Saving Your Analog Data from Oblivion

Transferring VHS tapes, audio cassettes, and other analog home media formats to a digital format, such as Ogg or Matroska, can be a complex and expensive process with archival-grade conversions. In this article, I show you a simple and inexpensive method for digitizing your VHS tapes that is perfect for personal use.

Shopping List

To convert a VHS tape to a digital format, you need two pieces of equipment: a playback device for the original medium and a capture device to read the playback device's output. In addition, you need transcoding software to process the data that the capture device retrieves from the playback device. For a list of what I used in this project, see the "Software and Hardware Requirements" box.

For the playback device, you can find a used VCR for less than EUR100 online ($100-$200+ for NTSC/PAL/multisystem). Keep in mind that your chosen playback device must match the color encoding system of the VHS tapes you intend to transfer. Trying to play a PAL VHS on an NTSC device won't work (see the box "PAL vs. NTSC").

For a capture device, I use a USBTV007 EasyCAP (Figure 1), which is inexpensive (less than EUR15/may be difficult to find in the US), performs captures of acceptable quality, and has Linux support. Keep in mind that "EasyCAP" is not a commercial brand name; it is a popular term used by Chinese manufacturers to designate cheap, simple USB capture cards. Finding a specific EasyCAP model can be challenging, since manufacturers usually don't provide the full specification list in their product description or similar use cases. See the box "The Other EasyCAP."

Figure 1: This EasyCAP device works as a cost-effective capture device.

You will also need an SCART to RCA adaptor (Figure 2), which may be purchased for less than EUR10 ($8). The SCART connector is plugged into the VCR output, and the RCA plugs into the EasyCAP connectors. Most RCA connectors are correctly color coded (Table 2).

Figure 2: This SCART to RCA adaptor has an extra pin for audio input, which I won't be using.

In addition, I used a laptop with an i5-2467M CPU for my tests, but you can use a computer with less horsepower. A dual-core CPU of 2.5OGHz for each core is the minimum requirement. However, tests showed that the procedures described in this article will result in a barely acceptable loss of frames in the encoding when using this weak of a CPU. Using a drive with a high write speed is also advisable, since the VHS will be transferred and saved to the filesystem in real time.

In terms of software, I used GStreamer with the Ugly plugin for the capture (see the "Why GStreamer?" box) and FFmpeg to transcode. FFmpeg requires x264 and libdfk-aac support. Finally, the SoX sound processing utility is used to clean up the transferred media's audio.

Hardware Detection

First, power up your computer and connect all the hardware as previously described. Then, make sure your operating system properly identifies the capture device.

The USBTV007 EasyCAP will show up as two different capture devices: one for audio and one for video. This model does not work with the PulseAudio sound server, so you might need to tell PulseAudio to ignore the device so that ALSA can manage it instead. To do this, use the graphical tool pavucontrol: Move to the Configuration tab and select the Off profile for USBTV007. If you prefer to work from the command line, then enter,

pactl set-card-profile $card_number off

where $card_number is the identifier of your EasyCAP device for PulseAudio. If you don't know what it is, you can find out by typing:

pactl list cards | grep -E 'device.product.name|device.string'

To list your currently detected audio inputs, enter the command:

arecord -l

This will display the working capture devices detected by ALSA.

Video inputs will show up as files named /dev/video* (e.g., /dev/video0, /dev/video1, etc.). If you have more than one video input – for example, you have a webcam in addition to an EasyCAP – and you are not sure which one is which, you can extract information from each device with the v4l2-ctl command. The following command will display some known properties of /dev/video0:

v4l2-ctl --device=/dev/video0 --list-inputs

The Capture Process

Once everything is connected and detected, it is time to capture the video. To do this, insert a VHS tape into the VCR and play it, and then turn on your capture software, which takes the VCR's output and dumps it into a file on the fly.

The raw video and audio stream can take up a lot of storage, probably more than 100GB, which is why most people prefer encoding that stream into something more manageable as it is captured. Because the computer is capturing the raw VCR output in real time, your encoder must be both fast and CPU-friendly, so as not to affect the data capture negatively. If you try to use an encoding configuration that is too hard on the CPU, it will not have enough time to process each piece of incoming data before the next one arrives, which will result in lost frames and data loss.

Listing 1 performs a lossless capture of the analog input. While not very practical, Listing 1 is provided as a reference. Notice that the buffers are set to zero in order to avoid problems during the real-time video capture. The pixel format, YUY2, is set to the format that the EasyCAP feeds to the computer. Audio is captured in a lossless format at a sampling rate of 48KHz in stereo. The output is dumped to a Matroska file. Listing 1 is tuned for PAL devices (if you are using NTSC, see the "PAL vs. NTSC" box.)

gst-launch-1.0 -q v4l2src device="$videodevice" do-timestamp=true norm="PAL" pixel-aspect-ratio=1
    ! video/x-raw,format=YUY2,framerate=25/1,width=720,height=576
    ! queue max-size-buffers=0 max-size-time=0 max-size-bytes=0
    ! mux.
  alsasrc device="$alsadevice" do-timestamp=true
    ! audio/x-raw,format=S16LE,rate=48000,channels=2
    ! queue
  max-size-buffers=0 max-size-time=0 max-size-bytes=0
    ! mux. matroskamux name=mux
    ! queue max-size-buffers=0 max-size-time=0 max-size-bytes=0
    ! filesink location=vhs.mkv

In Listing 1, replace "$videodevice" with a video input device (e.g., /dev/video1). The "$alsadevice" (in the alsasrc device line) describes an audio input such as hw:1,0 or hw:2,0. The ALSA audio inputs are always named in a "hw:$card,$device" format.

Listing 2 performs lossy compression while carrying the analog stream's capture, resulting in a smaller file. It is configured to capture video for 1 hour and 40 minutes, but you can stop the recording by pressing q at any time.

ffmpeg -loglevel 32 -t 01:40:00
       -i <( gst-launch-1.0 -q v4l2src device="$videodevice1"
             do-timestamp=true norm="PAL" pixel-aspect-ratio=1
             ! video/x-raw,format=YUY2,framerate=25/1,width=720,height=576
             ! queue max-size-buffers=0 max-size-time=0 max-size-bytes=0
             ! mux.
            alsasrc device="$alsadevice" do-timestamp=true
             ! audio/x-raw,format=S16LE,rate=48000,channels=2
             ! queue max-size-buffers=0 max-size-time=0 max-size-bytes=0
             ! mux.
            matroskamux name=mux
             ! queue max-size-buffers=0 max-size-time=0 max-size-bytes=0
             ! fdsink fd=1
           )
       -c:v libx264 -preset ultrafast -x264opts crf=18:keyint=50:min-keyint=5
       -pix_fmt yuyv422 -c:a flac
       -f matroska file:vhs.mkv

Listing 2 offers a good trade-off. GStreamer captures the VHS output and pipes it to FFmpeg for transcoding (see Table 3). It will encode the raw stream without a noticeable loss of quality and will diminish the output file size by a factor of ten. Because it is CPU-friendly, it allows the capture of the analog data without frame loss. The resulting file is, however, still too big for most users. It is also likely to suffer from audio and video noise, since old analog data is susceptible to acquiring defects because of the media's age. Therefore, the captured video will require postprocessing.