As many of you know, scanning video to MIDI file results vary widely.
There's smooth ones and there's ones with tempo jitter or rough chords,
but then sometimes a video will produce a MIDI file scan so accurate
it could've been made with a CIS roll scanning system! It's this kind
of scan that I made today from this video:
https://www.youtube.com/watch?v=I3sUO2Qu-SI
 

I've attached the scanned MIDI file; the only error in the scan file,
as far as I can tell, is the sustain pedal which had a really bad spot
about two-thirds of the way through the roll where a dark spot on the
video formed, triggering the sustain sensor for too long. The scan has
as well been uploaded by me to the IAMMP archive at
http://www.pianorollmusic.org/ 

(I don't know the [copyright] date of the roll so right now it's
blocked but I am suspecting it to be from about 1925 based on where
it falls in the gap between two other scans.)

Piotr Barcz

 [ https://www.mmdigest.com/Attachments/22/06/23/220623_135117_Supertone%204467%20Aren't%20We%20All%20-%20John%20Matthews.mid