Video on oscilloscope

My friend sent me this video a few days ago:



And I immediately wanted to do the same, but it's a huge project if done properly. It has a ILDA interface/driver, high-speed custom DACs, etc.
So I tried a simple alternative - creating a raster monitor from oscilloscope.

Here is the idea:

• Put oscilloscope to the X-Y mode
• Take a standard composite video (from DVD player in my case)
• Separate horizontal and vertical sync signals from the composite video
• Use these sync signals to generate two saw-tooth signals for raster. Vertical sawtooth (frame signal) scans top-bottom and horizontal sawtooth (line signal) scans left-right.
• With these two signals I should get a raster/white screen on the scope
• Assuming that I'll be fine without gray shades, just black and white image
• Keep a raster running all the time, but when the beam is supposed to be passing over the black area then quickly change the horizontal (or vertical?) signal to the constant voltage. But choose this voltage to cause the beam to jump to the outside of the scope screen. This would form a dark spot.

This sound simple and could be done without microcontrollers, but using just a handful of analog ICs.

The only tricky part could be to separate sync signals from the composite video. I used to repair TVs a long time ago and vaguely remembered that it used to be a specialized IC to do exactly that. Sure enough google remembers it better than me: LM1881 Video Sync Separator

Here is the adapter schematic:

Horizontal sawtooth signal generated by R1, D1, C2 and buffed by the U1. The signal turned on/off by the analog switch U2A.
Vertical sawtooth signal is generated by the R6, D2, C5. It is inverted by the U2B, so it actually becomes an inverted sawtooth. This is because in the oscilloscope world higher voltage causes beam to move up, but TV scans top-bottom.
U4 is a comparator to decide when it is time to turn the beam off (based on the brightness). Actually it's not a "turning the beam off", but instead it disconnects the horizontal sawtooth signal from the scope input, which is equivalent to applying 0V to the input. This causes the beam to move to the left, beyond the scope screen, so the beam is not visible.
The alternative solution would be to use scope's blanking input (z-axis) to turn the beam off, but I thought that it'll be too much like a standard TV.

А few pictures/oscillograms to illustrate how it works. The source video were paused at this frame:

Below I am displaying both channels in standard scope mode (not X-Y).
The bottom sawtooth is vertical channel. I am triggering on the falling edge of this signal (vertical retrace).
The lame picture above the vertical signal is my horizontal trace.
The bright line just above the vertical trace signal is where my horizontal trace parks for the "beam off" period.

The same as above, just zoomed into the highlighted area (head).

And this is the result, playing the same video as the original project (touhou - "bad apple"):



Note: it looks like the video is losing frames around 0:30, but this is from the original video and have nothing to do with my adapter.