kernel module. I have
a device with id 2304:021a. You can check if the device is visible using
lsusb command after connecting the device to the USB port:
$ lsusb | fgrep DVC100
Bus 001 Device 025: ID 2304:021a Pinnacle Systems, Inc. Dazzle DVC100 Audio Device
Use udevadm info command from udev package to check whether V4L subsystem
properly configured the device. The following command will find the name of the
device created by V4L:
udevadm info /dev/video* | egrep 'DEVNAME|ID_V4L_PRODUCT' | fgrep -B 1 'Dazzle'
Use following command to find proper audio device:
$ cat /proc/asound/cards | fgrep DVC100
1 [DVC100 ]: USB-Audio - DVC100
Pinnacle Systems GmbH DVC100 at usb-0000:05:00.4-1, high speed
First column contains the number of ALSA card. It should be combined
with hw: prefix to form the name of ALSA device. For the example above the
name of the device is hw:1.
For a sake of simplicity export names of the video and audio devices as environment variables:
export VIDEO=/dev/video0
export AUDIO=hw:1
Set PAL as a video standard, proper frame size and unmute an audio capture
device using v4l2-ctl:
v4l2-ctl --device $VIDEO \
--set-standard 5 --set-fmt-video=width=720,height=576 \
--set-ctrl mute=0 --set-ctrl volume=31
Check V4L settings are applied:
v4l2-ctl --device $VIDEO --get-standard --get-fmt-video --get-ctrl mute --get-ctrl volume
Use ALSA mixer to increase volume of the audio. Min volume is 0, max volume is
amixer -D $AUDIO sset Line 13
Check that video is captured properly using ffplay:
ffmpeg -ar 44100 -thread_queue_size 1024 -f alsa -i $AUDIO \
-r 25 -thread_queue_size 1024 -i $VIDEO \
-codec copy -f matroska - | ffplay -
On this stage video may have:
Such inconsistencies depend on the video recorder/player/capturer in use and can easily be removed during and after capturing. But the center of the frame should contain a color picture, you should hear the sound and it should not be too noisy.
Typical issues that should be fixed on this stage are using incorrect color coding and incorrect sound rate. If picture is black and white or too noisy then incorrect color coding system is used probably. There are three main standards PAL, NTSC and SECAM. If it is unclear which one was used for the record you can try switching between them:
v4l2-ctl --device $VIDEO --list-standards
v4l2-ctl --device $VIDEO --set-standard <standard>
Second typical issue is an incorrect sound capturing rate. If sound is to noisy
or FFmpeg complains about audio synchronization it may mean sound capturing
rate is incorrect. It is specified by -ar option of FFmpeg. Typical values
are 44100 Hz and 48000 Hz. 44100 worked to me so it is used in FFmpeg
examples provided.
To keep as much quality as possible it is better to split a capturing process
on two stages: fast lossless capturing and compressing. The lossless capturing
is saving content using a lossless compression algorithm. Following command use
ffv1 codec for video and pcm_s16le for audio. Couple of fast filters are
applied before saving in order to simplify postprocessing.
I have only single audio channel on my camera and two channels on Dazzle
capturing device. I connected camera audio output to the left channel but
capture two channels, so I used channelmap FFmpeg filter to leave only left
channel.
Filter crop is applied to remove blurred and black stripes from the video
frame. Four parameters are used:
w and h);x
and y).To determine the values capture one frame from the input video using VLC or
other video player and measure the width of stripes using picture viewer or
editor. In my case resulting values are: crop=w=700:h=556:x=8:y=0.
Check the overall flow using ffplay before capturing:
ffmpeg -ar 44100 -thread_queue_size 1024 -f alsa -i $AUDIO \
-itsoffset -0.5 -r 25 -thread_queue_size 1024 -i $VIDEO \
-filter_complex 'channelmap=FL-0' -filter_complex 'crop=w=700:h=556:x=8:y=0' \
-map 1:v -map 0:a -codec:v ffv1 -codec:a pcm_s16le \
-f matroska - | ffplay -
At this stage the picture and sound should be correct. Audio and video can still be not synchronized but it is difficult to fix on this stage, so we will fix it after capturing.
Last thing is to set priority of the capturing stream to maximum to allow
capturing with near realtime speed. It is not necessary if our system is not
loaded and have disk which is fast enough, but I used HDD on my home server to
capture so I used ionice and nice to prioritize the process over other
userspace processes on ther system.
Final command for capturing:
sudo ionice -c 1 -n 0 nice -n -20 ffmpeg \
-ar 44100 -thread_queue_size 1024 -f alsa -i $AUDIO \
-r 25 -thread_queue_size 1024 -i $VIDEO \
-filter_complex 'channelmap=FL-0' -filter_complex 'crop=w=700:h=556:x=8:y=0' \
-map 1:v -map 0:a -codec:v ffv1 -codec:a pcm_s16le \
raw.mkv
After capturing is done use q button to stop FFmpeg gracefully.
Watch the result of the capturing using video player and find the shift between
audio and video. It can be done with VLC using “Audio delay” feature. Use keys
j and k to slightly change audio delay while it is totaly synchronized. Use
some scene with talking people to do that. Then use the delay in seconds to
shift the audio track in the resulting video. For example if delay is -500ms
(audio is later than video) then use the following command:
ffmpeg -i raw.mkv -itsoffset -0.5 -i raw.mkv -map 0:v -map 1:a -codec copy shifted.mkv
I found that delay can be different for different capture sessions. So be careful and alwasy check the synchronization after capturing.
You may watch the resulting video trim and split is on several fragments using FFmpeg command like this:
ffmpeg -i raw.mkv -ss 00:00:00 -t 00:14:00 -codec copy trimmed.mkv
Before final compression select the codecs and find out parameters which suits
your case best. I used h264 and mp3. FFmpeg wiki has a nice guide on
selecting the parameters, see H.264 Video Encoding
Guide and MP3 Encoding
Guide.
You need applying a filter which removes interlaced artifacts before
compression. I used yadif, see some related notes
here.
-pix_fmt yuv420p is used in command below to make result compatible with
older players.
I also found a recommendation to use -flags +ilme+ildct in the FFmpeg FAQ.
But using these flags led to shaking video effects in dynamic scenes, so I
totally removed these flags.
The final command for compression:
ffmpeg -i shifted.mkv -filter:v yadif \
-codec:a libmp3lame -q:a 5 \
-codec:v libx264 -crf 26 -preset slow \
-tune film -pix_fmt yuv420p compressed.mkv
Avidemux is a perfect tool for post compression cutting and splitting. It allows interactively moving between key frames and momentarily remove pieces of video before key frame without recompression.